U.S. patent application number 16/692043 was filed with the patent office on 2020-06-18 for memory system topologies including a buffer device and an integrated circuit memory device.
The applicant listed for this patent is Rambus Inc.. Invention is credited to Craig Hampel, Ian Shaeffer, Ely Tsern.
Application Number | 20200194052 16/692043 |
Document ID | / |
Family ID | 39734893 |
Filed Date | 2020-06-18 |
![](/patent/app/20200194052/US20200194052A1-20200618-D00000.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00001.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00002.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00003.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00004.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00005.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00006.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00007.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00008.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00009.png)
![](/patent/app/20200194052/US20200194052A1-20200618-D00010.png)
View All Diagrams
United States Patent
Application |
20200194052 |
Kind Code |
A1 |
Shaeffer; Ian ; et
al. |
June 18, 2020 |
Memory System Topologies Including A Buffer Device And An
Integrated Circuit Memory Device
Abstract
Systems, among other embodiments, include topologies (data
and/or control/address information) between an integrated circuit
buffer device (that may be coupled to a master, such as a memory
controller) and a plurality of integrated circuit memory devices.
For example, data may be provided between the plurality of
integrated circuit memory devices and the integrated circuit buffer
device using separate segmented (or point-to-point link) signal
paths in response to control/address information provided from the
integrated circuit buffer device to the plurality of integrated
circuit buffer devices using a single fly-by (or bus) signal path.
An integrated circuit buffer device enables configurable effective
memory organization of the plurality of integrated circuit memory
devices. The memory organization represented by the integrated
circuit buffer device to a memory controller may be different than
the actual memory organization behind or coupled to the integrated
circuit buffer device. The buffer device segments and merges the
data transferred between the memory controller that expects a
particular memory organization and actual memory organization.
Inventors: |
Shaeffer; Ian; (Los Gatos,
CA) ; Tsern; Ely; (Los Altos, CA) ; Hampel;
Craig; (Los Altos, CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Rambus Inc. |
Sunnyvale |
CA |
US |
|
|
Family ID: |
39734893 |
Appl. No.: |
16/692043 |
Filed: |
November 22, 2019 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
16214986 |
Dec 10, 2018 |
10535398 |
|
|
16692043 |
|
|
|
|
15832468 |
Dec 5, 2017 |
10381067 |
|
|
16214986 |
|
|
|
|
15389409 |
Dec 22, 2016 |
9865329 |
|
|
15832468 |
|
|
|
|
14801723 |
Jul 16, 2015 |
9563583 |
|
|
15389409 |
|
|
|
|
14015648 |
Aug 30, 2013 |
9117035 |
|
|
14801723 |
|
|
|
|
13149682 |
May 31, 2011 |
8539152 |
|
|
14015648 |
|
|
|
|
12703521 |
Feb 10, 2010 |
8108607 |
|
|
13149682 |
|
|
|
|
12424442 |
Apr 15, 2009 |
7685364 |
|
|
12703521 |
|
|
|
|
11697572 |
Apr 6, 2007 |
7562271 |
|
|
12424442 |
|
|
|
|
11460899 |
Jul 28, 2006 |
7729151 |
|
|
11697572 |
|
|
|
|
11236401 |
Sep 26, 2005 |
7464225 |
|
|
11460899 |
|
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G11C 11/4096 20130101;
H01L 2224/32145 20130101; G11C 11/4076 20130101; H01L 2225/1005
20130101; G06F 13/4068 20130101; G11C 5/04 20130101; G06F 13/4027
20130101; H01L 2225/1058 20130101; H01L 2924/15331 20130101; H01L
2924/14 20130101; G11C 7/1006 20130101; G11C 5/06 20130101; H01L
25/0657 20130101; G11C 11/4091 20130101; H01L 2924/3011 20130101;
H01L 25/0652 20130101; G11C 7/22 20130101; H01L 2924/15192
20130101; H01L 2225/1023 20130101; G11C 5/025 20130101; H01L
2924/3025 20130101; G11C 11/4093 20130101; G11C 11/4094 20130101;
H01L 25/18 20130101; H01L 24/73 20130101; H01L 2924/15311 20130101;
H01L 2224/73265 20130101; G06F 13/16 20130101; G11C 7/222 20130101;
H01L 2224/48227 20130101; H01L 25/105 20130101; H01L 2924/3011
20130101; H01L 2924/00 20130101; H01L 2924/3025 20130101; H01L
2924/00 20130101; H01L 2224/73265 20130101; H01L 2224/32145
20130101; H01L 2224/48227 20130101; H01L 2924/00012 20130101; H01L
2924/14 20130101; H01L 2924/00 20130101 |
International
Class: |
G11C 11/4093 20060101
G11C011/4093; H01L 25/065 20060101 H01L025/065; H01L 25/10 20060101
H01L025/10; G11C 5/02 20060101 G11C005/02; G11C 5/04 20060101
G11C005/04; G11C 5/06 20060101 G11C005/06; G11C 7/10 20060101
G11C007/10; G11C 7/22 20060101 G11C007/22; G06F 13/40 20060101
G06F013/40; G06F 13/16 20060101 G06F013/16; G11C 11/4076 20060101
G11C011/4076; G11C 11/4091 20060101 G11C011/4091; G11C 11/4094
20060101 G11C011/4094; G11C 11/4096 20060101 G11C011/4096 |
Claims
1. (canceled)
2. A memory controller to control the operation of an external
memory module, the memory controller comprising: a circuit to
receive configuration information from non-volatile storage of the
external memory module, the configuration information representing
presence of a plurality of memory die stacks disposed on the
external memory module; and an interface to transmit, as part of a
memory access operation, a memory access command to specify a
memory access, a chip select signal to select the memory die stacks
for the memory access; and information representing selection of a
particular die within each die stack of the plurality of memory die
stacks.
3. The memory controller of claim 2, wherein the non-volatile
storage comprises a serial presence detect (SPD) register, and
wherein circuit to receive the configuration information is to
access the SPD register.
4. The master of claim 2, wherein the configuration information
comprises values representing timing parameters of the plurality of
memory die stacks, wherein the timing parameters include row access
time, column access time, time between accesses to successive rows,
and time between accesses to successive columns, and wherein the
memory controller transmits memory access command timed according
to the timing parameters.
5. The memory controller of claim 2, wherein the information
representing selection of a particular die within each die stack of
the plurality of die stacks comprises one or more bits, wherein the
one or more bits are to effectuate selection of the particular die
within each die stack of the plurality of memory die stacks with
each memory access command.
6. The memory controller of claim 5, wherein the one or more bits
are included in a row address provided with the memory access
command to each die stack of the plurality of memory die
stacks.
7. The memory controller of claim 2, wherein the interface is to
transmit a row address to activate a row in the particular die
within each die stack of the plurality of memory die stacks,
wherein the information representing selection of the particular
die is provided with the row address.
8. The memory controller of claim 2, wherein the particular die in
each die stack of the plurality of memory die stacks provides a
respective slice of data as part of the memory access
operation.
9. The memory controller of claim 2, wherein the configuration
information specifies a number of memory dies within each die stack
of the plurality of memory die stacks.
10. A method of operation of a memory controller, the memory
controller to control the operation of memory dies disposed on an
external memory module, the method comprising: receiving
configuration information from non-volatile storage of the external
memory module, the configuration information representing presence
of a plurality of memory die stacks disposed on the external memory
module; and as part of a memory access operation, transmitting, a
memory access command to specify a memory access, a chip select
signal to select the memory die stacks for the memory access, and
information representing selection of a particular die within each
die stack of the plurality of memory die stacks.
11. The method of claim 10, wherein the configuration information
specifies a number of memory dies within each die stack of the
plurality of memory die stacks.
12. The method of claim 10, wherein receiving configuration
information is performed during an initialization sequence.
13. The method of claim 10, wherein the configuration information
comprises values representing timing parameters of the plurality of
memory die stacks, wherein transmitting the memory access command
comprises transmitting the memory access command according to the
timing parameters.
14. The method of claim 13, wherein the timing parameters include
row access time, column access time, time between accesses to
successive rows, and time between accesses to successive columns of
the plurality of memory dies.
15. The method of claim 10, wherein the information representing
selection of a particular die within each die stack of the
plurality of memory die stacks comprises one or more bits, wherein
the one or more bits are to effectuate selection of the particular
die within each die stack of the plurality of memory die stacks
with each memory access command.
16. The method of claim 15, further comprising transmitting a row
address to the plurality of memory die stacks, wherein the wherein
the one or more bits are included in the row address.
17. The method of claim 10, further comprising transmitting a row
address to activate a row in the particular die in each stack of
the plurality of memory die stacks, wherein the information
representing selection of the particular die is transmitted with
the row address.
18. The method of claim 10, wherein the particular die in each die
stack of the plurality of memory die stacks provides a respective
slice of data as part of the memory access operation.
19. A memory controller to control the operation of a plurality of
memory die stacks, the memory controller comprising: a circuit to
receive configuration information representing a number of memory
die within each stack of the plurality of memory die stacks; and an
interface to transmit to the plurality of memory die stacks, as
part of a memory access operation, a memory access command to
specify a memory access, a chip select signal to select the memory
die stacks for the memory access, and information representing
selection of a particular die within each die stack of the
plurality of memory die stacks.
20. The memory controller of claim 19, wherein the configuration
information is to be received from non-volatile storage of an
external memory module containing the plurality of memory die
stacks.
21. The memory controller of claim 20, wherein the information
representing selection of a particular die within each die stack of
the plurality of memory die stacks comprises one or more bits,
wherein the one or more bits are to effectuate selection of the
particular die within each die stack of the plurality of memory die
stacks with each memory access command.
Description
RELATED APPLICATIONS
[0001] This application is a continuation of U.S. patent
application Ser. No. 16/214,986, filed on Dec. 10, 2018, which is a
continuation of U.S. patent application Ser. No. 15/832,468 filed
on Dec. 5, 2017 (now U.S. Pat. No. 10,381,067), which is a
continuation of U.S. patent application Ser. No. 15/389,409 filed
on Dec. 22, 2016 (now U.S. Pat. No. 9,865,329), which is a
continuation of U.S. patent application Ser. No. 14/801,723 filed
on Jul. 16, 2015 (now U.S. Pat. No. 9,563,583), which is a
continuation of U.S. patent application Ser. No. 14/015,648 filed
on Aug. 30, 2013 (now U.S. Pat. No. 9,117,035), which is a
continuation of U.S. patent application Ser. No. 13/149,682 filed
on May 31, 2011 (now U.S. Pat. No. 8,539,152), which is a
continuation of U.S. patent application Ser. No. 12/703,521 filed
on Feb. 10, 2010 (now U.S. Pat. No. 8,108,607) which is a
continuation of U.S. patent application Ser. No. 12/424,442 filed
on Apr. 15, 2009 (now U.S. Pat. No. 7,685,364) which is a
divisional of U.S. patent application Ser. No. 11/697,572 filed on
Apr. 6, 2007 (now U.S. Pat. No. 7,562,271) which is a
continuation-in-part of U.S. patent application Ser. No. 11/460,899
filed on Jul. 28, 2006 (now U.S. Pat. No. 7,729,151) which is a
continuation-in-part of U.S. patent application Ser. No. 11/236,401
filed on Sep. 26, 2005 (now U.S. Pat. No. 7,464,225).
FIELD OF THE INVENTION
[0002] The present invention generally relates to integrated
circuit devices, high speed signaling of such devices, memory
devices, and memory systems.
BACKGROUND
[0003] Some contemporary trends predict that processors, such as
general purpose microprocessors and graphics processors, will
continue to increase system memory and data bandwidth requirements.
Using parallelism in applications such as multi-core processor
architectures and multiple graphics pipelines, processors should be
able to drive increases in system bandwidths at rates some predict
will be doubled every three years for the next ten years. There are
several major trends in dynamic random access memory ("DRAM") that
may make it costly and challenging to keep up with increasing data
bandwidth and system memory requirements. For example, transistor
speed relative to feature size improvements in a given DRAM
technology node, and the rising costs of capital investment
required to move DRAM technology to greater memory densities for a
given DRAM die adversely affect the rate at which DRAM technology
can keep pace with the increasing data bandwidth and system
capacity requirements.
BRIEF DESCRIPTION OF THE DRAWINGS
[0004] Embodiments are illustrated by way of example, and not by
way of limitation, in the figures of the accompanying drawings and
in which like reference numerals refer to similar elements and in
which:
[0005] FIG. 1 illustrates a memory module topology including a
plurality of integrated circuit memory devices and a plurality of
integrated circuit buffer devices;
[0006] FIG. 2 illustrates a memory module topology having a split
multi-drop control/address bus;
[0007] FIG. 3 illustrates a memory module topology having a single
multi-drop control/address bus;
[0008] FIG. 4 illustrates a memory module topology that provides
data between each integrated circuit buffer device and a memory
module connector interface;
[0009] FIG. 5 illustrates a memory module topology including a
plurality of integrated circuit memory devices and a plurality of
integrated circuit buffer devices with an integrated circuit buffer
device for control and address information;
[0010] FIG. 6 illustrates termination of a control/address signal
path in a memory module topology of FIG. 5;
[0011] FIG. 7 illustrates termination of data signal paths in a
memory module topology of FIG. 5;
[0012] FIG. 8 illustrates termination of a split control/address
signal path in a memory module topology of FIG. 5;
[0013] FIG. 9A illustrates a top view of a memory module topology
including a plurality of integrated circuit memory devices and a
plurality of integrated circuit buffer devices;
[0014] FIG. 9B illustrates a side view of a memory module topology
including a plurality of integrated circuit memory devices and a
plurality of integrated circuit buffer devices;
[0015] FIG. 9C illustrates a bottom view of a memory module
topology including a plurality of integrated circuit memory devices
and a plurality of integrated circuit buffer devices;
[0016] FIG. 10 is a block diagram illustrating a topology of a
device having a plurality of integrated circuit memory dies and an
integrated circuit buffer die;
[0017] FIG. 11 illustrates a multi-chip package ("MCP") device
having a plurality of integrated circuit memory dies and an
integrated circuit buffer die;
[0018] FIG. 12 illustrates a device having a plurality of
integrated circuit memory dies and a buffer die;
[0019] FIG. 13 illustrates a device having a plurality of
integrated circuit memory devices and a buffer device that are
disposed on a flexible tape;
[0020] FIG. 14 illustrates a device having a plurality of
integrated circuit memory dies and a buffer die that are disposed
side-by-side and housed in a package;
[0021] FIG. 15 illustrates a device having a plurality of
integrated circuit memory dies and a buffer die that are housed in
separate packages and integrated together into a larger
package-on-a-package ("POP") device;
[0022] FIG. 16 illustrates a memory module topology including a
serial presence detect device ("SPD");
[0023] FIG. 17 illustrates a memory module topology with each data
slice having an SPD;
[0024] FIG. 18 is a block diagram of an integrated circuit buffer
die;
[0025] FIG. 19 is a block diagram of a memory device;
[0026] FIGS. 20A-B illustrate signal paths between memory module
interface portions and a plurality of integrated circuit buffer
devices;
[0027] FIGS. 21A-D illustrate memory system point-to-point
topologies including a master and at least one memory module (shown
as buffer 101a) having a plurality of integrated circuit memory
devices;
[0028] FIGS. 22A-C illustrate memory system daisy chain topologies
including a master and at least one memory module having a
plurality of integrated circuit memory devices;
[0029] FIGS. 23A-C and 24A-B illustrate memory system topologies
including a master to provide control/address information to a
plurality of integrated circuit buffer devices;
[0030] FIGS. 25A-B illustrate memory modules having different sized
address spaces, or memory capacity;
[0031] FIGS. 26A-B illustrate a memory system including a master
and two memory modules operating during a first and second mode of
operation (bypass mode);
[0032] FIG. 27 illustrates a memory system including a master and
at least four memory modules;
[0033] FIGS. 28A-B illustrate memory systems including a master and
four memory modules operating during a first mode and second mode
of operation (bypass mode);
[0034] FIG. 29 illustrates a bypass circuit;
[0035] FIGS. 30A-B illustrate timing diagrams for an integrated
circuit buffer device;
[0036] FIG. 31 illustrates a method to levelize memory modules
according to an embodiment;
[0037] FIGS. 32A-E illustrate tree topologies (data and/or
control/address information) between an integrated circuit buffer
device and a plurality of integrated circuit memory devices;
[0038] FIGS. 33A-B illustrate fly-by topologies (data and/or
control/address information) between an integrated circuit buffer
device and a plurality of integrated circuit memory devices;
[0039] FIG. 34 illustrates point-to-point (also known as segmented)
topology (data and/or control/address information) between an
integrated circuit buffer device and a plurality of integrated
circuit memory devices;
[0040] FIG. 35 illustrates an MCP (or system-in-a-package ("SIP")
topology (data and/or control/address information) between an
integrated circuit buffer die and a plurality of integrated circuit
memory dies;
[0041] FIG. 36 is a block diagram of an integrated circuit buffer
device;
[0042] FIGS. 37A-B illustrate timing diagrams of an integrated
circuit buffer device;
[0043] FIG. 38 illustrates a buffer device and a plurality of
integrated circuit memory devices in different ranks;
[0044] FIG. 39 illustrates a system for accessing individual memory
devices that function as respective memory ranks;
[0045] FIG. 40 illustrates a method of operation in an integrated
circuit buffer device.
DETAILED DESCRIPTION
[0046] Systems, among other embodiments, include topologies for
transferring data and/or control/address information between an
integrated circuit buffer device (that may be coupled to a master,
such as a memory controller) and a plurality of integrated circuit
memory devices. For example, data may be provided between the
plurality of integrated circuit memory devices and the integrated
circuit buffer device using separate segmented (or point-to-point
link) signal paths in response to control/address information
provided from the integrated circuit buffer device to the plurality
of integrated circuit buffer devices using a single fly-by (or bus)
signal path. Other topology types may include forked, star, fly-by,
segmented and topologies used in SIP or MCP embodiments.
[0047] An integrated circuit buffer device enables configurable
effective memory organization of a plurality of integrated circuit
memory devices. The memory organization represented by the
integrated circuit buffer device to a memory controller may be
different than the actual memory organization behind or coupled to
the integrated circuit buffer device. For example, control/address
information may be provided to the buffer device from a memory
controller that expects a memory organization having a
predetermined number of memory devices and memory banks as well as
page size and peak bandwidth, but the actual memory organization
coupled to the buffer device is different. The buffer device
segments and/or merges the data transferred between the memory
controller that expects a particular memory organization and the
actual memory organization. The integrated circuit buffer device
may merge read data from separate memory devices into a stream of
read data. Likewise, the integrated circuit memory device may
segment a write data into write data portions that are stored on a
plurality of memory devices.
