High Bandwidth Memory Module Architecture
The HB-DIMM architecture addresses memory bandwidth limitations by employing multiple data buffer chips and a register clock driver circuit to achieve double data rates and reduce signal loading, enhancing performance for high-bandwidth applications.
Patent Information
- Application Number
- JP2024533935
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-06-24
- Filing Date
- 2022-11-04
- Publication Date
- 2024-11-28
- Estimated Expiration
- 2042-11-04
AI Technical Summary
Current DDR DRAM chips fail to meet the increasing memory bandwidth requirements for high-performance graphics processors and servers due to limitations in existing DIMM architectures, necessitating improved DIMM designs to handle higher data rates and reduce signal loading.
The implementation of a High Bandwidth DIMM (HB-DIMM) architecture that includes multiple data buffer chips and a register clock driver circuit, allowing for twice the data rate of memory chips through independently addressable pseudo channels, and supporting flexible configurations like quad-rank and dual-channel modes to optimize signal transmission.
HB-DIMM achieves twice the data rate of conventional DDR DRAM chips, reduces signal loading, and extends the operational lifespan of DDR5 standards by enabling higher operating frequencies and supporting various application requirements.
Smart Images

Figure 2024544217000001_ABST
Abstract
Description
[Background technology]
[0001] Modern dynamic random-access memory (DRAM) provides high memory bandwidth by increasing the speed of data transmission on a bus that connects the DRAM to one or more data processors, such as a graphics processing unit (GPU) or central processing unit (CPU). DRAM is typically inexpensive and dense, which allows for large amounts of DRAM to be integrated per device. Most DRAM chips sold today conform to various double data rate (DDR) DRAM standards promoted by the Joint Electron Devices Engineering Council (JEDEC). Typically, several DDR DRAM chips are combined on a single printed circuit board to form a memory module that is relatively fast but also offers scalability.
[0002] However, while these enhancements have improved the speed of DDR memory used for main memory in computer systems, further improvements are required, particularly as the memory bandwidth required for applications such as high performance graphics processors and servers with multiple cores and corresponding increased bandwidth requirements per core is outpacing the roadmap for bandwidth improvements in DDR DRAM chips. Improved DIMM architectures are needed to meet such requirements with current DDR chip technologies such as DDR5.
[0003] One such architecture is the load-reduced dual inline memory module (LRDIMM). LRDIMM is a load-reduced (LR) DIMM (typically employed in servers) that supports higher density than existing DIMMs. LRDIMM contains a memory buffer (MB) chip rather than a register to reduce the load on the memory bus. The memory buffer reduces each clock, command / address, and data input to a single impedance load on the memory bus data lines, which increases memory speed.
[0004] Another proposed architecture for improving bandwidth performance is the high-bandwidth DIMM (HB-DIMM), which attempts to use the advantages of the high-bandwidth memory (HBM) format in a DIMM form factor. [Brief description of the drawings]
[0005] [Figure 1] 1 is a block diagram illustrating a data processing system according to some embodiments. [Diagram 2] FIG. 1 is a block diagram illustrating a DDR5 Load Reduced Dual In-line Memory Module (LRDIMM) according to the prior art. [Diagram 3] 1 is a block diagram illustrating a high-bandwidth DIMM (HB-DIMM) in accordance with some embodiments. [Figure 4] 4 is a timing diagram illustrating several signals illustrating the operation of the HB-DIMM of FIG. 3. [Diagram 5] 1 is a block diagram illustrating an HB-DIMM according to some other embodiments. [Figure 6] 1 is a block diagram illustrating an HB-DIMM according to some other embodiments. [Figure 7] 6 is a timing diagram illustrating several signals illustrating the operation of the HB-DIMM of FIG. 5 in a 1N mode of operation. [Figure 8] 6 is a timing diagram illustrating several signals illustrating the operation of the HB-DIMM of FIG. 5 in a 2N mode of operation. [Figure 9] 1 is a table containing various options available for configuring an HB-DIMM, according to various embodiments. [Figure 10] 1 is a table including exemplary clock speeds and data rates for HB-DIMMs according to some embodiments. [Figure 11] 1 is a table containing exemplary speed limits for various configurations of HB-DIMMs, according to various embodiments. [Figure 12] 1 is a table containing various HB-DIMM module configurations according to various embodiments. [Figure 13] FIG. 2 is a block diagram illustrating a module register clock driver (MRCD) according to some embodiments. [Figure 14] 1 is a mixed circuit and block diagram illustrating a module data buffer (MDB) in accordance with some embodiments. [Figure 15] 4 is a flowchart illustrating a process for operating an HB-DIMM in accordance with some embodiments. [Figure 16] FIG. 2 is a block diagram illustrating an HB-DIMM according to some embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0006] In the following description, the use of the same reference numbers in different figures indicates similar or identical items. Unless otherwise noted, the word "coupled" and its related verb forms include both direct and indirect electrical connections by means well known in the art, and unless otherwise noted, any description of a direct connection also refers to alternative embodiments using a suitable form of indirect electrical connection.