[0048] An integrated circuit buffer device may include data path,
address translation, data path router, command decode and control
(or register set) circuits. The buffer device also includes an
interface that may be configured into at least three different
segmentation modes: 1) Four 4-bit interfaces (4.times.4), 2) Two
4-bit interfaces (2.times.4) or 3) Two 8-bit interfaces
(2.times.8). The different configurations allow flexibility in
memory module or memory stack configurations. The buffer device may
also include a pattern generator and internal memory array circuit
to emulate storing and retrieving data from the plurality of
integrated circuit memory devices.
[0049] The buffer device may increase memory system performance by;
for example, eliminating a "time bubble" or idle time for a signal
path (bus) turnaround between memory transactions to different
ranks of integrated circuit memory devices coupled to segmented
data signal paths. A memory rank may also include a single
integrated circuit memory device. Eliminating the need for the
memory controller to track memory rank access and inserting time
bubbles may reduce memory controller complexity. Memory modules or
memory rank capacity may be expanded using segmented data signal
paths without decreasing bandwidth caused by bubble time insertion.
Memory modules may include more memory devices or dies while still
emulating a single rank memory module.
[0050] According to embodiments, a system includes a master device
and a first memory module having a plurality of integrated circuit
memory devices and a plurality of integrated circuit buffer devices
that operate in first and second modes of operation (bypass mode).
In a first mode of operation, a first memory module provides read
data from the plurality of integrated circuit memory devices (via
an integrated circuit buffer device) on a first signal path to the
master and a second memory module simultaneously provides read data
from its plurality of integrated circuit memory devices (via
another integrated circuit buffer device on the second module) on a
third signal path coupled to the master device. In a second mode of
operation, the first memory module provides first read data from
its plurality of integrated circuit memory devices (via the
integrated circuit buffer device) on the first signal path and
second read data from its plurality of integrated circuit memory
devices (via the integrated circuit buffer device) on a second
signal path that is coupled to a second memory module. An
integrated circuit buffer device in the second memory module then
bypasses the second read data from the second signal path and
provides the second read data on a third signal path coupled to the
master device. The first memory module may have a larger address
space or capacity, such as twice as large, as compared to the
second memory module.
[0051] Similarly, write data may be provided from the master device
to the first and second memory modules during the first and second
modes of operation.
[0052] According to embodiments, the second memory module includes
a bypass circuit (such as in the integrated circuit buffer device,
interface or in continuity memory module) to transfer the second
read data from the second signal path to the third signal path. The
bypass circuit may include a jumper, signal trace and/or
semiconductor device. The bypass circuit may also include delay
circuits for adding delay in outputting the read data (or
levelizing) from a memory module.
[0053] According to embodiments, a system includes a master device
and at least four memory modules wherein at least two memory
modules have different capacities than the other two memory
modules. The four memory modules are coupled to a plurality of
signal paths. The system may operate in a bypass mode in which one
or more memory modules use a bypass circuit to provide read data
from at least one larger capacity memory module to a master
device.
[0054] According to embodiments, a system includes a master and a
plurality of memory modules that may be disposed in a variety of
topologies, such as point-to-point or daisy chain topologies.
Memory modules may include a plurality of integrated circuit buffer
devices that are coupled using a variety of topologies to receive
control information, such as dedicated, fly-by, stub, serpentine or
tree topologies, singly or in combination.
[0055] According to embodiments, a method determines a mode of
operation of a system including a master and a plurality of memory
modules. In a bypass mode of operation, delays are provided to read
data from at least one memory module to levelize or ensure that
read data from different capacity memory modules using different
signal paths arrive at the master at approximately the same
time.
[0056] According to embodiments, a memory module includes a
plurality of signal paths that provide data to a memory module
connector from a plurality of respective integrated circuit buffer
devices (or dies) that access the data from an associated plurality
of integrated circuit memory devices (or dies). In a specific
embodiment, each integrated circuit buffer device is also coupled
to a bussed signal path that provides control and/or address
information that specifies an access to at least one integrated
circuit memory device associated with the respective integrated
circuit buffer device.
[0057] According to embodiments, a memory module connector includes
a control/address interface portion and a data interface portion. A
control/address bus couples a plurality of integrated circuit
buffer devices to the control/address interface portion. A
plurality of data signal paths couple the plurality of respective
integrated circuit buffer devices to the data interface portion.
Each integrated circuit buffer device includes 1) an interface to
couple to at least one integrated circuit memory device, 2) an
interface to couple to the control/address bus and 3) an interface
to couple to a data signal path in the plurality of data signal
paths.
[0058] According to embodiments, a memory module may include a
non-volatile memory location, for example using an electrically
erasable programmable read only memory ("EEPROM") (also known as a
Serial Presence Detect ("SPD") device), to store information
regarding parameters and configuration of the memory module. In
embodiments, at least one integrated circuit buffer device accesses
information stored in the SPD device.
[0059] In a package embodiment, a package houses an integrated
circuit buffer die and the plurality of integrated circuit memory
dies. In the package, a plurality of signal paths transfer data
(read and/or write data) between the integrated circuit buffer die
and the plurality of integrated circuit memory dies. The integrated
circuit buffer die provides control signals from an interface of
the package to the plurality of integrated circuit memory dies.
Data stored in memory arrays of the plurality of integrated circuit
memory dies is provided to a signal path disposed on the memory
module via the integrated circuit buffer die in response to the
control signals. In an embodiment, the package may be a multichip
package ("MCP"). In an embodiment, the plurality of integrated
circuit memory dies may be housed in common or separate packages.
In an embodiment described below, the memory module may include a
series of integrated circuit dies (i.e., memory die and buffer die)
stacked on top of one another and coupled via a signal path.
[0060] As described herein, an integrated circuit buffer device is
also referred to as a buffer or buffer device. Likewise, an
integrated circuit memory device is also referred to as a memory
device. A master device is also referred to as a master.
[0061] In an embodiment, an integrated circuit memory device is
distinguished from a memory die in that a memory die is a
monolithic integrated circuit formed from semiconductor materials
for storing and/or retrieving data or other memory functions,
whereas an integrated circuit memory device is a memory die having
at least some form of packaging or interface that allows the memory
die to be accessed.
[0062] Likewise in an embodiment, an integrated circuit buffer
device is distinguished from a buffer die in that a buffer die is a
monolithic integrated circuit formed from semiconductor materials
and performs at least one or more buffer functions described
herein, whereas an integrated circuit buffer device is a buffer die
having at least some form of packaging or interface that allows
communication with the buffer die.
[0063] In the embodiments described in more detail below, FIGS. 1-8
illustrate control/address and data signal path topologies
including a plurality of integrated circuit memory devices (or
dies) and a plurality of integrated circuit buffer devices (or
dies) situated on a memory module. FIGS. 10, 18, and 19 also
illustrate signal path topologies including integrated circuit
memory devices (or dies) and integrated circuit buffer devices (or
dies) situated on a memory module as well as the operation of an
integrated circuit buffer device (or die) and memory device (or
die) in embodiments among other things. FIGS. 21A-D, 22A-C, 23A-C
and 24A-B illustrate system topologies. FIGS. 26A-B, 28A-B and 31
illustrate operating a memory system in a first and second mode of
operation (bypass mode). FIGS. 32A-E, 33A-B, 34 and 35 illustrate
topologies between an integrated circuit buffer device and a
plurality of integrated circuit memory devices. FIG. 36 is a block
diagram of an integrated circuit buffer device and FIGS. 37A-B
illustrates timing diagrams of an integrated circuit buffer device.
FIGS. 38 and 39 illustrate a buffer device and a plurality of
integrated circuit memory devices in different memory ranks. FIG.
40 illustrates a method of operation in an integrated circuit
buffer device.
[0064] FIG. 1 illustrates a memory module topology including a
plurality of integrated circuit memory devices and a plurality of
associated integrated circuit buffer devices. In an embodiment, a
memory module 100 includes a plurality of buffer devices 100a-d
coupled to a common address/control signal path 121. Each buffer
device of the plurality of buffer devices 100a-d provides access to
a plurality of respective integrated circuit memory devices 101a-d
via signal paths 102a-d and 103. In an embodiment, respective data
slices a-d are formed by one of buffers 100a-d and sets of memory
devices 101a-d. Each of buffer devices 100a-d is coupled to a
respective set of signal paths 120a-d, that transfer data (read and
write data) between the buffer devices 100a-d and a memory module
connector interface. In an embodiment, mask information is
transferred to buffer devices 100a-d from a memory module connector
interface using signal paths 120a-d, respectively.
[0065] In an embodiment, a data slice is a portion of the memory
module data signal path (or bus) that is coupled to the respective
integrated circuit buffer device. The data slice may include the
full data path or portions of data paths to and from a single
memory device disposed on the memory module.
[0066] Integrated circuit memory devices may be considered as a
common class of integrated circuit devices that have a plurality of
storage cells, collectively referred to as a memory array. A memory
device stores data (which may be retrieved) associated with a
particular address provided, for example, as part of a write or
read command. Examples of types of memory devices include dynamic
random access memory ("DRAM"), including single and double data
rate synchronous DRAM, static random access memory ("SRAM"), and
flash memory. A memory device typically includes request or command
decode and array access logic that, among other functions, decodes
request and address information, and controls memory transfers
between a memory array and signal path. A memory device may include
a transmitter circuit to output data, for example, synchronously
with respect to rising and falling edges of a clock signal, (e.g.,
in a double data rate type of memory device). Similarly, the memory
device may include a receiver circuit to receive data, for example,
synchronously with respect to rising and falling edges of a clock
signal or outputs data with a temporal relationship to a clock
signal in an embodiment. A receiver circuit also may be included to
receive control information synchronously with respect to rising
and falling edges of a clock signal. In an embodiment, strobe
signals may accompany the data propagating to or from a memory
device and that data may be captured by a device (e.g., memory
device or buffer, or controller) using the strobe signal.
[0067] In an embodiment, an integrated circuit buffer device is an
integrated circuit that acts as an interface between a memory
module connector interface and at least one integrated circuit
memory device. In embodiments, the buffer device may store and/or
route data, control information, address information and/or a clock
signal to at least one integrated circuit memory device that may be
housed in a common or separate package. In an embodiment, the
buffer isolates, routes and/or translates data, control information
and a clock signal, singly or in combination, between a plurality
of memory devices and a memory module connector interface. An
embodiment of a memory module connector interface is described
below and shown in FIGS. 9A-C.
[0068] At least one signal path 121, as shown in FIG. 1, disposed
on memory module 100, transfers control and/or address
(control/address) information between at least one of the buffer
devices 100a-d and a memory module connector interface in various
embodiments. In an embodiment, signal path 121 is a multi-drop bus.
As illustrated in FIGS. 2-8 and described below, alternate
topologies for transferring control/address information, data and
clock signals between one or more buffer devices 100a-d and a
memory module connector interface may be used in alternate
embodiments. For example, a split multi-drop control/address bus,
segmented multi-drop control/address bus, and point-to-point and/or
daisy chain topologies for a data bus may be employed.
[0069] In an embodiment, clock signals and/or clock information may
be transferred on at least one signal line in signal path 121.
These clock signal(s) provide one or more clock signals having a
known frequency and/or phase. In an embodiment, a clock signal is
synchronized with or travels along side the control/address
information. In an embodiment, an edge of the clock signal has a
temporal relationship with an edge of a control/address signal
representing the control/address information. In an embodiment, a
clock signal is generated by a clock source, master device (e.g.,
controller device) and/or buffer device.
[0070] In an embodiment, a clock signal and/or clock information
may be transferred on at least one signal line in respective signal
paths 120a-d. Buffer devices 100a-d may receive and/or transmit a
clock signal with data on signal paths 120a-b. In an embodiment,
write data is provided to buffer devices 100a-d on signal paths
120a-d and a clock signal is provided on signal paths 120a-d along
side write data. In an embodiment, a clock signal (such as a
clock-to-master ("CTM")) is provided from buffer devices 100a-d on
signal paths 120a-d along side read data on signal paths 120a-d. In
an embodiment, a clock signal is synchronized with or travels along
side the write and/or read data. An edge of the clock signal has a
temporal relationship or is aligned with an edge of a data signal
representing write and/or read data. Clock information can be
embedded in data, eliminating the use of separate clock signals
along with the data signals.
[0071] In an embodiment, a read, write and/or bidirectional strobe
signal may be transferred on at least one signal line in respective
signal paths 120a-d. Buffer devices 100a-d may receive and/or
transmit a strobe signal with data on signal paths 120a-b. In an
embodiment, write data is provided to buffer devices 100a-d on
signal paths 120a-d and a strobe signal is provided on signal paths
120a-dalong side write data. In an embodiment, a strobe signal is
provided from buffer devices 100a-d on signal paths 120a-d along
side read data on signal paths 120a-d. In an embodiment, a strobe
signal is synchronized with or travels along side the write and/or
read data. An edge of the strobe signal has a temporal relationship
or is aligned with an edge of a data signal representing write
and/or read data.
[0072] In an embodiment, addresses (for example, row and/or column
addresses) for accessing particular memory locations in a
particular integrated circuit memory device and/or commands are
provided on signal path 121 from a memory module connector
interface. In an embodiment, a command relates to a memory
operation of a particular integrated circuit memory device. For
example, a command may include a write command to store write data
at a particular memory location in a particular integrated circuit
memory device and/or a read command for retrieving read data stored
at a particular memory location from a particular integrated
circuit memory device. Also, multiple memory devices in different
data slices can be accessed simultaneously. In embodiments, a
command may include row commands, column commands such as read or
write, mask information, precharge and/or sense command. In an
embodiment, control information is transferred on signal path 121
over a common set of lines in the form of a time multiplexed packet
where particular fields in the packet are used for including
command operation codes and/or addresses. Likewise, packets of read
data may be transferred from integrated circuit memory devices via
buffers 100a-d on respective signal paths 120a-d to a memory module
connector interface. In an embodiment, a packet represents one or
more signals asserted at particular bit windows (or a time
interval) for asserting a signal on particular signal lines.
[0073] In an embodiment, chip select information may be transferred
on one or more signal lines in signal path 121. In an embodiment,
chip select information may be one or more chip select signals on
respective signal lines having predetermined voltage values or
states (or logic values) that select and enable operation of a
"chip" or integrated circuit memory device/buffer device.
[0074] In embodiments, memory module 100 communicates (via a memory
module connector interface) with a master device (e.g., a processor
or controller).
[0075] FIG. 2 illustrates an embodiment of a memory module topology
having a split multi-drop control/address/clock bus. In particular,
memory module 200 includes a split multi-drop control/address bus
221 coupled to buffers 100a-d and a memory module connector
interface. With reference to FIG. 2, a first portion of bus 221 is
terminated by termination 230 and a second portion of bus 221 is
terminated by termination 231. In an embodiment, the impedance of
termination 230 matches the impedance of the first portion of bus
221 (Z0) coupled to buffers 100c-d and the impedance of termination
231 matches the impedance of the second portion of bus 221 (Z1)
coupled to buffers 100a-b. In an embodiment, impedance Z0 equals
impedance Z1. In embodiments, terminations 230 and 231, singly or
in combination, are disposed on memory module 100, buffer devices
100a and 100d or packages used to house buffer devices 100a and
100d.
[0076] FIG. 3 illustrates a memory module topology having a single
multi-drop control/address/clock bus terminated by termination 330.
In an embodiment, the impedance of termination 330 matches the
impedance of signal path 121 (or control/address/clock bus). In
embodiments, termination 330, singly or in combination, is disposed
on memory module 300 or on buffer device 100d.
[0077] FIG. 4 illustrates a memory module topology that provides
data between each integrated circuit buffer device and a memory
module connector interface. In an embodiment, each signal path
120a-d is terminated by an associated termination 420a-d,
respectively. In an embodiment, terminations 420a-d have respective
impedances that match the impedance Z0 of each of the signal paths
120a-d. In embodiments, terminations 420a-d, singly or in
combination, are disposed on memory module 400, each of buffer
devices 100a-d or packages used to house buffer devices 100a-d.
[0078] Referring to FIG. 1, a control/address signal rate ratio of
signal path 121 to signal path 103 may be 2:1 (or other multiples
such as 4:1, 8:1, etc.) so that a memory module connector interface
is able to operate as fast as specified while memory devices 101a-d
may operate at half (quarter, eighth, etc) the control/address
signaling rate so that relatively lower cost memory devices may be
used. Similarly, a data signal rate of one of signal paths 120a-d
to one of signal paths 102a-d may be 2:1 (or other multiple such as
4:1, 8:1, etc) so that a memory module connector interface is able
to operate as fast as specified while memory devices 101a-d may
operate at half (quarter, eighth, etc.) the data signaling rate so
that relatively lower cost memory devices may be used.
[0079] FIG. 5 illustrates a memory module topology including a
plurality of integrated circuit memory devices and a plurality of
integrated circuit buffer devices with an integrated circuit buffer
device 501 for control, address and/or clock information. Memory
module 500 is similar to memory module 100 except that buffer
device 501 is coupled to signal paths 121 and 121a-b. Buffer device
501 outputs control, address and/or clock information to buffer
devices 100a-b on signal path 121a and to buffer devices 100c-d on
signal path 121b. In an embodiment buffer device 501 copies
control, address and/or clock information received on signal path
121 and repeats the control, address and/or clock information on
signal paths 121a-b. In an embodiment, buffer device 501 is a
clocked buffer device that provides a temporal relationship with
control and address information provided on signal paths 121a-b. In
an embodiment, signal paths 121a-b include at least one signal line
to provide a clock signal and/or clock information. In an
embodiment, buffer device 501 includes a clock circuit 1870 as
shown in FIG. 18. In an embodiment, buffer device 501 receives
control information, such as a packet request, that specifies an
access to at least one of the integrated circuit memory devices
101a-d and outputs a corresponding control signal (on signal path
121a and/or 121b) to the specified integrated circuit memory
device.
[0080] FIG. 6 illustrates a memory module topology similar to that
illustrated in FIG. 5 except that a termination 601 is coupled to
signal path 121 on memory module 600. In an embodiment, the
impedance of termination 601 matches the impedance Z0 of signal
path 121. In embodiments, termination 601 is disposed on memory
module 600 or buffer device 501 or a package used to house buffer
device 501.
[0081] FIG. 7 illustrates a memory module topology that provides
data to and/or from each integrated circuit buffer device and
terminations coupled to signal paths. In an embodiment, each signal
path 120a-d is terminated by associated terminations 701a-d,
respectively. In an embodiment, terminations 701a-d have respective
impedances that match the impedance Z0 of each of the signal paths
120a-d. In embodiments, terminations 701a-d, singly or in
combination, are disposed on memory module 700, buffer devices
100a-d or packages used to house buffer devices 100a-d.
[0082] FIG. 8 illustrates a memory module topology having a split
multi-drop signal path between a buffer device for control, address
and/or clock information and the plurality of buffer devices. In
particular, memory module 800 includes a split multi-drop
control/address bus 121a-b coupled to buffers 100a-d and a buffer
device 501. In an embodiment, a first portion of bus 121a is
terminated by termination 801 and a second portion of bus 121b is
terminated by termination 802. In an embodiment, the impedance of
termination 801 matches the impedance of the first leg (Z0) and the
impedance of termination 802 matches the impedance of the second
leg (Z1). In an embodiment, impedance Z0 equals impedance Z1. In
embodiments, terminations 801 and 802, singly or in combination,
are disposed on memory module 800, buffer devices 100a and 100d or
packages used to house buffer devices 100a and 100d.
[0083] Referring to FIG. 5, a control/address signal rate ratio of
signal path 121 to signal path 121a (or 121b) to signal path 103
may be 2:1:1 (or other multiples such as 4:1:1, 8:1:1, etc.) so
that other multi-drop bus topology embodiments using signal paths
121a (or 121b) and signal path 103 do not have to necessarily
operate as high a signal rate as an embodiment that uses signal
path 121 as shown in FIG. 1. Also like FIG. 1, a control/address
signal rate ratio of signal path 121 to signal path 103 may be 2:1
(or other multiples such as 4:1, 8:1, etc.) so that a memory module
connector interface is able to operate as fast as specified while
memory devices 101a-d may operate at half (or quarter, eighth,
etc.) the control/address signaling rate so that relatively lower
cost memory devices may be used. Similarly, a data signal rate of
one of signal paths 120a-d to one of signal paths 102a-d may be 2:1
(or other multiple such as 4:1, 8:1, etc.) so that a memory module
connector interface is able to operate as fast as the specified
signaling rate while memory devices 101a-d may operate at half (or
quarter, eighth, etc.) the data signaling rate so that relatively
lower cost memory devices may be used.
[0084] FIG. 9A illustrates a top view of a memory module topology
including a plurality of integrated circuit memory devices and a
plurality of integrated circuit buffer devices coupled to a
connector interface. In an embodiment, memory module 900 includes a
substrate 910 having a standard dual in-line memory module ("DIMM")
form factor or other module form factor standards, such as small
outline DIMM ("SO-DIMM") and very low profile DIMM ("VLP-DIMM"). In
alternate embodiments, substrate 910 may be, but is not limited to,
a wafer, printed circuit board ("PCB"), package substrate like BT
epoxy, flex, motherboard, daughterboard or backplane, singly or in
combination.
[0085] In an embodiment, memory module 900 includes pairs of memory
devices 101a-b and buffer devices 100a-d disposed on a first side
of substrate 910. In alternate embodiments, more or less memory
devices and buffer devices are used. In an embodiment, pairs of
memory devices 101c-d are also disposed on a second side of memory
module 900 as shown in a side and bottom view of memory module 900
in FIGS. 9B and 9C. In an embodiment, each memory device and buffer
device are housed in separate packages. In alternate embodiments,
memory devices and buffer devices may be housed in MCP package
embodiments described herein.
[0086] Memory module 900 includes connector interface 920 that has
different interface portions for transferring data and
control/address/clock signals. For example, a first side of memory
module 900 includes connector interface portions 920a-d used to
transfer data signals and a connector interface portion 930a used
to transfer control/address signals. In an embodiment, connector
interface portion 930a also transfers a clock signal and/or clock
information. In an embodiment, a second side of memory module 900
including connector interface portions 920e-h are used to transfer
data signals and a connector interface portion 930b is used to
transfer control/address signals. In an embodiment, connector
interface portion 930b also transfers a clock signal and/or clock
information.
[0087] In an embodiment, connector interface 920 is disposed on an
edge of substrate 910. In an embodiment, a memory module 900 is
inserted into a socket 940 disposed on substrate 950. In an
embodiment, substrate 950 is a main board or PCB with signal paths
960a-b for transferring signals on substrate 950. In an embodiment,
signal paths 960a and 960b are signal traces or wires. In an
embodiment, signal paths 960a and 960b are coupled to other sockets
disposed on substrate 950 that may have another memory module
inserted and/or coupled to a master.
[0088] In an embodiment, connector interface portions include at
least one contact or conducting element, such as a metal surface,
for inputting and/or outputting an electrical signal. In alternate
embodiments, a contact may be in the form of any one of or a
combination of a ball, socket, surface, signal trace, wire, a
positively or negatively doped semiconductor region and/or pin,
singly or in combination. In an embodiment, a connector interface
as described herein, such as connector interface 920, is not
limited to physically separable interfaces where a male connector
or interface engages a female connector (or socket 940) or
interface. A connector interface also includes any type of physical
interface or connection, such as an interface used in a
system-in-a-package ("SIP") where leads, solder balls or
connections from a memory module are soldered to a circuit
board.
[0089] In an alternate embodiment, memory module 900 is included in
an embedded memory subsystem, such as one in a computer graphics
card, video game console or a printer. In an alternate embodiment,
memory module 900 is situated in a personal computer or server.
[0090] In an embodiment, a master communicates with memory modules
illustrated in FIGS. 1-9 and 16-17. A master may transmit and/or
receive signals to and from the memory modules illustrated in FIGS.
1-9 and 16-17. A master may be a memory controller, peer device or
slave device. In embodiments, a master is a memory controller,
which may be an integrated circuit device that contains other
interfaces or functionality, for example, a Northbridge chip of a
chipset. A master may be integrated on a microprocessor or a
graphics processor unit ("GPU") or visual processor unit ("VPU"). A
master may be implemented as a field programmable gate array
("FPGA"). Memory modules, signal paths, and a master may be
included in various systems or subsystems such as personal
computers, graphics cards, set-top boxes, cable modems, cell
phones, game consoles, digital television sets (for example, high
definition television ("HDTV")), fax machines, cable modems,
digital versatile disc ("DVD") players or network routers.
[0091] In an embodiment, a master, memory modules and signal paths
are in one or more integrated monolithic circuits disposed in a
common package or separate packages.
[0092] FIG. 10 is a block diagram illustrating an embodiment of a
device 1000 having a plurality of integrated circuit memory devices
101a-d and a buffer 100a. Here, data (read and/or write) may be
transferred between the plurality of integrated circuit memory
devices 101a-d and buffer 100a on a signal path 1006 (data). Signal
path 1006 is a signal path situated internal to device 1000 and
corresponds to signal paths 1113a-d and 1114 shown in FIG. 11.
Signal path 1006 is a bus for providing bidirectional data signals
between a plurality of integrated circuit memory devices 101a-d and
buffer 100a. An example of bidirectional data signals includes
signals traveling from one or more of integrated circuit memory
devices 101a-d to buffer 100a and also signals traveling from
buffer 100a to one or more of integrated circuit memory devices
101a-d. Signal path 1005 is a signal path internal to device 1000
and corresponds to signal paths 1116a-d and 1117 shown in FIG. 11.
Signal path 1005 is a bus for providing unidirectional
control/address/clock signals from a buffer 100a to a plurality of
integrated circuit memory devices 101a-d. In an example of a
unidirectional bus, signals travel in only one direction, i.e., in
this case, from only buffer 100a to one or more of integrated
circuit memory devices 101a-d. Signal path 1005 includes individual
control signal lines, for example, a row address strobe line,
column address strobe line, chip select line, etc., and address
signal lines. Signal path 1005 may include a fly-by clock line to
transfer a clock signal from buffer 100a to integrated circuit
memory devices 101a-d. Signal path 1005 may transfer a clock signal
from one or more integrated circuit memory devices 101a-d to buffer
100a.
[0093] In an embodiment, buffer 100a communicates with a serial
presence detect ("SPD") device to store and retrieve parameters and
configuration information regarding device 1000 and/or memory
module 900. In an embodiment, an SPD 1002 is a non-volatile storage
device. Signal path 1004 couples SPD 1002 to buffer 100a. In an
embodiment, signal path 1004 is an internal signal path for
providing bidirectional signals between SPD 1002 and buffer
100a.
[0094] In an embodiment, SPD 1002 is an EEPROM device. However,
other types of SPD 1002 are possible, including but not limited to
a manual jumper or switch settings, such as pull-up or pull-down
resistor networks tied to a particular logic level (high or low),
which may change state when a memory module is added or removed
from a system.
[0095] In an embodiment, SPD 1002 is a memory device that includes
registers that stores configuration information that can be easily
changed via software during system operation, allowing a high
degree of flexibility, and making configuration operations that are
transparent to an end user.
[0096] In an embodiment illustrated in FIG. 18, functionality of
the SPD mentioned above may be integrated into buffer device 100a
using a register set, such as configuration register set 1881.
Referring to FIG. 18, SPD logic and interface 1820c may be
preconfigured with information pertaining to the buffer and memory
devices connected to the buffer, or may store information
pertaining to only one of the memory devices or the buffer device
100a. Control inputs to the buffer may determine when a storage
node within the register set will sample the information to preload
or preconfigure the SPD logic and interface 1820c. The term
register may apply either to a single-bit-wide register or
multi-bit-wide register.
[0097] In an embodiment illustrated by FIG. 10, SPD 1002 stores
information relating to configuration information of memory module
900 or a memory system. For example, configuration information may
include repair and redundancy information to repair a defective
memory device, defective memory cells or peripheral circuits on a
memory device, and/or signal path. In an embodiment, SPD
configuration information includes memory module population
topology, such as a number, a position and a type of memory device
in a package and/or on a memory module, or rank, if any. SPD
configuration information may include an amount of memory capacity
of one or more memory modules and/or timing information to levelize
signals between memory modules and a master device in a memory
system. In an embodiment, SPD configuration information includes a
serialization ratio for interfaces in a buffer and/or information
regarding configuring the width of a buffer. In an embodiment, SPD
configuration information includes a first value that represents
the desired width of buffer device 100a or includes multiple values
that represent the range of possible widths of the buffer device
100a, and a second value that represents the desired width of
interface 1820b as illustrated in FIG. 18.
[0098] In an embodiment, SPD configuration information includes
timing information or parameters for accessing memory devices, such
as a time to access a row or the memory device, a time to access a
column of the memory device, a time between a row access and a
column access, a time between a row access and a precharge
operation, a time between a row sense applied to a first bank of a
memory array and a row sense applied to a second bank of the memory
array and/or a time between a precharge operation applied to a
first bank in a memory array and a precharge operation applied to a
second bank of the memory array.
[0099] In an embodiment, the stored timing information may be
expressed in terms of time units where a table of values maps
specific time units to specific binary codes. During an
initialization or calibration sequence, a master or a buffer may
read SPD configuration information and determine the proper timing
information for one or more memory devices. For example, a master
may also read information representing the clock frequency of a
clock signal from an SPD 1002, and divide the retrieved timing
information by a clock period of a clock signal. (The clock period
of the clock signal is the reciprocal of the clock frequency of the
clock signal). Any remainder resulting from this division may be
rounded up to the next whole number of clock cycles of the clock
signal.
[0100] Signal paths 120a and 121, as shown in FIG. 10, are coupled
to buffer 100a. In an embodiment, signal path 121 transfers
unidirectional control/address/clock signals to buffer 100a. In an
embodiment, signal path 120a transfers bidirectional or
unidirectional data signals to and from buffer 100a. Other
interconnect and external connect topologies may also be used for
device 1000 in alternate embodiments. For example, buffer 100a may
be coupled to a single multi-drop control bus, a split multi-drop
control bus, or a segmented multi-drop bus.
[0101] In an embodiment, device 1000 has two separate power
sources. Power source V1 supplies power to one or more memory
devices (memory devices 101a-d) on memory module 900. Power source
V2 supplies power to one or more buffers (buffer 100a) on memory
module 900. In an embodiment, the buffer 100a has internal power
regulation circuits to supply power to the memory devices
101a-d.
[0102] FIG. 11 illustrates a device 1100 including a plurality of
integrated circuit memory dies 1101a-d and a buffer die 1100a
housed in or upon a common package 1110 according to embodiments.
As described herein in other embodiments and illustrated in FIGS.
12-15 and 35, a plurality of integrated circuit memory dies 1101a-d
and buffer die 1100a are disposed in multiple package type
embodiments. For example, a plurality of integrated circuit memory
dies 1101a-d and a buffer die 1100a may be stacked, on a flexible
tape, side-by-side or positioned in separate packages on a device
substrate. Buffer die 1100a is used to provide signals, including
control/address/clock information and data, between a plurality of
integrated circuit memory dies 1101a-d and a device interface 1111
that includes contacts 1104a-f. In an embodiment, one or more
contacts 1104a-f is similar to contacts of connector interface 920.
Contacts 1104a-f are used to couple device 1100 to substrate 910,
and in particular to signal paths 120a and 121, of memory module
100 in an embodiment. Device interface 1111 also includes signal
paths 1118 and 1115 to transfer signals between contacts 1104a-f
and buffer 100a via buffer interface 1103. Signals are then
transferred between a plurality of memory dies 1101a-d and buffer
die 1100a via buffer interface 1103 and signal paths 1117 (disposed
in device interface 1111) and 1116a-d as well as signal paths 1114
(disposed in device interface 1111) and 1113a-d. In an embodiment,
spacers 1102a-c are positioned between integrated circuit memory
dies 1101a-d. In an embodiment, spacers 1102a-c are positioned to
dissipate heat. Similarly, buffer die 1100a is disposed away from a
plurality of integrated circuit memory dies 1101a-d to alleviate
heat dissipation near the memory devices. In an embodiment, signal
paths are coupled to each other and integrated circuit memory dies
1101a-d by a solder ball or solder structure.
[0103] FIG. 12 illustrates a stacked package device 1200 having a
package 1210 containing a plurality of integrated circuit memory
dies 1101a-d and a separate package 1290 having a buffer die 1100a.
Both packages 1210 and 1290 are stacked and housed to make device
1200. In an embodiment, a plurality of integrated circuit memory
dies has separate packages and is stacked on package 1290. Device
1200 has similar components illustrated in FIG. 11. Buffer die
1100a communicates with a plurality of integrated circuit memory
dies 1101a-d as described herein. Device 1200 has memory dies
1101a-d stacked upon buffer die 1100a and separated by contacts
1201a-d. In an embodiment, contacts 1201a-d are solder balls that
couple signal paths 1117 and 1114 to signal paths 1202 and 1203
that are coupled to buffer interface 1103.
[0104] FIG. 13 illustrates devices 1300 and 1301 having a plurality
of integrated circuit memory devices 101a-b (101a-c in device 1301)
and a buffer device 100a that are disposed on a flexible tape 1302
according to embodiments. Buffer device 100a communicates with a
plurality of integrated circuit memory devices as described herein.
Signal path 1305 disposed on or in flexible tape 1302 transfers
signals between a plurality of integrated circuit memory devices
101a-c and buffer 100a. Contacts, such as a grid array of balls
1304, couple each integrated circuit memory device in a plurality
of integrated circuit memory devices 101a-c and a buffer 100a to
signal path 1305 in flexible tape 1302 in an embodiment. Adhesive
1303 may be used to couple a plurality of integrated circuit memory
devices 101a-c to each other and to a buffer 100a in an embodiment.
Device 1300 and 1301 are disposed in common package in an
embodiment.
[0105] FIG. 14 illustrates a device 1400 having a plurality of
integrated circuit memory dies 1101a-dand 1401a-d and a buffer die
1100a that are disposed side-by-side and housed in a package 1410.
Device 1400 has similar components illustrated in FIG. 11. Buffer
die 1100a communicates with a plurality of integrated circuit
memory dies 1101a-d and 1401a-d as described herein. In an
embodiment, a plurality of integrated circuit memory dies 1101a-d
and 1401a-d and a buffer die 1100a are disposed side-by-side on a
substrate 1450 that is coupled to device interface 1411. A
plurality of integrated circuit memory dies 1401a-d is separated by
spacers 1402a-c. In an embodiment, a single integrated circuit
memory die 1101d and a single integrated circuit memory die 1401d
are disposed side-by-side with buffer die 1100a. Device interface
1411 includes contacts 1104a-f. Signals are transferred between
buffer interface 1103 and contacts 1104a-f by signal paths 1418 and
1415. Signals are transferred between buffer interface 1103 and
signal paths 1116a-d (or integrated circuit memory dies 1101a-d) by
signal path 1417. Similarly, signals are transferred between buffer
interface 1103 and signal paths 1113a-d (or integrated circuit
memory dies 1401a-d) by signal path 1414.
[0106] FIG. 15 illustrates a device 1500 having a plurality of
integrated circuit memory dies 1101a-b and a buffer die 1100a that
are housed in separate packages 1501, 1505 and 1520, respectively.
Device 1500 has similar components illustrated in FIG. 11. Buffer
die 1100a communicates with integrated circuit memory dies 1101a-b
as described herein. Integrated circuit memory dies 1101a-b and a
buffer die 1100a are disposed on substrate 1530 that includes
signal paths 1504, 1509, 1515 and 1518.
[0107] Integrated circuit memory die 1101a includes memory
interface 1507 having contacts 1508. Integrated circuit memory die
1101b includes memory interface 1503 having contacts 1541. Buffer
die 1100a includes a buffer interface 1103 having contacts 1560.
Signals are transferred between buffer interface 1103 and contacts
1104a-f by signal paths 1515 and 1518. Signals are transferred
between buffer interface 1103 and integrated circuit memory die
1101a by signal path 1509 via memory interface 1507 and contacts
1508. Similarly, signals are transferred between buffer interface
1103 and integrated circuit memory die 1101b by signal path 1504
via memory interface 1503 and contacts 1541. As described herein,
device 1500 is coupled to a memory module 900 via contacts
1104a-f.
[0108] FIG. 16 illustrates a memory module having an SPD 1603
according to an embodiment. Memory module 1610 includes a plurality
of integrated circuit memory devices (or dies) and buffer devices
(or dies) disposed on substrate 930 along with SPD 1603. FIG. 16
illustrates a memory module 1610 having a single SPD 1603 that can
be accessed by each buffer device 100a-b positioned on substrate
930. Signal path 1601 allows access to SPD 1603 from connector
interface 920 and one or more buffers 100a-b. In an embodiment,
signal path 1601 is a bus. SPD 1603 may have configuration and/or
parameter information written to or read by a master by way of
connector interface 920 and signal path 1601. Likewise, buffers
100a-b may write to or read from SPD 1603 via signal path 1601.
[0109] FIG. 17 illustrates a memory module 1710 with each device
1711a-b or data slice a-b having an associated SPD 1720a-b, buffer
device (or die) 100a-b and at least one integrated circuit memory
device 101a (or die) according to an embodiment. The plurality of
buffers 100a-b and associated plurality of SPDs 1720a-b are
disposed on substrate 930. Configuration and/or parameter
information is accessed from SPDs 1720a-b using signal path 1701,
which is coupled, to connector interface 920 and each SPD 1720a-b.