[0007] A high bandwidth dual in-line memory module (HB-DIMM) is adapted to couple to a host bus and includes a plurality of memory chips, a plurality of data buffer chips, and a register clock driver (RCD) circuit. The plurality of data buffer chips are coupled to respective sets of memory chips and capable of transmitting data from the memory chips over the host bus at a data rate twice that of the memory chips. The RCD circuit includes a host bus interface for coupling to the host bus and a memory interface coupled to the plurality of memory chips. The RCD circuit is capable of executing memory access commands received over the host bus by routing command / address (C / A) signals to the memory chips to provide at least two independently addressable pseudo channels, and the RCD circuit addresses each pseudo channel based on chip identifier (CID) bits derived from the C / A signals.
[0008] A method for operating an HB-DIMM includes receiving a memory access command over a memory interface bus at a register clock driver (RCD) circuit. In the RCD circuit, the method includes routing a command / address (C / A) signal for the memory access command to a selected one of a plurality of memory chips to provide at least two independently addressable pseudo channels, the RCD circuit addressing each respective pseudo channel based on a chip identifier (CID) bit of the C / A signal. In the RCD circuit, the method includes preparing a data buffer command code that indicates a pseudo channel associated with the selected memory access command on which the command is to be executed.
[0009] The data processing system includes a host bus, a memory controller, and an HB-DIMM. The memory controller is coupled to the host bus for executing memory access commands via at least two independently addressable pseudo channels. The HB-DIMM is coupled to the host bus and includes a plurality of memory chips, a plurality of data buffer chips, and a register clock driver. The plurality of data buffer chips are coupled to respective sets of memory chips and capable of transmitting data from the memory chips via the host bus at a data rate twice that of the memory chips. The register clock driver (RCD) circuit includes a host bus interface for coupling to the host bus and a memory interface coupled to the plurality of memory chips. The RCD circuit executes memory access commands received via the host bus by routing command / address (C / A) signals to the memory chips for the at least two independently addressable pseudo channels. The RCD circuit addresses each pseudo channel based on a chip identifier (CID) bit of the C / A signal.
[0010] 1 is a block diagram illustrating a data processing system 100 according to some embodiments. Data processing system 100 generally includes data processors in the form of a graphics processing unit (GPU) 110, a host central processing unit (CPU) 120, double data rate (DDR) memory 130, and HB-DIMM memory 140.
[0011] GPU 110 is a discrete graphics processor with very high performance for optimized graphics processing, rendering and display, but requires high memory bandwidth to perform these tasks. GPU 110 generally includes a set of command processors 111, a graphics single instruction multiple data (SIMD) core 112, a set of caches 113, a memory controller 114, a DDR physical interface circuit (DDR PHY) 115, and a HB-DIMM PHY 116.