In particular, signal path 1701 couples SPD 1720a-b of device
1711a-b to connector interface 920. In an embodiment, signal path
1701 is a bus. In an alternate embodiment, signal path 1701 couples
SPD 1720a and SPD 1720b in a daisy chain or serial topology. In an
embodiment, one or more buffer devices 100a-b of devices 1711a-b
may access (read and/or write) respective SPDs 1720a-b. Likewise, a
master may access (read and/or write) respective SPDs 1720a-b using
signal path 1701. In an embodiment, configuration and/or parameter
information is transferred using a header field or other identifier
so that SPDs coupled in a daisy chain may forward the SPD
information to the intended destination SPD.
[0110] FIG. 18 illustrates a block diagram of a buffer device 100a
(or die, such as buffer die 1100a) according to embodiments. Buffer
100a includes buffer interface 1103a, interfaces 1820a-c,
redundancy and repair circuit 1883, multiplexer 1830, request and
address logic circuit 1840, data cache and tags circuit 1860,
computations circuit 1865, configuration register set 1881, and
clock circuit 1870, singly or in combination.
[0111] In a memory read operation embodiment, buffer 100a receives
control information (including address information) that may be in
a packet format from a master on signal path 121 and in response,
transmits corresponding signals to one or more, or all of memory
devices 101a-d on one or more signal paths 1005. One or more of
memory devices 101a-d may respond by transmitting data to buffer
100a which receives the data via one or more signal paths 1006 and
in response, transmits corresponding signals to a master (or other
buffer). A master transmits the control information via one or more
signal paths 121 and receives the data via one or more signal paths
120a.
[0112] By bundling control and address information in packets,
protocols required to communicate to memory devices 101a-d are
independent of the physical control/address interface
implementation.
[0113] In a memory write operation embodiment, buffer 100a receives
control information (including address information) that may be in
a packet format from a master on signal path 121 and receives the
write data for one or more memory devices 101a-d that may be in a
packet format from a master on signal path 120a. Buffer 100a then
transmits corresponding signals to one or more, or all of memory
devices 101a-d on one or more signal paths 1006 so that the write
data may be stored.
[0114] A master transmits the control/address/clock information via
one or more signal paths 121 and transmits the write data via one
or more signal paths 120a.
[0115] In an embodiment, simultaneous write and/or read operations
may occur for different memory devices in memory devices
101a-d.
[0116] In an embodiment, control information that is provided to
buffer 100a causes one or more memory operations (such as write
and/or read operations) of one or more memory devices 100a-d, while
the same control information may be provided to buffer 100b which
causes the same memory operations of one or more memory devices
100a-d associated with buffer 100b. In another embodiment, the same
control information may be provided to buffer 100a and buffer 100b,
yet different memory operations occur for the one or more memory
devices 100a-d associated with each buffer 100a-b.
[0117] In an embodiment, buffer interface 1103a couples signal
paths 121 and 120a to buffer 100a as shown in FIG. 10. In an
embodiment, buffer interface 1103a corresponds to buffer interface
1103 shown in FIGS. 11, 12, 14 and 15. In an embodiment, buffer
interface 1103a includes at least one transceiver 1875 (i.e.
transmit and receive circuit) coupled to signal path 120a to
transmit and receive data and at least one receiver circuit 1892
coupled to signal path 121 to receive control/address/clock
information. In an embodiment, signal paths 121 and 120a include
point-to-point links. Buffer interface 1103a includes a port having
at least one transceiver 1875 that connects to a point-to-point
link. In an embodiment, a point-to-point link comprises one or a
plurality of signal lines, each signal line having no more than two
transceiver connection points. One of the two transceiver
connection points is included on buffer interface 1103a. Buffer
interface 1103a may include additional ports to couple additional
point-to-point links between buffer 100a and other buffer devices
on other devices and/or memory modules. These additional ports may
be employed to expand memory capacity as is described in more
detail below. Buffer 100a may function as a transceiver between a
point-to-point link and other point-to-point links. In an
embodiment, buffer interface 1103a includes a repeater circuit 1899
to repeat data, control information and/or a clock signal. In an
embodiment, buffer interface 1103a includes a bypass circuit 1898
to transfer signals between connector interface portions.
[0118] In an embodiment, termination 1880 is disposed on buffer
100a and is connected to transceiver 1875 and signal path 120a. In
this embodiment, transceiver 1875 includes an output driver and a
receiver. Termination 1880 may dissipate signal energy reflected
(i.e., a voltage reflection) from transceiver 1875. Termination
1880, as well as other termination described herein, may be a
resistor or capacitor or inductor, singly or a series/parallel
combination thereof. In alternate embodiments, termination 1880 may
be external to buffer 100a. For example, termination 1880 may be
disposed on a substrate 910 of a memory module 900 or on a package
used to house buffer 100a.
[0119] Interface 1820a includes at least one transmitter circuit
1893 coupled to signal path 1005 to transmit control/address/clock
information to one or more memory devices. In an embodiment,
interface 1820a includes a transceiver that may transfer
control/address/clock information between buffers disposed on a
common memory module or different memory modules.
[0120] Interface 1820b includes a transceiver 1894 coupled to
signal path 1006 to transfer data between buffer 100a and one or
more memory devices 101a-d as illustrated in FIG. 10. SPD logic and
interface 1820c includes a transceiver 1896 coupled to signal path
1004 to transfer configuration and/or parameter information between
buffer 100a and an SPD 1002 as illustrated in FIG. 10. In an
embodiment, interface 1820c is used to transfer configuration
and/or parameter information as illustrated in FIGS. 16 and 17.
[0121] According to an embodiment, multiplexer 1830 may perform
bandwidth-concentrating operations between buffer interface 1103a
and interface 1820b as well as route data from an appropriate
source (i.e. target a subset of data from memory devices, internal
data, cache or write buffer). The concept of bandwidth
concentration involves combining the (smaller) bandwidth of each
data path coupled to a memory device in a multiple data signal path
embodiment to match the (higher) overall bandwidth utilized by
buffer interface 1103a. In an embodiment, multiplexing and
demultiplexing of throughput between the multiple signal paths that
may be coupled to interface 1820b and buffer interface 1103a is
used. In an embodiment, buffer 101a utilizes the combined bandwidth
of multiple data paths coupled to interface 1820b to match the
bandwidth of interface buffer interface 1103a.
[0122] In an embodiment, data cache and tags circuit 1860 (or cache
1860) may improve memory access time by providing storage of most
frequently referenced data and associated tag addresses with lower
access latency characteristics than those of the plurality of
memory devices. In an embodiment, cache 1860 includes a write
buffer that may improve interfacing efficiency by utilizing
available data transport windows over an external signal path to
receive write data and address/mask information. Once received,
this information is temporarily stored in a write buffer until it
is ready to be transferred to at least one memory device over
interface 1820b.
[0123] Computations circuit 1865 may include a processor or
controller unit, a compression/decompression engine, etc., to
further enhance the performance and/or functionality of buffer
100a. In an embodiment, computations circuit 1865 controls the
transfer of control/address/clock information and data between
buffer interface 1103a and interfaces 1820a-c.
[0124] Clock circuit 1870 may include a clock generator circuit
(e.g., Direct Rambus.RTM. Clock Generator), which may be
incorporated onto buffer 101a and thus may eliminate the need for a
separate clock generating device.
[0125] In an alternate embodiment, clock circuit 1870 include clock
alignment circuits for phase or delay adjusting an internal clock
signal with respect to an external clock signal, such as a phase
lock loop ("PLL") circuit or delay lock loop ("DLL") circuit. Clock
alignment circuits may utilize an external clock from an existing
clock generator, or an internal clock generator to provide an
internal clock, to generate internal synchronizing clock signals
having a predetermined temporal relationship with received and
transmitted data and/or control information.
[0126] In an embodiment, clock circuit 1870 receives a first clock
signal having a first frequency via signal path 121 and generates a
second clock signal (via interface 1820a) to memory device 101a
using the first clock signal and also generates a third clock
signal (via interface 1820a) to memory device 101b using the first
clock signal. In an embodiment, the second and third clock signals
have a predetermined temporal (phase or delay) relationship with
the first clock signal.
[0127] In an embodiment, a transmit circuit (such as in
transceivers 1875, 1896 and 1894 shown in FIG. 18) transmits a
differential signal that includes encoded clock information and a
receiver circuit (such as in transceiver 1875, 1896 and 1894)
receives a differential signal that includes encoded clock
information. In this embodiment, a clock and data recovery circuit
(such as clock circuit 1870) is included to extract the clock
information encoded with the data received by the receiver circuit.
Likewise, clock information may be encoded with data transmitted by
the transmit circuit. For example, clock information may be encoded
onto a data signal, by ensuring that a minimum number of signal
transitions occur in a given number of data bits.
[0128] In an embodiment, a transceiver 1875 transmits and receives
a first type of signal (for example, a signal having specified
voltage levels and timing), while transceivers 1894 (and/or
transmit circuit 1893) transmits and receives a second different
type of signal. For example, transceiver 1875 may transmit and
receive signals for a DDR2 memory device and transceivers 1894 may
transmit and receive signals for a DDR3 memory device.
[0129] In an embodiment, the control information and/or data that
is provided to buffer 100a (by way of signal paths 121 and 120) may
be in a different protocol format or have different protocol
features than the control information and/or data provided to one
or more memory devices 100a-d from buffer 100a. Logic (for example
computation circuit 1865) in buffer 100a performs this protocol
translation between the control information and/or data received
and transmitted. A combination of the different
electrical/signaling and control/data protocol constitute an
interface standard in an embodiment. Buffer 100a can function as a
translator between different interface standards--one for the
memory module interface (for example connector interface 920) and
another for one or more memory devices 100a-d. For example, one
memory module interface standard may require reading a particular
register in a particular memory device disposed on the memory
module. Yet, a memory module may be populated with memory devices
that do not include the register required by the memory module
interface standard. In an embodiment, buffer 100a may emulate the
register required by the memory module interface standard and thus
allow for the use of memory devices 100a-d that operate under a
different interface standard. This buffer functionality, combined
with the module topology and architecture, enables a memory module
to be socket compatible with one interface standard, while using
memory devices with a different interface standard.
[0130] In an embodiment, buffer 100a includes a redundancy and
repair circuit 1883 to test and repair the functionality of memory
cells, rows or banks of a memory device, entire memory devices (or
periphery circuits) and/or signal paths between buffer 100a and
memory devices 101a-d. In an embodiment, redundancy and repair
circuit 1883 periodically, during a calibration operation and/or
during initialization, tests one or more of memory devices 101a-d
by writing a predetermined plurality of values to a storage
location in a selected memory device (for example, using
transceiver 1894 and a look-up table storing the predetermined
values) using a selected data path and then reading back the stored
predetermined plurality of values from the selected memory device
using the selected data path. In an embodiment, when the values
read from the storage location of the selected memory device do not
match the values written to the storage location, redundancy and
repair circuit 1883 eliminates access by buffer 100a to the
selected memory device and/or selected signal path. In an
embodiment, a different signal path to a different memory device
may be selected and this testing function may be performed again.
If selecting the different signal path results in an accurate
comparison of read predetermined values to the predetermined values
in redundancy and repair circuit 1883 (or a pass of the test), the
different memory address to a different memory location, within or
to another memory device, is selected or mapped thereafter.
Accordingly, future write and/or read operations to the defective
memory location will not occur.
[0131] In an embodiment, any multiplexed combination of control
information (including address information) and data intended for
memory devices 101a-d coupled with buffer 100a is received via
buffer interface 1103a, which may, for example extract the address
and control information from the data. For example, control
information and address information may be decoded and separated
from multiplexed data on signal path 120a and provided on signal
path 1895 to request and address logic circuit 1840 from buffer
interface 1103a. The data may then be provided to configurable
serialization/deserialization circuit 1891. Request and address
logic circuit 1840 generates one or more control signals to
transmitter circuit 1893.
[0132] Interfaces 1820a and 1820b include programmable features in
embodiments. A number of control signal lines and/or data signal
lines between buffer 100a and memory devices 101a-d are
programmable in order to accommodate different numbers of memory
devices. Thus, more dedicated control signal lines are available
with an increased number of memory devices. Using programmable
dedicated control lines and/or data lines avoids any possible load
issues that may occur when using a bus to transfer control signals
between memory devices and a buffer 100a. In another embodiment,
additional data strobe signals for each byte of each memory device
may be programmed at interface 1820b to accommodate different types
of memory devices, such as legacy memory devices that require such
a signal. In still a further embodiment, interfaces 1820a and 1820b
are programmable to access different memory device widths. For
example, interfaces 1820a and 1820b may be programmed to connect to
16 "x4" width memory devices, 8 "x8" width memory devices or 4
"x16" width memory devices. Likewise, buffer interface 1103a has a
programmable width for signal path 120a.
[0133] Configurable serialization/deserialization circuit 1891
performs serialization and deserialization functions depending upon
a stored serialization ratio. As a memory device access width is
reduced from its maximum value, memory device access granularity
(measured in quanta of data) is commensurately reduced, and an
access interleaving or multiplexing scheme may be employed to
ensure that all storage locations within memory devices 101a-d can
be accessed. The number of signal paths 1006 may be increased or
decreased as the memory device access width changes. Signal path
1006 may be subdivided into several addressable subsets. The
address of the transaction will determine which target subset of
signal path 1006 will be utilized for the data transfer portion of
the transaction. In addition, the number of transceiver,
transmitter and/or receiver circuits included in interfaces 1820a
and 1820b that are employed to communicate with one or more memory
devices 101a-d may be configured based on the desired serialization
ratio. Typically, configuration of the transceivers may be
effectuated by enabling or disabling how many transceivers are
active in a given transfer between one or more memory devices
101a-d and buffer interface 1103a. In an embodiment, a data rate of
transferring data at buffer interface 1103a is a multiple or ratio
of a data rate of transferring data on one or more signal paths
1006 coupled to memory devices 101a-d.
[0134] Buffer 100a provides a high degree of system flexibility.
New interface standards of memory devices may be phased in to
operate with a master or a memory system that supports older
interface standards by modifying buffer 100a. In an embodiment, a
memory module may be inserted using an older memory module
interface or socket, while newer generation memory devices may be
disposed on the memory module. Backward compatibility with existing
generations of memory devices may be preserved. Similarly, new
generations of masters, or controllers, may be phased in which
exploit features of new generations of memory devices while
retaining backward compatibility with existing generations of
memory devices. Similarly, different types of memory devices that
have different costs, power requirements and access times may be
included in a single common package for specific applications.
[0135] FIG. 19 illustrates an integrated circuit memory device 1900
(or a memory die) in an embodiment. Integrated circuit memory
device 1900 corresponds to one or more integrated circuit memory
devices 101a-d in embodiments. Integrated circuit memory device
1900 includes a memory core 1900b and a memory interface 1900a.
Signal paths 1950a-b, 1951a-b, 1952 and 1953 are coupled to memory
interface 1900a. Signal paths 1950a-b transfer read and write data.
Signal paths 1951a-btransfer address information, such as a row
address and a column address in packets, respectively. Signal path
1952 transfers control information. Signal path 1953 transfers one
or more clock signals. In an embodiment, signal paths 1950a-b
correspond to signal path 120a shown in FIG. 10 and signal paths
1951a-b, 1952 and 1953 correspond to signal path 121 in FIG.
10.
[0136] Memory interface 1900a includes at least one transmitter
and/or receiver for transferring signals between memory device 1900
and signal paths 1950a-b, 1951a-b, 1952 and 1953. Write
demultiplexer ("demux") 1920 and read multiplexer ("mux") 1922 are
coupled to signal path 1950a, while write demux 1921 and read mux
1923 are coupled to signal path 1950b. Write demux 1920-21 provide
write data from signal paths 1950a-b to memory core 1900b (in
particular sense amplifiers 0-2a and 0-2b). Read mux 1922-23
provide read data from memory core 1900b to signal paths 1950a-b
(in particular sense amplifiers Na and Nb).
[0137] Demux and row packet decoder 1910 is coupled to signal path
1951a and Demux and column packet decoder 1913 is coupled to signal
path 1951b. Demux and row packet decoder 1910 decodes a packet and
provides a row address to row decoder 1914. Demux and Column packet
decoder 1913 provides a column address and mask information to
column and mask decoder 1915.
[0138] Control registers 1911 are coupled to signal path 1952 and
provide control signals to row decoder 1914 and column and mask
decoder 1915 in response to register values.
[0139] A clock circuit is coupled to signal path 1953 to provide a
transmit clock signal TCLK and a receive clock signal RCLK in
response to one or more clock signals transferred on signal path
1953. In an embodiment, write demux 1920 and 1921 provide write
data from signal paths 1950a-b to memory core 1900b in response to
an edge of receive clock signal RCLK. In an embodiment, read mux
1922 and 1923 provide read data from memory core 1900b to signal
paths 1950a-b in response to an edge of a transmit clock signal
TCLK. In an embodiment, the clock circuit generates a clock signal
on signal path 1953 (to a buffer device) that has a temporal
relationship with read data that are output on signal paths
1950a-b.
[0140] Row decoder 1914 and column and mask decoder 1915 provide
control signals to memory core 1900b. For example, data stored in a
plurality of storage cells in a memory bank is sensed using sense
amplifiers in response to a row command. A row to be sensed is
identified by a row address provided to row decoder 1914 from demux
and row packet decoder 1910. A subset of the data sensed by a sense
amplifier is selected in response to a column address (and possible
mask information) provided by demux and column packet decoder
1913.
[0141] A memory bank in memory banks 0-N of memory core 1900b
includes a memory array having a two dimensional array of storage
cells. In embodiments, memory banks 0-N include storage cells that
may be DRAM cells, SRAM cells, FLASH cells, ferroelectric RAM
("FRAM") cells, magnetoresistive or magnetic RAM ("MRAM") cells, or
other equivalent types of memory storage cells. In an embodiment,
integrated circuit memory device 1900 is a DDR integrated circuit
memory device or later generation memory device (e.g., DDR2 or
DDR3). In an alternate embodiment, integrated circuit memory device
1900 is an XDR.TM. DRAM integrated circuit memory device or Direct
Rambus.RTM. DRAM ("DRDRAM") memory device. In an embodiment,
integrated circuit memory device 1900 includes different types of
memory devices having different types of storage cells housed in a
common package.
[0142] FIGS. 20A-B illustrate signal paths between memory module
interface portions and a plurality of integrated circuit buffer
devices. In particular, FIG. 20A illustrates how each buffer device
100a-d has signal paths for data signals coupled to each connector
interface portion 920a-h. In an embodiment, FIGS. 20A-B illustrate
signal paths between buffer devices and connector interfaces of
memory module 900 that include a plurality of memory devices as
shown in FIGS. 9A-C. For example, FIG. 20B which shows an expanded
section of FIG. 20A, illustrates how data signal paths 2003 and
2004 provide data signals between connector interface portions 920a
and 920e and buffer device 100a. FIG. 20A also illustrates how
signal paths for control/address signals, such as control/address
signal paths 2001 and 2002, couple connector interface portions
930a and 930b to buffer devices 100a-d. In an embodiment, each
signal path 2001 and 2002 is a multi-drop bus as shown in FIG.
1.
[0143] FIGS. 21A-D illustrate memory system point-to-point
topologies including a master 2101 and at least one memory module
having a plurality of integrated circuit memory devices (The
plurality of memory devices on respective memory modules are not
illustrated in FIGS. 21A-D, 22A-C, 23A-C and 24A-B for clarity). In
an embodiment, FIGS. 21A-D, 22A-C, 23A-C and 24A-B illustrate
signal paths between memory modules, such as memory module 900 as
shown in FIGS. 9A-C, and other memory modules and/or masters. FIGS.
21A-D illustrate expanding memory capacity and bandwidth as well as
different configurations. In particular, master 2101 is coupled to
interfaces (such as sockets) 2102 and 2103 by signal paths 2120,
2121a-b, 2122 and 2123 in Dynamic Point-to-Point ("DPP") system
2100a. In an embodiment, master 2101, interfaces 2102 and 2103 as
well as signal paths 2120, 2121a-b, 2122 and 2123 are disposed on a
substrate, such as a printed circuit board ("PCB"). In an
embodiment, memory modules may be inserted and/or removed
(unpopulated) from interfaces 2102 and 2103. In an embodiment,
signal paths 2120, 2121a-b, 2122 and 2123 are signal traces on a
PCB. In an embodiment, signal paths 2120 and 2121a-b provide data
between data signal paths on a memory module, such as signal paths
120a and 120b shown in FIG. 1, and master 2101. In an embodiment,
signal paths 2122 and 2123 provide control/address information to
the memory modules (via interfaces 2102 and 2103 and in particular
connector interface portions 930b of the memory modules) from
master 2101. In particular, control/address information is provided
from signal paths 2122 and 2123 to a signal path on the memory
modules, such as signal path 121 shown in FIG. 1.
[0144] FIG. 21A illustrates a DPP system 2100a that simultaneously
accesses two buffer devices in memory modules coupled to interfaces
2102 and 2103. In response to control and address information
provided on signal paths 2122 and 2123 from master 2101, the two
buffers 101a output data simultaneously from connector interface
portions 920a and 920e, respectively, onto signal paths 2120 and
2121a, that are coupled to master 2101. In an embodiment, signal
paths 2120 and 2121a are point-to-point links. In an embodiment, a
point-to-point link includes one or a plurality of signal lines,
each signal line generally having two transceiver connection
points, each transceiver connection point coupled to a transmitter
circuit, receiver circuit or transceiver circuit. For example, a
point-to-point link may include a transmitter circuit coupled at or
near one end and a receiver circuit coupled at or near the other
end. The point-to-point link may be synonymous and interchangeable
with a point-to-point connection or a point-to-point coupling.
[0145] In an embodiment, the number of transceiver points along a
signal line may distinguish between a point-to-point link and a
bus. For example, a point-to-point link generally includes only two
transceiver connection points while a bus generally includes more
than two transceiver points. In some instances a point to point
link can be mixed with bussed signal lines, where the bussed single
lines may be used to provide sideband functionality such as
maintenance, initialization or test.
[0146] Several embodiments of point-to-point links include a
plurality of link topologies, signaling, clocking and signal path
types. Embodiments having different link architectures include
simultaneous bi-directional links, time-multiplexed bi-directional
links and multiple unidirectional links. Voltage or current mode
signaling may be employed in any of these link topologies.
[0147] FIG. 21B illustrates a DPP with Continuity Module system
2100b for accessing a buffer device 101a in a memory module coupled
to interface 2103 while a continuity memory module 2105 is coupled
to interface 2102. In an embodiment, master 2101 outputs a single
set of control/address information on signal paths 2122 and 2123.
Data is output from connector interfaces 920a and 920e of the
memory module coupled to interface 2103 in response to the single
set of control/address information. Data is provided to master 2101
on signal path 2120 via signal path 2121b and a bypass circuit in
continuity memory module 2105. The bypass circuit passes the data
from connector interface portion 920e to connector interface
portion 920a in continuity memory module 2105. Data is also
provided to master 2101 by signal path 2121a.
[0148] FIG. 21C illustrates a DPP bypass system 2100c similar to
system 2100b except that a buffer device 101a (rather than
continuity memory module 2105) in a memory module includes a bypass
circuit for passing the data from connector interface portion 920e
to connector interface portion 920a of the memory module inserted
in interface 2102.
[0149] FIG. 21D illustrates a DPP bypass system 2100d similar to
system 2100c except that data is accessed from buffer device 101a
of the memory module coupled to interface 2102 and buffer device
101a of the memory module coupled to interface 2103 includes a
bypass circuit for passing the data from connector interface
portion 920a to connector interface portion 920e.
[0150] In an embodiment, a clock signal or clock information is
provided on signal paths 2122 and 2123, on a separate signal path
from a clock source or master 2101, or along the data signal paths
2121a-b.
[0151] FIGS. 22A-C illustrate memory system daisy chain topologies
including a master 2101 and at least one memory module having a
plurality of integrated circuit memory devices. In particular,
FIGS. 22A-C illustrate how half of the bandwidth, as compared to
system 2100a-d, is obtained when accessing a single memory module
in an embodiment. FIG. 22A illustrates a Daisy Chain system 2200a
that includes a buffer 101a in a memory module coupled to interface
2103 that provides data (by way of connector interface portion
920e) on signal path 2121a in response to a single set of
control/address information output by master 2101 onto signal paths
2122 and 2123. No module is coupled to interface 2102.
[0152] FIG. 22B illustrates a Daisy Chain system 2200b that is
similar to system 2200a except a memory module is coupled to
interface 2102.
[0153] FIG. 22C illustrates a Daisy Chain system 2200c similar to
system 2200b except that data accessed from a buffer device 101a in
a memory module is coupled to interface 2102 rather than interface
2103. Buffer device 101a in a memory module coupled to interface
2103 provides a bypass circuit to allow data to be received at
interface portion 920a and output at interface portion 920e of the
memory module coupled to interface 2103. Data is thus passed from
data path 2121b to data path 2121a and ultimately to master
2101.
[0154] FIGS. 23A-C and 24A-B illustrate memory system topologies
including a master to provide control/address information to a
plurality of integrated circuit buffer devices. In particular, FIG.
23A illustrates a Dedicated/Fly-by system 2300a that includes a
master 2101 that provides control/address information to memory
modules 2301a and 2301b (in particular to integrated circuit buffer
devices 101a-d on each memory module) by signal paths 2311 and
2310, respectively. In an embodiment, signal paths 2310 and 2311
are separate and carry control/address information for each
respective memory module. In an embodiment, signal path 2311 does
not pass through or include a signal path in memory module 2301b.
In an embodiment, signal path 2311 does not pass through or include
an interface, such as a socket, used for memory module 2301b. The
double headed arrow in FIGS. 23A-C, 24A-B and 25A-B illustrate the
data information (read and write data) transferred on separate data
paths between memory modules 2301a-b (and in particular from buffer
devices) and master 2101. In an embodiment, a clock signal or clock
information is provided on signal paths 2310 and 2311, on a
separate signal path from a clock source or master 2101, or along
the data signal paths.
[0155] Signal path 2311 is terminated by termination 2350a and
signal path 2310 is terminated by termination 2350b. In an
embodiment, the impedance of termination 2350a matches the
impedance of a portion of the signal path 2311 (multi-drop bus
2320a) on memory module 2310a, (Z0) and the impedance of
termination 2350b approximately matches the impedance of a portion
of the signal path 2310 (multi-drip bus 2320b) on memory module
2301b (Z1). In an embodiment, impedance Z0 approximately equals
impedance Z1. In embodiments, terminations 2350a and 2350b, singly
or in combination, are disposed on memory module, buffer devices or
packages used to house buffer devices. FIG. 23B illustrates a
Stub/Fly-by system 2300b similar to system 2300a except that a
single signal path 2320 provides control/address information from
master 2101 to memory modules 2301a and 2301b (in particular to
integrated circuit buffer devices 101a-d on each memory module). In
an embodiment, memory modules 2301a and 2301b include
stubs/internal signal paths (multi-drop bus) 2320a-b coupled to a
single common signal path 2320 that are disposed on memory modules
2301a-b. In an embodiment, a portion of signal path 2320 passes
through or includes an interface, such as a socket, used for memory
module 2301b. Memory modules 2301a and 2301b are terminated similar
to system 2300a.
[0156] FIG. 23C illustrates a Serpentine system 2300c similar to
system 2300a except that a single signal path 2320 provides
control/address information from master 2101 to memory modules
2301a and 2301b (in particular to integrated circuit buffer devices
101a-d on each memory module) without using stubs on respective
memory modules as illustrated in FIG. 23B. In an embodiment, a
single signal path 2330 couples master 2101 to memory modules 2301a
and 2301b. In an embodiment signal path 2330 includes a first
external signal path portion between master 2101 and memory module
2301b; a second signal path portion disposed on the memory module
2301b and coupled to the first signal path portion as well as to
respective buffer devices 101a-d; a third external signal path
portion 2331 coupled to the second signal path portion and also
coupled to memory module 2301a; and a fourth signal path portion
disposed on the memory module 2301a and coupled to the third signal
path portion 2331 as well as to respective buffer devices 101a-d on
memory module 2301a. Termination 2350a, in an embodiment, is not
disposed on memory module 2301a in order to ensure that memory
modules are interchangeable. Termination 2350a may be disposed on a
PCB or elsewhere in a system.
[0157] FIG. 24A illustrates a Dedicated/Tree system 2400a similar
to system 2300a except that memory modules 2401a-b include buffer
devices 101a-d that are coupled by way of a tree structure/topology
signal path 2413. A tree structure/topology may also be referred to
as a "forked," "T" or "hybrid T" topology. In particular, memory
module 2401a is coupled to signal path 2311 by signal path 2413a
disposed on memory module 2401a that then branches in to signal
paths 2413b and 2413c. Signal path 2413b then is coupled to buffer
devices 101a and 101b by branches or signal paths 2413d and 2413e.
Signal path 2413c, likewise, is coupled to buffer devices 101c and
101d by branches or signal paths 2413f and 2413g. In an embodiment,
memory module 2401b has a similar tree structure signal path 2413
to couple buffer devices 101a-d to signal path 2310.
[0158] FIG. 24B illustrates a Stub/Tree system 2400b similar to
system 2400a shown in FIG. 24A that includes tree structure signal
path 2413 in memory modules 2401a-b. System 2400b illustrates
signal path 2320 including stubs/ signal paths 2320a and 2320b that
couple master 2101 to memory modules 2401a and 2401b, respectively.
Stub/signal path 2320a is coupled to signal path 2413a disposed on
memory module 2401a and stub/signal path 2320b is coupled to signal
path 2413a disposed on memory module 2401b.
[0159] In embodiments, termination may be disposed on buffers
101a-d, memory modules 2401a-b and/or elsewhere in a system, such
as on a PCB.
[0160] FIGS. 25A-B illustrate memory modules having different
memory capacity or different sized address spaces. In particular,
memory module address space 2501 on a first memory module is larger
than memory module address space 2502 on a second memory module. In
an embodiment, memory module address space 2501 is twice as large
as memory module address space 2502. For example, memory module
address space 2501 may store 2 gigabyte (GB) of information and
memory module address space 2502 may store 1 GB of information.
Increasing the number or density of integrated circuit memory
devices disposed on a memory module may increase address space.
[0161] FIG. 25A illustrates how half (or portion) of the available
signal path width, for example half of a bus width, is used to
access the first half of memory module address space 2501
(overlapping address space) while the other half of the available
signal path width is used to access memory module address space
2502.
[0162] FIG. 25B illustrates how a larger capacity memory module is
able to use a full signal path by accessing a first half (or
portion) of the available signal path width coupled directly to the
larger capacity memory module and by way of accessing a second half
(or portion) of the available signal path width coupled to the
smaller capacity memory module using bypassing through the smaller
capacity memory module. FIGS. 26-29 illustrate how non-overlap
address space of a larger memory module may be accessed in various
embodiments.
[0163] FIGS. 26A-B illustrate a system 2600 to access different
sized/capacity (address space) memory modules during different
modes of operation, a first mode of operation and a second mode of
operation (or bypass mode). System 2600 includes a master 2101
coupled to memory module 2601 by signal path 2610 and memory module
2602 by signal path 2612. Memory modules 2601 and 2602 are coupled
by signal path 2611. In an embodiment, memory modules 2601 and 2602
represent memory modules including integrated circuit memory
devices and buffer devices as described herein. In an embodiment,
memory module 2601 has a larger address space than memory module
2602. In an embodiment, signal paths 2610-2612 are point-to-point
links that provide read/write data. In embodiments,
control/address/clock information is provided on separate signal
paths as described herein. Memory modules 2601 and 2602 may include
bypass circuits 2630a-b.
[0164] In a first mode of operation (or a non-bypass mode)
illustrated in FIG. 26A, read data 2601a (stored in an overlapping
address space) is provided on signal path 2610 to master 2101 from
memory module 2601 in response to control/address information
provided by master 2101 to memory module 2601. Similarly, read data
2602a (stored in an overlapping address space) is provided on
signal path 2612 to master 2101 from memory module 2602 in response
to control/address information provided by master 2101 to memory
module 2602. In the first mode of operation, signal path 2611 is
not used.
[0165] In a second mode of operation (or a bypass mode) illustrated
in FIG. 26B, read data 2601b (stored in a non-overlapping address
space of memory module 2601) is provided on signal path 2610 to
master 2101 from memory module 2601 in response to control/address
information provided by master 2101 to memory module 2601. Read
data 2601c (stored in a non-overlapping address space of memory
module 2601) is provided on signal path 2611 to memory module 2602
in response to control/address information provided by master 2101
to memory module 2601. Bypass circuit 2630b then provides read data
2601c to signal path 2612 and eventually to master 2101.
[0166] Write data from master 2101 may be provided to memory
modules 2601 and 2602 similar to how read data is obtained during a
first and second mode of operation.
[0167] In embodiments, modes of operation are determined in
response to a control signal from master 2101, or other circuit or
in response to reading configuration information stored in a
separate storage circuit in a device, such as an SPD device or
register on the buffer or controller device, disposed on system
2600. Modes of operation may be determined at initialization,
periodically or during calibration of system 2600.
[0168] In embodiments, bypass circuits 2630a-b (as well as bypass
circuits 2630c-d shown in FIG. 27) correspond to bypass circuit
2900 as described below and shown in FIG. 29 and/or bypass circuit
1898 shown in FIG. 18. In embodiments, these bypass circuits can be
incorporated on the buffer devices on the module.
[0169] FIG. 27 illustrates a system 2700 including master 2101
coupled to at least four memory modules 2701-2704 by way of
interfaces 2701a-d. In an embodiment, interfaces 2701a-d are female
sockets disposed on a substrate, such as a backplane, motherboard
or PCB, to receive male edge interfaces of memory modules
2701-2704. In an embodiment, memory modules 2701-2704 represent
memory modules including integrated circuit memory devices and
buffer devices as described herein as well as at least one of
bypass circuits 2630a-d.
[0170] Master 2101 is coupled to memory module 2701 by signal path
2710. Signal path 2711 couples memory module 2701 to memory module
2704. In an embodiment, bypass circuit 2630a allows read and write
data to be transferred between signal paths 2711 and 2710 either to
or from master device 2101 in response to control/address
information provided to memory module 2704.
[0171] Master 2101 is coupled to memory module 2702 by signal path
2712. Signal path 2713 couples memory module 2702 to memory module
2703. Signal path 2714 couples memory module 2703 to memory module
2704. In an embodiment, bypass circuits 2630b and 2630c allow read
and write data to be transferred between signal paths 2712 and
2713, as well as signal paths 2713 and 2714, either to or from
master device 2101 in response to control/address information
provided to memory modules 2702-04.
[0172] Master 2101 is coupled to memory module 2703 by signal path
2714. Signal path 2716 couples memory module 2703 to memory module
2704. In an embodiment, bypass circuit 2630c allows read and write
data to be transferred between signal paths 2714 and 2716 either to
or from master device 2101 in response to control/address
information provided to memory modules 2703-04.
[0173] Master 2101 is coupled to memory module 2704 by signal path
2717. In an embodiment, read and write data is transferred on
signal path 2717 to or from master device 2101 in response to
control/address information provided to memory module 2704.
[0174] FIGS. 28A-B illustrate a system 2700 to access different
capacity/sized (address space) memory modules during different
modes of operation that is similar in operation to that of system
2600. FIG. 28A illustrates accessing data in a first mode of
operation, such as accessing read data from different sized memory
modules that may be disposed in interfaces 2701a-d. Table 2810
illustrates how different sized memory modules may be disposed in
respective interfaces 2701a-d during a first mode of operation. For
example, interfaces 2701a-d may be coupled to all "small" sized
memory modules as indicated by the first row of Table 2810.
Alternatively, interface 2701a may be coupled to a "large" sized
memory module; interface 2701b may be coupled to a "small" sized
memory module; interface 2701c may be coupled to a "large" sized
memory module; and interface 2701d may be coupled to a "small"
sized memory module, as indicated by the second from last row of
Table 2810.
[0175] In a first mode of operation (non-bypass mode) as
illustrated by FIG. 28A, data 2810a is provided on signal path
2717; data 2820a is provided on signal path 2714; data 2830 is
provided on signal path 2712; and data 2840 is provided on signal
path 2710.
[0176] Table 2820 illustrates how different sized memory modules
may be disposed in respective interfaces 2701a-d during a second
mode of operation (bypass mode). For example, interfaces 2701c-d
may be coupled to "small" sized memory modules and interfaces
2701a-b include bypass circuits 2802 and 2801 as indicated by the
first row of Table 2820. Alternatively, interface 2701c may be
coupled to a "large" sized memory module; and interface 2701d may
be coupled to a "small" sized memory module. Interfaces 2701a-b
include bypass circuits 2802 and 2801, as indicated by Table
2820.
[0177] In a second mode of operation (bypass mode) as illustrated
by FIG. 28B, read data 2810b is provided on signal path 2717 and
read data 2810c is provided on signal paths 2711 and 2710 (via
bypass circuit 2802). Read data 2820b is provided on signal path
2714 and read data 2820c is provided on signal paths 2713 and 2712
(via bypass circuit 2801).
[0178] In embodiments, bypass circuits 2801 and/or 2802 are
disposed in a continuity module, integrated circuit buffer device,
interface (for example a socket) and/or memory module. In an
embodiment, bypass circuits 2801 and 2802 are conductive elements,
such as metal traces or wires that may be disposed manually on an
interface or memory module. In an embodiment, bypass circuits 2801
and 2802 correspond to bypass circuit 2900 shown in FIG. 29.