[0012] Command processor 111 is used to interpret high-level graphics instructions such as those specified in the OpenGL programming language. Command processor 111 has a bidirectional connection to memory controller 114 for receiving high-level graphics instructions such as OpenGL instructions, a bidirectional connection to cache 113, and a bidirectional connection to graphics SIMD core 112. In response to receiving the high-level instructions, command processor issues low-level instructions for rendering, geometric processing, shading, and rasterization of data such as frame data, using cache 113 as temporary storage. In response to the graphics instructions, graphics SIMD core 112 executes the low-level instructions on large data sets in a massively parallel manner. Command processor 111 and cache 113 are used for temporary storage of input data and output (e.g., rendered and rasterized) data. Cache 113 also has a bidirectional connection to graphics SIMD core 112 and a bidirectional connection to memory controller 114.
[0013] The memory controller 114 has a first upstream port connected to the command processor 111, a second upstream port connected to the cache 113, a first downstream bidirectional port to the DDR PHY 115, and a second downstream bidirectional port to the HB-DIMM PHY 116. As used herein, the "upstream" port is the side of the circuitry towards the data processor and away from the memory, and the "downstream" port is in the direction away from the data processor and towards the memory. The memory controller 114 controls the timing and sequencing of data transfers between the DDR memory 130 and the HB-DIMM 140. The DDR memory has asymmetric access, i.e., access to open pages in the memory is faster than access to closed pages. The memory controller 114 stores memory access commands and processes them out of order for efficiency, for example, by prioritizing access to open pages while adhering to certain quality of service targets.
[0014] The DDR PHY 115 has an upstream port connected to a first downstream port of the memory controller 114 and a downstream port bidirectionally connected to the DDR memory 130. The DDR PHY 115 meets all specified timing parameters of a version of the DDR memory 130, such as DDR version 5 (DDR5), and performs timing calibration operations at the direction of the memory controller 114. Similarly, the HB-DIMM PHY 116 has an upstream port connected to a second downstream port of the memory controller 114 and a downstream port bidirectionally connected to the HB-DIMM 140. The HB-DIMM PHY 116, in this embodiment, employs DDR5 DRAM chips and a new high-speed PHY interface standard in which the data rate on the PHY is twice the data rate of the memory chip, but the HB-DIMM form factor and connector are compatible with the existing LRDIMM specification. The HB-DIMM 140 includes a set of mode registers 141 that are programmable via the HB-DIMM PHY 116 to configure the HB-DIMM 140 for operation.
[0015] During operation, the data processing system can be used as a graphics card or accelerator due to the high bandwidth graphics processing performed by the graphics SIMD cores 112. The host CPU 120, running an operating system or application programs, sends graphics processing commands to the GPU 110 through the DDR memory 130, which serves as a unified memory for the GPU 110 and the host CPU 120. It may send commands to the GPU interface, for example, using OpenGL commands or through any other host CPU. OpenGL is a cross-language, cross-platform application programming interface for rendering 2D and 3D vector graphics developed by the Khronos Group. The host CPU 120 interacts with the GPU 110 using an application programming interface (API) to provide hardware accelerated rendering.
[0016] 2 is a block diagram illustrating a prior art DDR5 Load Reduced Dual In-line Memory Module (LRDIMM) 250 (LRDIMM 250). LRDIMM 250 is generally embodied on a printed circuit board (PCB) or other module format and includes multiple DRAM chips 252, multiple data buffers (DB) 254, and a register clock driver (RCD) 256.
[0017] Shown are the operating speeds of the data (DQ) lines of the memory bus (6.4 Gbps), the host clock DCK (3.2 GHz), and the command / address (DCA_a) lines operating in either a 6.4 Gbps "1N" mode or a 3.2 Gbps "2N" mode. There are four ranks of DRAM chips 252, and DB 254 works with RCD buffer 256 to isolate the electrical load of DRAM chips 252 from the memory bus, making LRDIMM 250 appear to the system as a dual-rank DIMM. As can be seen, LRDIMM 250 transmits and receives data over the memory bus at the speed of the DRAM chips (6.4 Gbps) and does not provide increased memory bandwidth. Data and C / A signaling are transmitted at a 1:1 ratio to the speed adopted by the DDR5 standard. RCD 256 and DB 254 do not handle quad-rank logic or timing.