[0179] FIG. 29 illustrates a bypass circuit 2900 used in a write
operation according to an embodiment. Bypass circuit 2900 includes
receiver and transmitter circuits 2901a-e and 2902a-d coupled to a
signal path including signal paths DQ[0:3] and RQ. In an
embodiment, bypass circuit 2900 is included in an integrated
circuit buffer device, such as corresponding to bypass circuit 1898
in buffer interface 1103a, disposed on a memory module and/or
corresponding to bypass circuits 2630a-d shown in FIGS. 26A-B and
27. For example, signal paths DQ[0:1] are coupled to connector
interface portion 920a and signal paths DQ[2:3] are coupled to
connector interface portion 920b as shown in FIGS. 20A-B. In an
embodiment, signal paths DQ[0:1] are coupled to an adjacent master
or memory module and signal paths DQ[2:3] are coupled to a memory
module in a memory system.
[0180] Receiver circuits 2901a-d receive write data signals from
signal paths DQ[0:3] and provide write data to data width
translator circuit 2950 and/or back out to a signal path by way of
transmitters 2902a-d and bypass elements 2905-2910. Receiver
circuit 2901e receives write address signals from signal path RQ
and provides write addresses to data width translator circuit 2950.
Receiver circuit 2901a is coupled to bypass elements 2906 and 2908
to reroute received data signals to transmitter circuits 2902b and
2902c in response to control signals (not shown) provided to bypass
elements 2906 and 2908. Receiver circuit 2901b is coupled to bypass
elements 2905 and 2910 to reroute received data signals to
transmitter circuits 2902a and 2902d in response to control signals
(not shown) provided to bypass elements 2905 and 2910. Receiver
circuit 2901c is coupled to bypass element 2907 to reroute received
data signals to transmitter circuit 2902a in response to control
signals (not shown) provided to bypass element 2907. Receiver
circuit 2901d is coupled to bypass element 2909 to reroute received
data signals to transmitter circuit 2902b in response to control
signals (not shown) provided to bypass element 2909.
[0181] As can be seen, write data may be rerouted from a single
signal path DQ0 to another single signal path DQ1. Write data may
be also rerouted from two signal paths DQ0 and DQ1 to signal paths
DQ2 and DQ3.
[0182] In an embodiment, bypass elements 2905-2910 function
independently as respective switches to allow a signal (represented
by a voltage level) to be passed from a receiver circuit to a
transmitter circuit. In an embodiment, bypass elements 2905-2910
are semiconductors such as negative and/or positive-channel
metal-oxide (NMOS/PMOS) semiconductors with a control signal (such
as a voltage) provided to a gate of the semiconductor while a
source and/or a drain is coupled to a transmitter and/or receiver
circuit. In an alternate embodiment, other types of semiconductors
or switches may be used. In an embodiment, control signals (not
shown) provided to bypass elements 2905-2910 are provided by master
2101 or from a programmable register, such as an SPD device. In an
embodiment, control signals are provided by a master after reading
memory capacity information of memory modules stored in one or more
SPD devices. In an embodiment, control signals provided to bypass
elements may be provided in response to a manual jumper,
programmable fuse or register. In an embodiment, control signals
provided to bypass elements may be provided by one or more
integrated circuit buffer devices in response to one or more
integrated circuit buffer devices reading a received
address/control information. For example, when an address is
received that identifies a memory location that is not provided on
a particular memory module (non-overlapping address space or
smaller capacity memory module), control signals are provided to
bypass elements from the integrated circuit buffer device that
received the address/control information (in a bypass mode) to
enable data to be rerouted from the larger capacity memory module
to another destination, such as a master.
[0183] In an embodiment, bypass elements 2905-2910 may be disposed
before or left of receiver and transmitter circuits 2901a-d and
2902a-d as well as in or after (right of) data width translator
circuit 2950 (for example, after a clock barrier or boundary).
Bypass elements 2905-2910 may be disposed in a master, an interface
(such as a socket) and/or a memory module (outside of a buffer
device). Bypass elements 2905-2910 may also be disposed internal to
an integrated circuit buffer, as opposed to an interface of an
integrated circuit buffer device, or in an integrated circuit
memory device.
[0184] In an embodiment, rerouted write data may be resynchronized
by a transmitter circuit using a different or the same clock signal
that is used by the receiver circuit in receiving the read data.
Also, write data that has been rerouted by bypass elements may be
transmitted in a fast analog mode.
[0185] Stored read data from integrated circuit memory devices
disposed on a memory module are provided on signal paths
DQ_DRV[0:3] by way of an integrated circuit buffer device. Read
data is levelized or delays are provided to the read data by a
selector circuit, such as multiplexers (mux) 2903a-d, and delay
circuits 2904a-d in response to DELAY[0:3] control signals. Signal
paths DQ_DRV[0:3] are input to delay circuits 2904a-d and a first
input ("0 input") of mux 2903a-d, while an output of delay circuits
2904a-d is provided to a second input ("1 input") of mux 2903a-d.
DELAY[0:3] control signals select an output of mux 2903a-d or
whether a delay is introduced into read data on signal paths
DQ_DRV[0:3]. In an embodiment, delay circuits 2904a-d may introduce
a programmable delay in response to a control signal (not shown).
Control signals provided to delay circuits 2904a-d as well as
DELAY[0:3] control signals may be provided similar to control
signals provided to bypass elements 2905-2910 as described
above.
[0186] In an embodiment, delay circuits 2904a-d are inverters,
registers and/or a series of inverters and/or registers that may
introduce programmable delay to a read signal on signal paths
DQ_DRV[0:3]. The amount of delay provided to read data by delay
circuits 2904a-d may be longer than the amount of time for
providing read data to delay circuits 2904a-d, or longer than a
data cycle time.
[0187] In an embodiment, multiplexers 2903a-d and delay circuits
2904a-d may be disposed before or left of receiver and transmitter
circuits 2901a-d and 2902a-d. For example, multiplexers 2903a-d and
delay circuits 2904a-d may be disposed in a master, interface (such
as a socket) and/or memory module. In an embodiment, multiplexers
2903a-d and delay circuits 2904a-d may be disposed in data width
translator circuit 2950 and/or left of data width translator
circuit 2950. For example, multiplexers 2903a-d and delay circuits
2904a-d may be disposed internal to an integrated circuit buffer,
as opposed to an interface of an integrated circuit buffer device,
or in an integrated circuit memory device.
[0188] Levelization or the amount of delay (if any) provided to
read data on signal paths DQ_DRV[0:3] is dependent upon the signal
path (between a memory module and a master) used by a system to
provide the read data to the master (or flight time or amount of
time to transfer read data from a memory module to a master and/or
another memory module). For example in a system 2600 shown in FIG.
26B, delay is introduced into data 2601b so that data 2601b arrives
at master 2101 at the approximate same time data 2601c arrives at
master 2101 because data 2601c travels a longer path (as compared
to data 2601b) on signal paths 2611 and 2612 as well as through
memory module 2602 (or at least through an integrated buffer
device/interface of memory module 2602).
[0189] Data width translator circuit 2950 may be configurable to
translate data of various widths into data suitable for a
fixed-width memory die or device disposed on a memory module. Data
width translator circuit 2950, in accordance with some embodiments,
uses a data-mask signal to selectively prevent memory accesses to
subsets of physical addresses. This data masking divides physical
address locations of the memory die into two or more temporal
subsets of the physical address locations, effectively increasing
the number of uniquely addressable locations in a particular memory
die. As used herein, the term "width" refers to the number of bits
employed to represent data.
[0190] A data width translator circuit 2950 allows memory modules,
such as memory modules 2601 and 2602, to vary the effective width
of their external memory module interfaces without varying the
width of the internal memory device/die interfaces. A memory system
thus may support a first mode of operation and a second mode of
operation (bypass mode). In the bypass mode of operation, memory
module 2601 uses both signal path 2610 and signal paths 2611 and
2612 (via memory module 2602).
[0191] In accordance with an embodiment, data width translator
circuit 2950 can translate data of width one, two, or four on
signal paths DQ[0:3] into four-bit-wide data on signal path
IDQ[0:3]. Address translator circuit 2970 translates address
signals on signal path RQ to signal path IRQ which is coupled to
one or more memory devices. This flexibility allows one or a
combination of memory modules to be used in an extensible
point-to-point memory topology. Similarly, data width translator
circuit 2950 can translate data of width one, two, or four on
signal paths IDQ[0:3] into four-bit-wide data on signal path
DQ[0:3].
[0192] Data width translator circuit 2950 includes a data
translator circuit 2960, an address translator circuit 2970, and a
DLL 2980. DLL 2980 produces an internal differential clock signal
ICLK locked (or having a temporal relationship) to a
like-identified incoming differential clock signal CLK, typically
from an associated master or a clock-generator device. Though not
shown, a memory device disposed on a memory module may receive the
same or a similar clock signal CLK from data width translator
circuit 2950 or a master. Data translator circuit 2960 and address
translator circuit 2970, responsive to a configuration signal CFG,
translate the data on one, two, or four of data signal paths
DQ[0:3] into four-bit-wide data on signal paths IDQ[0:3] for write
cycles; and conversely translate four-bit-wide data on signal paths
IDQ[0:3] into one, two, or four-bit-wide data on one or more of
external signal paths DQ[0:3] for read cycles. In one embodiment,
plugging a second memory module into a two-connector mother board
automatically asserts configuration signal CFG, causing each of two
memory modules to configure themselves as half-width (e.g., two
bits instead of four) modules. In other embodiments, configuration
signal CFG comes from a register on a memory module (e.g., within
data width translator circuit 2950) that is addressable by a master
and is set, such as via the BIOS, at boot time. In other
embodiments, a configuration signal CFG is provided after reading
values stored in a SPD device. In general, an external memory
module interface conveys data signals of data-width N, an internal
memory device interface conveys signals of data-width M, and
configuration signal CFG is indicative of the ratio of N to M. Some
embodiments use a PLL instead of DLL 2980.
[0193] A fixed-width memory device disposed on a memory module may
include a mask line/signal path or pin that can be used in support
of partial-write operations. For example, double data rate "DDR"
memory die include a data-mask pin DM and single data rate "SDR"
memory die include a data-mask pin DQM. Memory modules detailed
herein may employ data-mask functionality to create variable-width
modules using fixed-width memory devices. In an embodiment, a
data-mask signal DM is output from data translator circuit 2960 to
one or more memory devices in order to synchronize write
operations. FIGS. 30A-B, described below, illustrate a write
operation using data width translator circuit 2950 in an
embodiment.
[0194] In an embodiment, bypass circuit 2900 includes bypass
elements 2905-2910 and not multiplexers 2903a-d and delay circuits
2904a-d. In an alternate embodiment, bypass circuit 2900 includes
multiplexers 2903a-d and delay circuits 2904a-d and not bypass
elements 2905-2910. For example, memory module 2601 shown in FIG.
26B, and in particular bypass circuit 2630a, may include
multiplexers 2903a-d and delay circuits 2904a-d to provide a delay
to data 2601a and not bypass elements 2905-2910. Conversely, memory
module 2602, and in particular bypass circuit 2630b, may include
bypass elements 2905-2910 to reroute data 2601c but not
multiplexers 2903a-d and delay circuits 2904a-d to provide a delay.
In an embodiment, bypass circuit 2900 is disposed in a memory
system that does not include an integrated circuit buffer
device.
[0195] FIGS. 30A-B illustrate a pair of timing charts 3000 and 3001
depicting the operation of a memory system, or memory module, using
data width translator circuit 2950 in a first mode of operation and
a second mode of operation (bypass mode). Data to be written to a
common address A in a single memory device disposed on a memory
module may be transmitted over external signal paths DQ[0:3] as
four eight-symbol bursts (a single eight-symbol burst 0A-0H on
signal path DQ0 is shown in FIG. 30B) and an address A on signal
path RQ. For example, signal path DQ0 conveys eight binary symbols
0A through 0H for storage at physical address location A in a fixed
width memory device on the memory module. In embodiments, the three
remaining signal paths DQ[1:3] likewise may convey eight symbols
for storage at address location A. When all signal paths DQ[0:3]
are used, the total number of symbols to be stored at a given
address A may be thirty-two (four times eight). Data width
translator circuit 2950 may convey the thirty-two symbols and
corresponding address A to a memory device via signal paths
IDQ[0:3] and IRQ. The burst length can be longer or shorter in
other embodiments.
[0196] In an embodiment, data width translator circuit 2950 uses
mask signal DM to divide the addressed physical locations in a
fixed-width memory device into subsets of memory locations
addressed separately in the time domain, a process that may be
referred to as "time slicing." For example, a most significant
bit(s) (MSB(s)), or any other bits in address A, causes data
translator circuit 2960 (via a signal from address translator
circuit 2970 to data translator circuit 2960) to assert a mask
signal DM (DM=1) to block writes to a first set of locations having
address A, and then de-asserts mask signal DM (DM=0) to allow
writes to the second set of locations having address A. This
process then may repeat.
[0197] FIG. 30A illustrates how data provided from two external
signal paths DQ[0:1] is output on signal paths IDQ[0:3] by data
width translator circuit 2950 in a bypass mode of operation (i.e.
memory modules 2701 and 2702 are bypassed as illustrated in FIGS.
27 and 28B). In an embodiment, signal path DQ0 is included in
signal path 2717 and signal path DQ1 is included in signal path
2711. Data 0A-0H is provided on signal path 2717 from master 2101
while data 1A-1H is also provided by master 2101 on signal path
2711 via memory module 2701 and signal path 2710.
[0198] In an embodiment, the address space in memory module 2704
(i.e. memory devices) is bisected in the time domain. One of the
external address bits of address A is employed to assert mask
signal DM every other time slot. In this embodiment, the MSB of the
external address A is zero, so mask signal DM is deasserted for
every time slot MSB=0 to allow writes during those time slots.
[0199] FIG. 30B illustrates how data provided from an external
signal path DQ0 (or signal paths DQ[0:3]) is output on signal paths
IDQ[0:3] by data width translator circuit 2950 in a non-bypass mode
of operation (i.e. data is provided to each of the memory
modules/sockets as illustrated in FIGS. 27 and 28A). In an
embodiment, signal path DQ0 is included in signal path 2717. Data
0A-0H is provided on signal path 2717 from master 2101. Similarly,
other data may be provided from master 2101 to memory modules
2701-2703 on signal paths DQ1, DQ2 and DQ3 that are included in
signal paths 2710, 2712 and 2714 .
[0200] FIG. 31 illustrates a method 3100 to adjust read and write
data delays in a system including memory modules having different
capacity and a bypass circuit. In embodiments, logic blocks
illustrated in FIGS. 31 and 40 are carried out by hardware,
software or a combination thereof. In embodiments, logic blocks
illustrated in FIGS. 31 and 40 illustrate actions or steps. In
embodiments, the circuits and/or systems illustrated herein, singly
or in combination, carry out the logic blocks illustrated in FIGS.
31 and 40. Other logic blocks that are not shown may be included in
various embodiments. Similarly, logic blocks that are shown may be
excluded in various embodiments. Also, while methods 3100 and 4000
shown in FIGS. 31 and 40 are described in sequential logic blocks,
steps or logic blocks of methods 3100 and 4000 are completed very
quickly or almost instantaneously.
[0201] Method 3100 begins at logic block 3101 where a determination
is made whether to levelize or adjust delays to read and write data
in a memory system. In an embodiment, this determination may be
made at initialization, periodically or during calibration
(testing). If levelization is not desired, method 3100 ends.
Otherwise, integrated circuit buffer devices are set to a typical
or first mode of operation as illustrated by logic block 3102. In
an embodiment, a control signal from a master, such as master 2101
shown in FIG. 26A-B, generates a control signal to memory modules,
and in particular to integrated circuit buffer devices of the
memory modules to operate in a first mode of operation which
includes providing read and write data on separate signal paths
(signal paths 2610 and 2612) to or from a master as illustrated in
FIG. 26A. In the first mode of operation, no additional delay is
provided to read and write data, as compared to the second mode of
operation described below.
[0202] Logic block 3103 illustrates levelizing read data or
providing delays to read data to take into account different flight
times or distances the read data must travel on different signal
paths in reaching a master. For example, signal path 2612 has a
longer signal path than signal path 2610. Therefore, in order for
read data 2601a and 2602a from both memory modules 2601 and 2602 to
reach master 2101 at the approximate same time, a delay should be
introduced into the read data 2601a to account for the longer
flight time or distance of signal path 2612. In an embodiment,
delays are provided in response to delay values stored in registers
on the integrated circuit memory devices and programmed by the
master. In alternate embodiments, delays corresponding to
respective memory modules are provided and programmed in the
master. Test symbols or test data may be written and read from the
integrated circuit memory devices to determine the programming of
the delay values.
[0203] A determination is then made whether a memory system
includes different capacity memory modules as illustrated by logic
block 3104. If different capacity memory modules are not present,
control transitions to logic block 3107. Otherwise, control
transitions to logic block 3105. In an embodiment, the
determination illustrated by logic block 3104 may be completed by a
master reading configuration information of a system stored in an
SPD.
[0204] Integrated circuit buffer devices are then set to a second
mode of operation (bypass mode) as illustrated in logic block 3105.
In an embodiment, the bypass mode of operation is set by providing
control signals to a bypass circuit in an integrated circuit buffer
device, for example bypass elements 2905-2910 in a bypass circuit
2900 as illustrated in FIG. 29.
[0205] Read data from a larger capacity memory module is then
levelized as illustrated by logic block 3106. For example, delays
are added to read data 2601b of memory module 2601 (larger
capacity) as illustrated in FIG. 26B. In an embodiment, Delay[0:3]
control signals are provided to multiplexers 2903a-d to select
additional delay to data signal on signal path DQ_DRV[0:3] of
bypass circuit 2900 shown in FIG. 29. The delay provided in logic
block 3106 is in addition to any delay provided in logic block
3103.
[0206] Integrated circuit buffers in a smaller capacity memory
module are set to a first mode of operation (or a non-bypass mode)
as illustrated by logic block 3109. For example, memory module 2602
in FIG. 26A has an integrated circuit buffer device that is set to
a typical mode of operation.
[0207] Read data levelization for the smaller capacity memory
module is then performed as illustrated by logic block 3108.
[0208] Write data levelization for data written to memory modules
is performed in logic block 3107.
[0209] A determination is then made whether a memory system
includes different capacity memory modules as illustrated by logic
block 3110. If different capacity memory modules are not present,
method 3100 ends. Otherwise, control transitions to logic block
3111. In an embodiment, the determination illustrated by logic
block 3110 may be completed by a master reading configuration
information of a system stored in a SPD.
[0210] Integrated circuit buffer devices are then set to a second
mode of operation (bypass mode) as illustrated in logic block 3111.
In an embodiment, the bypass mode of operation is set by providing
control signals to a bypass circuit in an integrated circuit buffer
device, for example bypass elements 2905-2910 in a bypass circuit
2900 as illustrated in FIG. 29.
[0211] Write data to larger capacity memory modules is then
levelized (in addition to the write data levelization illustrated
in logic block 3107) as illustrated by logic block 3112. In an
embodiment, additional write delays are added, in response to
stored write delay values, to the write data at a master,
integrated circuit buffer device and/or memory device. Delays to
write data may be selected based on whether write data is
transferred through a memory module having an integrated circuit
buffer device in a bypass mode of operation. For example, write
data provided to memory module 2601 on signal path 2610 from master
2101 may be delayed compared to write data provided to memory
module 2601 on signal paths 2612 and 2611 (by way of bypass circuit
2630b) from master 2101 so that the write data may arrive at
approximately the same time.
[0212] FIGS. 32A-E, 33A-B, 34 and 35 illustrate at least a portion
of memory system topologies including an integrated circuit buffer
device 3201 to provide control/address information (RQ) to a
plurality of integrated circuit memory devices 101a-d as well as
transferring data (DQ) between the integrated circuit buffer device
3201 and the plurality of integrated circuit memory devices 101a-d.
While each of FIGS. 32A-E, 33A-B, 34 and 35 illustrate one or more
signal paths to transfer either control/address information (RQ) or
data (DQ), other topologies or signal paths in other Figures may be
combined and used to transfer control/address information (RQ)
and/or data (DQ). For example, FIG. 33A illustrates a fly-by
topology having signal paths 3310 and 3310a-d that may be used for
transferring control/address information (RQ); while data (DQ) may
be transferred using a point-to-point (or segmented) topology or
signal paths 3410-3413 shown in FIG. 34. Numerous other topology
combinations may likewise be used in embodiments.
[0213] While topologies are illustrated with memory modules
3200a-e, 3300a-b and 3400, these illustrated topologies in FIGS.
32A-E, 33A-B and 34 may be used without a memory module. For
example, topologies illustrated in FIGS. 32A-E, 33A-B and 34 may be
used in an MCP or SIP embodiment. FIG. 35 illustrates a particular
topology in MCP device 3500.
[0214] In embodiments, a master, such as master 2101 may provide
control/address information and data to one or more integrated
circuit buffer devices 3201 in a topology illustrated in FIGS.
32A-E, 33A-B and 34. In an embodiment, a clock signal or clock
information is provided on signal paths from buffer device 3201
illustrated in FIGS. 32A-E, 33A-B and 34, or on a separate signal
path from a clock source, master, buffer device, or along the data
signal paths.
[0215] In embodiments, termination may be disposed on buffer 3201,
memory modules 3200a-e, 3300a-b and 3400, signal paths, memory
devices 101a-d and/or elsewhere in a system, such as on an PCB or
substrate. In embodiments, termination for the signal paths in the
topologies shown in FIGS. 32A-E, 33A-B and 34 may be similarly
disposed as shown in FIGS. 2-4, 6-8 and 23A-C. For example,
termination 420a-d shown in FIG. 4 may be similarly coupled to
signal paths 3410-3413 shown in FIG. 34.
[0216] FIGS. 32A-E illustrate forked (data and control/address
information) topologies between an integrated circuit buffer device
3201 and a plurality of integrated circuit memory devices 101a-d.
With respect to FIG. 32A, buffer device 3201 is coupled to signal
path 3210 disposed on memory module 3200a that then branches into
signal paths 3210a and 3210d. Signal path 3210a then is coupled to
memory devices 101a and 101b by branches or signal paths 3210b and
3210c. Signal path 3210d, likewise, is coupled to memory devices
101c and 101d by branches or signal paths 3210e and 3210f.
[0217] FIG. 32B illustrates a forked topology similar to the
topology illustrated in FIG. 32A. Signal path 3220 branches into
signal paths 3220a and 3220b that couple memory devices 101a-b to
buffer device 3201. Similarly, signal path 3230 branches into
signal paths 3230a and 3230b that couple memory devices 101c-d to
buffer device 3201.
[0218] FIG. 32C illustrates a forked/multi-drop bus topology.
Buffer device 3201 is coupled to signal path 3240 (or a stub) that
branches into signal paths 3240a and 3240b (or a bus) that are
coupled to signal paths (or stubs) 3240c-f coupled to memory
devices 101a-d. Other memory devices may be coupled to signal paths
3240a-b.
[0219] FIG. 32D illustrates a star topology. Signal path 3250
branches into signal path 3250a-d from a common node that couples
memory devices 101a-d to buffer device 3201.
[0220] FIG. 32E illustrates a forked topology similar to the
topology illustrated in FIG. 32B. Signal path 3260 branches into
signal paths 3260a and 3260b that couple memory devices 101a-b to
buffer device 3201.
[0221] FIGS. 33A-B illustrate fly-by topologies (data and/or
control/address information) between an integrated circuit buffer
device 3201 and a plurality of integrated circuit memory devices
101a-d. FIG. 33A illustrates a stub/fly-by topology including a
buffer device 3201 coupled to a signal path 3310 that is coupled to
signal paths (stubs) 3310a-d that are coupled to memory devices
101a-d. FIG. 33B illustrates a split/stub/fly-by topology. A buffer
device 3201 is coupled to a signal path 3320 that is coupled to
signal paths (stubs) 3320a-b that are coupled to memory devices
101a-b. The buffer device 3201 is also coupled to a signal path
3330 that is coupled to signal paths (stubs) 3330a-b that are
coupled to memory devices 101c-d. Split/stub/fly-by topologies may
be divided/split into even further sections in embodiments.
[0222] FIG. 34 illustrates point-to-point (also known as segmented)
topology (data and/or control/address information) between an
integrated circuit buffer device 3201 and a plurality of integrated
circuit memory devices 101a-d. Separate or segmented signal paths
3410-3413 (in particular point-to-point links) couple buffer device
3201 to memory devices 101a-d. A segmented topology for data using
separate point-to-point links is also illustrated in FIGS. 38-39
described below.
[0223] FIG. 35 illustrates an MCP (or SIP) topology (data and/or
control/address information) between an integrated circuit buffer
die 1100a and a plurality of integrated circuit memory dies
1101a-c. Device 3500 includes a plurality of integrated circuit
memory dies 1101a-c and a buffer die 1100a housed in or upon a
common package 3510 according to embodiments. A plurality of signal
paths 3501a-c are coupled to a signal path 3502 that provides data
between the integrated circuit buffer die 1100a and the plurality
of integrated circuit memory dies 1101a-c. Similarly, a plurality
of signal paths 3503a-c are coupled to a signal path 3504 that
provides control/address information from the integrated circuit
buffer die 1100a to the plurality of integrated circuit memory dies
1101a-c. As described above, a plurality of integrated circuit
memory dies 1101a-d and buffer die 1100a may be disposed with or
without spacers and in multiple package type embodiments.
[0224] FIG. 36 is a block diagram of an integrated circuit buffer
device 3600 (or a buffer die). Buffer device 3600, includes among
other circuit components, interfaces 3601 and 3611, register set
3605, data path 3606, data path router 3610, command decode 3607
and address translation 3608. Buffer device 3600 also includes
phase locked loop ("PLL") 3602, Joint Test Action Group or IEEE
1149.1 standard ("JTAG") interface 3603, Inter-IC ("I2C") interface
3604, pattern generator 3609 and internal memory array 3612 circuit
components.
[0225] In a memory read operation, buffer device 3600 operates
similar to buffer 100a shown in FIG. 18. Buffer device 3600
receives control information (including address information) that
may be in a packet format from a master on signal path 121 and in
response, transmits corresponding signals to one or more, or all of
memory devices 101a-d on one or more signal paths 1005. In an
embodiment, command decode 3607 and address translation 3608 output
control signals to data path 3606, data path router 3610 and
interface 3611 so that received read memory commands and received
read addresses are decoded and translated to corresponding
control/address signals output on signal path 1005. One or more of
memory devices 101a-d may respond by transmitting read data to
buffer device 3600 which receives the read data via one or more
signal paths 1006 and in response, transmits corresponding signals
to a master (or other buffer). In an embodiment, data path 3606 and
data path router 3610 (in response to control signals) merge
separate read data from more than one memory device into a single
merged read data or read stream output at interface 3601.
[0226] In an embodiment, memory devices 101a-d are configured into
memory ranks having segmented (point-to-point) signal paths 1006
and a shared fly-by bus signal path 1005 as illustrated in FIGS.
33A, 34, 38 and 39. A timing chart 3701 shown in FIG. 37B, and
described in detail below, illustrates an operation of buffer
device 3600 that may increase bandwidth by reducing a time bubble
when buffer device 3600 is coupled to ranked memory by segmented
signal paths as described below.
[0227] In a memory write operation embodiment, buffer 3600 operates
similar to buffer 100a. Buffer 3600 receives control information
(including address information) that may be in a packet format from
a master on signal path 121 and receives the write data for one or
more memory devices 101a-dthat may be in a packet format from a
master on signal path 120a. In an embodiment, command decode 3607
and address translation 3608 output control signals to data path
3606, data path router 3610 and interface 3611 so that received
write memory commands and received write addresses are decoded and
translated to corresponding control/address signals output on
signal path 1005. Buffer 3600 then transmits corresponding signals
to one or more, or all of memory devices 101a-d on one or more
signal paths 1006 so that the write data may be stored. In an
embodiment, data path 3606 and data path router 3610 (in response
to control signals) segments or parses received write data into two
or more write portions and directs the write portions to the
appropriate signal paths 1006 (via interface 3611) so that the
write portions will be stored in two or more memory devices.
Accordingly, buffer 3600 may receive write data having an
associated write address to a particular memory device and
parses/segments the received write data into a plurality of
different write data portions which are then routed to a plurality
of different memory devices at a plurality of different write
addresses for storage.
[0228] Interfaces 3601 and 3611 correspond to portions of
interfaces 1103a and interfaces 1820a-bshown in FIG. 18. For
example, interface 3601 may include one or more of transceiver 1875
and receiver circuit 1892 as well as termination 1880. Interface
3611 may include one or more of transceiver 1894 and transmitter
circuit 1893. In an embodiment, interface 3611 includes circuits to
interface with DDR3 memory devices and interface 3601 includes
circuits to interface with DDR2 memory devices or other type of
memory device.
[0229] In an embodiment, interface 3611 can be segmented into at
least three different configurations or segmentation modes: 1) Four
4-bit interfaces (4.times.4), 2) Two 4-bit interfaces (2.times.4)
or 3) Two 8-bit interfaces (2.times.8). The different
configurations allow flexibility in memory module or memory stack
configurations. Accordingly, buffer 3600 may interface with
high-capacity or lower-capacity entry level memory modules or in
particular memory devices. A four 4-bit interface may be used in
high capacity memory modules. A two 8-bit interface may be used for
low-cost memory modules. A two 4-bit interface may be used for
low-cost memory modules that still support ECC.
[0230] The assignment of strobe pins to data pin groupings is
adjusted depending upon the segmentation mode:
[0231] 4.times.4 segmentation mode: [0232] DQS[0]->DQ[3:0]
[0233] DQS[1]->DQ[7:4] [0234] DQS[2]->DQ[11:8] [0235]
DQS[3]->DQ[15:12]
[0236] 2.times.4 segmentation mode: [0237] DQS[0]->DQ[3:0]
[0238] DQS[1]->DQ[7:4] [0239] DQS[3:2], DQ[15:8] disabled
[0240] 2.times.8 segmentation mode: [0241] DQS[0]->DQ[7:0]
[0242] DQS[1]->DQ[15:8] [0243] DQS[3:2] disabled
[0244] Interface 3601 enters segmentation modes in response to bit
values stored in register set 3605 and/or one or more control
signals from address translation 3608.
[0245] Data path router 3610 routes read and write data between
data path 3606 and interface 3611. Control signals from command
decode 3607 and address translation 3608 determine the routing of
read/write data. Data path router also receives signals from
pattern generator 3609 and internal memory array 3612. In a mode of
operation that emulates operation with a memory device, all memory
transactions are routed to and from internal memory array 3612
rather than interface 3611. Interface 3611 may be disabled during
this mode of operation. In an embodiment, pattern generator 3609 is
used as an alternate source of data (or test pattern of data) as
well as a source for injecting ECC errors in modes of operation.
The test pattern of data may be transmitted on either interface
3601 or interface 3611 or some portion of both simultaneously.
Similarly, pattern generator 3609 may insert ECC errors on either
interface 3601 or interface 3611 or some portion of both
simultaneously. In an embodiment, data path router 3610 includes
XOR logic used for ECC error injection. In embodiments, read and
write data may proceed through data path 3606 in both directions
simultaneously. Modes of operation of buffer 3600 may be entered by
setting one or more bit values in multi-bit register set (or
storage circuit) 3605.
[0246] Data path router 3610 includes a write data router 3610a and
read data router 3610b. In an embodiment, write data router 3610a
outputs write data in response to a WCLK clock signal while the
read data router 3610b outputs read data in response to a RCLK
clock signal (either the positive or negative edge of RCLK clock
signal). The use of two clock domains may enable the buffer 3600 to
reduce latency and/or operate at a higher data rate.
[0247] During a typical mode of operation, write data router 3610a
receives write data and mask information from data path 3606 and
then routes the write data (or portion of the write data) to one of
four signal paths 1006 coupled to interface 3611. Similarly during
a read operation, read data is received from one of four signal
paths 1006 coupled to interface 3611 and routed to data path
3606.
[0248] Data path router 3610 includes a plurality of signal paths
used to merge read data from different memory devices as well as
parse write data into write data portions to be stored in multiple
memory devices.
[0249] Command decode 3607 includes a decoder to output control
signals to data path 3606, address translation 3608 and data path
router 3610 in response to control information received by
interface 3601 from signal path 121. In embodiments, the control
information may include memory transaction commands, such as read
or write commands. Other control information may include a command
to activate a particular memory bank in a particular memory device
or access information having a particular page size. In an
embodiment, command decode 3607 may remap/translate a received bank
address to a different bank address of one or more memory devices
coupled to signal paths 1006.
[0250] Address translation circuit 3608 receives an address
associated with a particular memory transaction command by way of
signal path 121 and interface 3601. For example, address
translation circuit 3608 receives an address for reading data
associated with a read command for a particular memory device in a
particular memory organization (for example, number of ranks,
number of memory devices, number of banks per memory device, page
size, bandwidth). Address translation 3608 then outputs control
signals (or a translated address and/or control signals) to
interface 3611 (and signal path 1005) so that the read data may be
read from different memory devices (via signal paths 1006) because
the memory organization coupled to interface 3611 is different than
indicated in the read command. In an embodiment, address
translation 3608 may include a storage circuit to store a look-up
table for translating addresses. Similarly, write addresses
associated with a write command are received by address translation
3608 which outputs control signals (translated write addresses) to
interface 3611 and signal path 1006 so that the corresponding write
data from data path 3606 may be written to one or more translated
write addresses of memory devices coupled to signal paths 1006.
[0251] In an embodiment, information in a received row address
field is used to output chip select signals. Buffer device 3600
outputs chip select information, such as chip select signals, from
interface 3611 to one or more integrated circuit memory devices in
response to information in a row address field received at
interface 3601. One or more row address bit values may be remapped
to chip select signals. For example, values of two particular row
address bits may be used to generate four one-hot chip select
signals from interface 3611 to four or more integrated circuit
memory devices.
[0252] In an embodiment, information in a received row address
field and received chip select signals are used to output chip
select signals. Buffer device 3600 receives chip select
information, such as chip select signals (via interface 3601) and
information in row address fields to generate one or more chip
select signals from interface 3611 to a plurality of integrated
circuit memory devices. For example, two one-hot chip select
signals received at interface 3601 along with two bit values in a
row address field may be used to output eight chip select signals
at interface 3611 to eight integrated circuit memory devices.
Similarly, four received chip select signals may be used with one
bit value in a row address field to output eight chip select
signals from interface 3611.
[0253] In an embodiment, information in a bank address field is
used to output chip select signals. Buffer device 3600 outputs chip
select information from interface 3611 to one or more integrated
circuit memory devices in response to the bank address information
received at interface 3601. Unused bank address fields/pins at
interface 3601 may be used to provide chip select information at
interface 3611. For example, interface 3601 may have 5 bank address
pins while four integrated circuit memory devices having 8 banks
each are coupled to interface 3611. The lower 3 pins, BA[2:0],
would identify a particular bank in a particular memory device
while the upper two bits BA[4:3] are used to decode/output chip
select signals. The four memory devices and buffer device 3600 then
may emulate one large memory die with 32 memory banks rather than 4
memory dies having 8 banks each.
[0254] In an embodiment, multiple chip select signals may be output
simultaneously from interface 3611 to multiple respective memory
devices in response to information in a row address field, chip
select information and/or bank address information, singly or in
combination.
[0255] Address translation circuit 3608 includes one or more
multiplexers to receive (via interface 3601) information in a row
address field, chip select information and/or bank address
information and outputs signals to interface 3611 that in turn
outputs chips select signals in an embodiment.
[0256] One or more column address bit values may be
re-tasked/remapped by buffer 3600 to perform time slicing, as
described above, in an embodiment. For example, the functions (or
portions thereof) of data width translator 2950 may be performed by
address translator 3608, command decode 3607, data path 3606 and
data path router 3610, singly or in combination. Also, bit values
in a column address field may also be used to initiate memory
device functions/operations. When information in a column address
field are re-tasked and this re-tasking uses lower order bit
values, the remaining address bit values may be shifted to fill the
lowest order column address bit values output at interface 3611.
For example, when bit values in column address A[4:3] in a column
address field are remapped to time slice address bits, column
address values in column address A[15:5] are shifted to column
address A[13:3] to fill the lowest order column address bits.
[0257] In an embodiment, column address bit values may not be
shifted when column address bit values are used to initiate a
memory device operation. For example, a bit value in column address
A[10] may be used to trigger an auto-precharge operation in a DDR3
memory device. When time slicing is used as described above, a bit
value in column address bit A[10] would be mapped to column address
bit A[10] (or not changed) while bit values in column addresses
A[15:11] and A[9:5] are shifted to fill the gap caused by
re-tasking bit values in column address A[4:3]. Another similar
example of not shifting a particular column address value includes
a bit value at column address A[12] used to trigger burst chop on
column address cycles in a DDR3 memory device. In a burst chop mode
of a DDR3 memory device, a portion of the read data (for example
the last 4 bits of 8 bit output data) is masked or not output from
an integrated circuit memory device.