[0018] FIG. 3 is a block diagram illustrating a high bandwidth DIMM (HB-DIMM) 300, according to some embodiments. The HB-DIMM 300 is embodied on a PCB or other module format and includes multiple DRAM chips 302, multiple data buffers (DB) 304, and a register clock driver (RCD) 306, and interfaces to a memory bus via pins generally shown along the bottom of the module. The HB-DIMM 300 includes two independently addressable pseudo channels, labeled "PC0" and "PC1." Generally, the HB-DIMM 300 provides a data (DQ) rate of 12.8 Gbps in this embodiment, which is double the data rate of the DRAM chips 302 (6.4 Gbps). Although a "x4" configuration is shown with one physical rank (only half of the module, one side of the HB-DIMM is shown), a "x8" configuration with two physical ranks is used in some other embodiments, as indicated by the dotted lines indicating the additional ranks of the DRAM chips 302.
[0019] As shown, pseudo channels PC0 and PC1 connect to DB 304 using separate 6.4Gbps links of four pins, with eight data pins connecting to each DB 304. Thus, DB 304 can provide a capacity of 12.8Gbps over the memory bus. RCD 306 circuitry includes a memory bus (host memory bus) interface for coupling to the memory bus and a memory interface connected to each of the DRAM chips 302. RCD 306 is capable of executing memory access commands received over the host bus by routing command / address (C / A) signals to the DRAM chips to provide at least two independently addressable pseudo channels. RCD 306 addresses each pseudo channel based on chip identifier (CID) bits derived from the C / A signals.
[0020] The HB-DIMM 300 has an operating speed of the data (DQ) lines of the memory bus of 12.8 Gbps, while the command / address ("DCA_a") lines operate in a "1N" mode at 12.8 Gbps or a "2N" mode at 6.4 Gbps. In this embodiment, the host clock DCK can also operate slower in a 2N mode, a capability not offered by LRDIMMs or other conventional DDRx DIMMs, where the DCK operates at 6.4 GHz in the 1N mode but 3.2 GHz in the 2N mode, a 2:1 ratio to the data rate on the DCA lines. This 2N mode of the DCK is made possible by clocking the C / A bus at double data rate, rather than a single data rate as employed in, for example, LRDIMMs.
[0021] FIG. 4 is a timing diagram 400 illustrating several signals indicative of the operation of the HB-DIMM 300 of FIG. 3. The horizontal axis indicates time. Shown are the host clock signal "DCK", the C / A command sequence "DCA", the local clock signals and buffered command bus clock signals labeled "QCK / BCK", the C / A command sequence for PC0 labeled "QCA PC0", the C / A sequence for PC1 labeled "QCA PC1", the data sequence for PC0 labeled "MDQ PC0", the data sequence for PC1 labeled "MDQ PC1", and the data sequence for the memory bus data (DQ) lines labeled "DQ". The commands and data for PC1 are shown with a hatched background in diagram 400. The bytes or commands in the various sequences are labeled with a "C" and their ordinal number, and the data bytes are labeled with a "D" and their ordinal number.
[0022] While other proposed HB-DIMM architectures transmit and receive time-interleaved data between pseudo-channels, the illustrated scheme holds the 64-byte data response together, as shown in data sequence DQ. This scheme provides simplified signaling and allows optional use of the "2N" C / A signaling mode supported by DDR5 DRAM chips, as described further below.
[0023] 5 is a block diagram illustrating a high-bandwidth DIMM (HB-DIMM) 500, according to some embodiments. The HB-DIMM 500 is embodied on a PCB or other module format and includes multiple DRAM chips 502, multiple module data buffers (MDBs) 504, and a module register clock driver (MRCD) 506, and interfaces to a memory bus via pins shown generally along the bottom of the module. The HB-DIMM 500 includes two independently addressable pseudo channels, labeled "PC0" and "PC1," with the two ranks in pseudo channel PC0 labeled "R0" and "R1," and the two ranks in pseudo channel PC1 labeled "R0" and "R1."
[0024] Generally, the HB-DIMM 500 provides a data (DQ) rate of 12.8 Gbps in this embodiment, which is double the data rate of the DRAM chips 502 (6.4 Gbps).