[0258] Buffer device 3600 may remap column bit values used to
initiate a memory device operation (i.e., auto-precharge, burst
chop, read sequence ordering) to particular column address bit
fields. For example, bit values in column address bits A[2:0] are
used to define bit ordering from a DDR memory device. Data on each
signal line coupled to an integrated circuit memory device will be
returned in a different order depending on the column bit values at
column address bits A[2:0]. When buffer device 3600 performs time
slicing, these column bit values are reassigned to a different
value to match a "time" address used to store data and to
efficiently move data from an integrated circuit memory device to
buffer device 3600. In an embodiment, data path 3606 rearranges the
data (from data path router 3610) in response to control signals
from address translation circuit 3608 which receives column bit
address values at column address A[2:0].
[0259] When less data is needed by buffer device 3600 than expected
by an integrated circuit memory device, such as in time slicing,
the buffer device 3600 may use burst chop to save I/O power from
the integrated circuit memory devices. This would be irrelevant of
the value of a column address bit A[12] (BCN). The received BCN bit
values may be stored in the data path 3605 or command decode
circuit 3607 that outputs signals to chop the data as originally
requested by way of interface 3601.
[0260] In an embodiment, received chip select information and bit
values in a received row address fields may be used by buffer
device 3600 to assign/remap column bit values in column addresses
output at interface 3611.
[0261] Address translation circuit 3608 includes one or more
multiplexers to receive (via interface 3601) information in a
column address fields and reassign/re-task column address bit
values during time slicing and/or otherwise as describe above.
[0262] Buffer device 3600 may receive row address values or chip
select information that then may be used to configure a memory
system that accesses different sized/capacity (address space)
memory modules during different modes of operations as described
above in regard to FIGS. 25-29. For example, row address values or
chip select information may be used to select whether particular
signal path widths are used in accessing different sized memory
modules during different modes of operation as illustrated in FIGS.
25A-B. In another example, row address values or chip select
information may be used to configure bypass circuit 2900 shown in
FIG. 29, such as enabling or disabling bypass paths (i.e. via
bypass elements 2905-2910) as well as selecting delay multiplexers
(i.e. outputting appropriate DELAY[0:3] control signals) shown in
FIG. 29.
[0263] In embodiments, buffer 3600 may include JTAG 3603 and/or I2C
3604 interfaces/circuits for accessing bit values in register set
3605. JTAG 3603 may include a port having test pins used during
testing of buffer 3600. An I2C 3604 may be used for outputting or
receiving bit values (by way of an I2C bus) for register set 3605
that outputs control signals to buffer device circuit components in
response to stored bit values that may represent particular buffer
configurations. In an embodiment, bit values in register set 3605
may be accessed (written/read) directly through interface 3601.
[0264] In an embodiment, register set 3605 corresponds to
configuration register set 1881 shown in FIG. 18. In an embodiment,
registers set 3605 stores one or more bit values that indicate
memory system topology so that interface 3611 may be configured
accordingly. For example, register set 3605 may include bit values
that indicate a number of integrated circuit memory devices
selected for a received memory transaction/operation. Buffer device
3600 then may configure interface 3611 (in response to register
value) in order to match the bandwidth associated with interface
3601.
[0265] In an embodiment, register set 3605 may store one or more
bit values indicating where to obtain information in received
control information (i.e. a request packet) that may be used in
determining/remapping and outputting chip select information or
signals to one or more integrated circuit memory devices. As
described below, information in row address fields, column address
fields, bank address fields as well as received chip select signals
may be used to decode and output predetermined chip select signals
from integrated circuit buffer device 3600 to the plurality of
integrated circuit memory devices.
[0266] In an embodiment, register set 3605 may store one or more
bit values to indicate a number of signal paths (i.e. width), type
of signal path topology, a number of signal lines per signal path
and/or a number (or existence) of data signal strobe signal lines
between integrated circuit buffer device 3611 (in particular
interface 3611) and a plurality of integrated circuit memory
devices.
[0267] In an embodiment, register set 3605 may store one or more
bit values to indicate how received column, row and/or bank
addresses are reordered and output from buffer device 3600.
[0268] PLL 3602 is used to synchronize the timing of receiving
and/or transmitting read and write data both internally and
externally to buffer 3600. In alternate embodiments, PLL 3602 may
be another clock alignment circuit that corresponds to clock
circuit 1870 shown in FIG. 18. In an embodiment, PLL 3602 outputs
WCLK and RCLK clock signals in response to a clock source that may
be provided to buffer 3600.
[0269] FIGS. 37A-B illustrate timing diagrams for an integrated
circuit buffer device. In particular, FIG. 37A illustrates a timing
chart 3700 that identifies when a buffer device, such as buffer
device 3600, receives and outputs control/address information as
well as receives and outputs read data when using a shared or
command data signal path.
[0270] Control information, such as commands to activate a memory
rank are illustrated by a shaded block A.sub.n that represents the
amount of time control signals are provided on a control/address
signal path (external (Ext.) RQ or internal (Int.) RQ signal paths)
during cycles of a Clock signal. For example, shaded block A.sub.a
on a row labeled Ext. RQ represents a buffer device receiving a
command to activate a memory rank "a" on an Ext. RQ signal path
during a first clock cycle of the Clock signal. Similarly a command
to read a particular memory bank is illustrated by shaded blocks
R.sub.n on signal paths Ext. RQ and Int. RQ. For example, timing
chart 3700 illustrates how a read command R.sub.a is received by a
buffer device via signal path Ext. RQ and a command R.sub.a is
output a clock cycle later onto signal path Int. RQ. In alternate
embodiments, more or less memory commands or control signals may be
received and generated.
[0271] Similarly, read data transferred on signal paths Ext. DQ and
Int. DQ to a memory controller or from a memory rank are
illustrated by a shaded block labeled Read Data.sub.n. Write data
may be similarly transferred.
[0272] Signal path Ext. RQ refers to a signal path that provides
control/address information from a memory controller to the buffer
device. Signal path Int. RQ refers to a signal path that provides
control/address information from the buffer device to a plurality
of integrated circuit memory devices or memory rank. Signal path
Ext. DQ refers to a signal path that provides Read Data, from the
buffer device to the memory controller. Signal path Int. DQ refers
to a signal path that provides Read Data, from a plurality of
integrated circuit memory devices or memory rank to the buffer
device. In an embodiment, Ext. RQ corresponds to signal path 121
and Int. RQ corresponds to signal path 1005; while Ext. DQ
corresponds to signal path 120a and Int. DQ corresponds to signal
path 1006.
[0273] Timing chart 3700 illustrates that when memory ranks are
coupled to the same (or shared/common) signal path that transfers
Read Data.sub.n, a memory system may have to be more complicated
and less efficient. In particular, a shared signal path among
memory ranks for transferring Read Data, may require a memory
controller to track accesses to memory ranks and insert bubbles
when changing access to different memory ranks. A "bubble" or "time
bubble" refers to an amount of idle time a memory controller must
insert in transferring data when switching between memory
transaction to the same memory rank. For example, a memory
controller may have to insert a bubble or idle time when switching
from accessing different memory ranks so as to allow the shared or
common bus to settle (or allow time for tri-state drivers in a
transceiver to switch to an alternate state as well as allow time
for another preamble signal) or for noise to dissipate before
initiating another memory rank access or (in the case of strobed
memory devices) to allow for a strobe preamble. This insertion of
bubbles reduces signal path utilization and may lower bandwidth on
both internal and external signal paths.
[0274] FIG. 37B illustrates a timing chart 3701 that eliminates the
need for a memory controller to track memory rank accesses and
insert bubbles thereby reducing memory controller complexity and
increasing bandwidth. Timing chart 3701 is similar to timing chart
3700 except rather than having a shared signal path for
transferring data between a buffer device and memory ranks,
segmented signal paths or dedicated signal paths Int. DQ(0)-(7) are
provided between the buffer device and each memory rank (8 memory
ranks). Bubbles are no longer present on the Ext. DQ signal path as
Read Data.sub.a-f are provided on separate signal paths Int.
DQ(0)-(7) from respective memory ranks.
[0275] FIG. 38 illustrates a system 3800 including a buffer device
3600 and a plurality of integrated circuit memory devices 101a-101n
organized in different memory ranks (1-4). System 3800 may be
included in a memory system including other buffer devices and/or
memory controllers as described herein.
[0276] A "memory rank" or "rank" refers to a number of integrated
circuit memory devices grouped to output a predetermined amount of
data bits or blocks of data, such as 72 data bits (64 data bits
plus 8 ECC bits provided by an ECC device), onto a signal path
during a predetermined period of time. For example a dual rank
system (using rank 1 and rank 2 shown in FIG. 38) may provide two
64 data bit blocks from two sets of integrated circuit memory
devices, rank 1 and rank 2. In an embodiment, the integrated
circuit memory devices may be x4 memory devices (memory devices
that produce 4 bits of data) or x8 memory devices (memory devices
that produce 8 bits of data). In this example, 8.times.8 memory
devices could produce a 64 data bit block or 16 x4 memory devices
could produce a 64 data bit block. In embodiments, different
numbers of ranks may be used.
[0277] Buffer device 3600 receives control/address information as
well as data from a memory controller via signal paths 120a and
121. In an embodiment, interface 3601 as illustrated in FIG. 36, is
used to receive control/address information and write data as well
as output read data from integrated circuit memory devices in
system 3800. Buffer device 3600 outputs translated (and/or decoded)
control/address information as well as selected write data to
integrated circuit memory devices 101a-n in memory ranks 1-4 using
interface 3611 of buffer 3600.
[0278] Interface 3611 is coupled to signal paths 3801-3804 and
signal path 3810. Signal paths 3801-3804 are segmented signal paths
to transfer read and write data between buffer device 3600 and
integrated circuit memory devices in ranks 1-4. Signal path 3801 is
coupled to memory devices 101a-n in rank 1. Signal path 3802 is
coupled to memory devices 101a-n in rank 2. Signal path 3803 is
coupled to memory devices 101a-n in rank 3. Signal path 3804 is
coupled to memory devices 101a-n in rank 4. In an embodiment, read
and write data is transferred using a segmented topology as
illustrated in FIG. 34.
[0279] In contrast, signal path 3810 provides control/address
information to memory ranks 1-4 on a shared/common signal path
3810, such as a fly-by topology shown in FIG. 33A. Each memory
device in each memory rank is coupled to shared signal path 3810.
In embodiments, clock signals or clock information may be provided
on either signal paths 3801-3804 or signal path 3810 or on another
separate signal path.
[0280] FIG. 39 illustrates a system 3900 for accessing individual
memory devices that function as respective memory ranks. System
3900 illustrates an embodiment similar to system 3800 except that
memory devices 3901a-h are included in respective memory ranks. In
an embodiment, memory devices 3900a-h are eight x4 DDR3 memory
devices. Accordingly, system 3900 is an eight rank system having
respective segmented data signal paths. Segmented signal paths
3904a-h transfer data bits DQ [0:3] between data segment
(segmentation) and merge circuit 3902 and respective memory devices
3901a-h. A data mask signal DM is provided to respective memory
devices 3901a-h from data segment and merge circuit 3902.
Similarly, clock signals or differential strobe signals DQS and
DQSN are provided from data segment and merge circuit 3902 for
synchronization of data signals. Control/address signals are
provided on signal path 3903 that is a shared signal path similar
to signal path 3810 shown in FIG. 38.
[0281] In an embodiment, data segment and merge circuit 3902
operates similar to one or more circuit components in buffer device
3600 shown in FIG. 36. Data segment and merge circuit 3902 merges
read data from a plurality of memory devices 3901a-h onto a single
signal path as a read data stream. Likewise, data segment and merge
circuit 3902 segments a single write data from a single signal path
into multiple write data output to multiple signal paths coupled to
multiple memory devices 3901a-h. For example, data segment and
merge circuit 3902 may include the functionality of data path
circuit 3606, data path router 3610, command decode 3607 and
address translation circuit 3608, singly or in combination. In an
embodiment, mux control and RQ state information is provided by a
control circuit, such as command decode 3607 and address
translation circuit 3608 shown in FIG. 36. Mux control and RQ state
information determines the source or destination of read/write
data.
[0282] FIG. 40 illustrates a method 4000 of operation in an
integrated circuit buffer device. In an embodiment, buffer device
3600 performs method 4000. Method 4000 begins at logic block 4001
where an integrated circuit buffer device is reset and/or power is
provided. In logic block 4002, an integrated circuit buffer device
receives first control information that indicates a read operation
for a first memory organization. In an embodiment, a master
provides the first control information to access a first memory
configuration that includes a first predetermined number of memory
devices, banks as well as predetermined page length/size and
bandwidth. However, the buffer device interfaces with a second
different memory organization that may include a second
predetermined number of memory devices, banks as well as
predetermined page length/size and bandwidth.
[0283] A virtual page size/length may be the size of data or memory
block that may be used by a processor or memory controller. For
example, if a process requests an operating system to allocate 64
bytes, but the page size is 4 KB, then the operating system must
allocate an entire virtual page or 4 KB to the process. In
embodiments, a physical page size/length may equal the amount of
data provided by a memory rank or the amount of data bits available
from a plurality of sense amplifiers in one or more banks of one of
more integrated circuit memory devices in the memory rank. A
virtual page size may equal a physical page size in an embodiment.
A memory controller may be able to adjust the virtual page size but
not the physical page size.
[0284] Logic blocks 4003 and 4004 illustrate outputting second and
third control information to a first signal path coupled to first
and second integrated circuit memory devices in the second memory
organization.
[0285] Logic blocks 4005 and 4006 illustrate receiving first and
second data from second and third signal paths coupled to the first
and second integrated circuit memory devices in the second memory
organization.
[0286] Logic block 4007 illustrates merging and output read data
that includes the first and second read data from the integrated
circuit buffer device in response to the first control
information.
[0287] In an embodiment, one or more logic blocks 4002-4007 may be
repeated.
[0288] Logic block 4008 illustrates ending method 4000 when power
is removed. In alternate embodiments, method 4000 may end without
power removed.
[0289] A method of operation of a buffer device that transfers
write data performs similar steps illustrated in method 4000.
However rather than receiving and outputting read data as
illustrated by blocks 4005-4007, write data may be segmented and
transferred to second and third signal paths in response to first
control information.
[0290] Signals described herein may be transmitted or received
between and within devices/circuits using signal paths and
generated using any number of signaling techniques including
without limitation, modulating the voltage or current level of an
electrical signal. The signals may represent any type of control
and timing information (e.g. commands, address values, clock
signals, and configuration/parameter information) as well as data.
In an embodiment, a signal described herein may be an optical
signal.
[0291] A variety of signals may be transferred on signal paths as
described herein. For example, types of signals include
differential (over a pair of signal lines), non-return to zero
("NRZ"), multi-level pulse amplitude modulation ("PAM"), phase
shift keying, delay or time modulation, quadrature amplitude
modulation ("QAM") and Trellis coding.
[0292] In an embodiment employing multi-level PAM signaling, a data
rate may be increased without increasing either the system clock
frequency or the number of signal lines by employing multiple
voltage levels to encode unique sets of consecutive digital values
or symbols. That is, each unique combination of consecutive digital
symbols may be assigned to a unique voltage level, or pattern of
voltage levels. For example, a 4-level PAM scheme may employ four
distinct voltage ranges to distinguish between a pair of
consecutive digital values or symbols such as 00, 01, 10 and 11.
Here, each voltage range would correspond to one of the unique
pairs of consecutive symbols.
[0293] In an embodiment, a clock signal is used to synchronize
events in a memory module and/or device such as synchronizing
receiving and transmitting data and/or control information. In an
embodiment, globally synchronous clocking is used (i.e., where a
single clock frequency source is distributed to various devices in
a memory module/system). In an embodiment, source synchronous
clocking is used (i.e., where data is transported alongside a clock
signal from a source to a destination such that a clock signal and
data become skew tolerant). In an embodiment, encoding data and a
clock signal is used. In alternate embodiments, combinations of
clocking or synchronization described herein are used.
[0294] In embodiments, signal paths described herein include one or
more conducting elements, such as a plurality of wires, metal
traces (internal or external), signal lines or doped regions
(positively or negatively enhanced), as well as one or more optical
fibers or optical pathways, singly or in combination. In
embodiments, multiple signal paths may replace a single signal path
illustrated in the Figures and a single signal path may replace
multiple signal paths illustrated in the Figures. In embodiments, a
signal path may include a bus and/or point-to-point connection. In
an embodiment, signal paths include signal paths for transferring
control and data signals. In an alternate embodiment, signal paths
include only signals paths for transferring data signals or only
signal paths for transferring control signals. In still other
embodiments, signal paths transfer unidirectional signals (signals
that travel in one direction) or bidirectional signals (signals
that travel in two directions) or combinations of both
unidirectional and bidirectional signals.
[0295] It should be noted that the various circuits disclosed
herein may be described using computer aided design tools and
expressed (or represented) as data and/or instructions embodied in
various computer-readable media, in terms of their behavior,
register transfer, logic component, transistor, layout geometries,
and/or other characteristics. Formats of files and other objects in
which such circuit expressions may be implemented include, but are
not limited to: formats supporting behavioral languages such as C,
Verilog, and HLDL; formats supporting register level description
languages like RTL; formats supporting geometry description
languages such as GDSII, GDSIII, GDSIV, CIF, MEBES; and any other
suitable formats and languages. Computer-readable media in which
such formatted data and/or instructions may be embodied include,
but are not limited to, non-volatile storage media in various forms
(e.g., optical, magnetic or semiconductor storage media) and
carrier waves that may be used to transfer such formatted data
and/or instructions through wireless, optical, or wired signaling
media or any combination thereof. Examples of transfers of such
formatted data and/or instructions by carrier waves include, but
are not limited to, transfers (uploads, downloads, e-mail, etc.)
over the Internet and/or other computer networks via one or more
data transfer protocols (e.g., HTTP, FTP, SMTP, etc.). When
received within a computer system via one or more computer-readable
media, such data and/or instruction-based expressions of the above
described circuits may be processed by a processing entity (e.g.,
one or more processors) within the computer system in conjunction
with execution of one or more other computer programs including,
without limitation, netlist generation programs, place and route
programs and the like, to generate a representation or image of a
physical manifestation of such circuits. Such representation or
image may thereafter be used in device fabrication, for example, by
enabling generation of one or more masks that are used to form
various components of the circuits in a device fabrication
process.
[0296] The foregoing description of several embodiments has been
provided for the purposes of illustration and description. It is
not intended to be exhaustive or to limit the embodiments to the
precise forms disclosed. Modifications and variations will be
apparent to practitioners skilled in the art. The embodiments were
chosen and described in order to explain inventive principles and
practical applications, thereby enabling others skilled in the art
to understand various embodiments and with the various
modifications as are suited to the particular use contemplated. It
is intended that the scope of the invention be defined by the
following claims and their equivalents.
* * * * *