[0025] As shown, pseudo channels PC0 and PC1 connect to the MDB 504 using separate 6.4 Gbps links of 8 pins, with 16 data pins connecting to each MDB 504. Thus, the MDB 504 can provide a capacity of 12.8 Gbps over the memory bus. The MRCD 506 circuitry includes a memory bus (host memory bus) interface for coupling to the memory bus and a memory interface connected to each of the DRAM chips 502. The MRCD 506 is capable of executing memory access commands received over the host bus by routing command / address (C / A) signals to the DRAM chips to provide at least two independently addressable pseudo channels. The MRCD 506 addresses each pseudo channel based on chip identifier (CID) bits derived from the C / A signals.
[0026] The HB-DIMM 500 has an operating speed of the data (DQ) lines of the memory bus (12.8 Gbps), the command / address (DCA_a) lines operate in a "1N" mode at 12.8 Gbps or a "2N" mode at 6.4 Gbps, and the host clock DCK supports slower speeds in the 2N mode by using double data rate clocking for the C / A bus as described with respect to FIG. 4 (6.4 GHz in 1N mode with single data rate clocking, 3.2 GHz in 2N mode with double data rate clocking).
[0027] The HB-DIMM 500 has several advantages over a similarly sized LRDIMM. First, the data lines from the DRAM chips 502 to the MDB 504 do not require decision feedback equalization (DFE) for higher operating frequencies because the QCA, QCS, QCK signals from the MRCD 506 to the DRAM chips 502 are clocked slower than the equivalent signals on the LRDIMM, as shown by the illustrated frequencies of 3.2 Gbps in 1N mode and 1.6 Gbps in 2N mode. Second, while a 1N frequency for the expected end of life (EOL) operating frequency of the DDR5 standard is achievable for the QCA, QCS, QCK signals due to the number of loads on each signal leaving the MRCD 506 (QCA has 20 loads in this configuration, and QCS and QCK both have 10 loads), such a frequency is not achievable in the LRDIMM configuration due to the increased number of loads. Third, both the MRCD 506 and the MDB 504 are configured to implement either two-rank logic and timing control per channel, or both quad-rank logic and timing control, allowing the HB-DIMM 500 to be configured with quad-ranks and no pseudo-channels (as shown in FIG. 6).
[0028] FIG. 6 is a block diagram illustrating a high bandwidth DIMM (HB-DIMM) 600, according to some embodiments. The HB-DIMM 600 shows the HB-DIMM 500 of FIG. 5 configured in quad-rank mode. The HB-DIMM 600 is embodied on a PCB or other module format and includes multiple DRAM chips 602, multiple data buffers (MDBs) 604, and a module register clock driver (MRCD) 606, and interfaces to a memory bus via pins generally shown along the bottom of the module. Generally, the HB-DIMM 600 provides a data (DQ) rate of 804 Gbps in this embodiment, double the data rate of the DRAM chips 602 (4.2 Gbps).
[0029] As shown, the quad-rank mode connects four ranks of DRAM chips 602 (labeled "R0", "R1", "R2", "R3") to the MDB 604 using individual 4.2 Gbps links of four pins, for a total of 16 data pins connecting to each MDB 604. The MRCD 606 circuitry includes a memory bus (host memory bus) interface for coupling to the memory bus and a memory interface connected to each of the DRAM chips 602. The MRCD 606 is capable of executing memory access commands received over the host bus by routing command / address (C / A) signals to the DRAM chips 601 to implement a quad-rank signaling configuration, which represents a programmable configuration of the HB-DIMM 500 (FIG. 5). Both the MRCD 606 and the MDB 604 are programmable to operate in either the quad-rank format or the dual-channel format illustrated in FIG. 5.
[0030] Figure 7 is a timing diagram 700 illustrating several signals illustrating the operation of the HB-DIMM 500 of Figure 5 in the 1N mode of operation. The first column lists the clock speed or data rate of each signal, and the second column lists the signal name. Shown are the host clock signal "DCK", a command sequence labeled "CMD", a C / A command sequence "DCA[6:0]" from the host to the MRCD, the local MRCD clock signal to the DRAM chip labeled "QCK", a command sequence for PC0 labeled "CMD PC0", a command sequence for PC1 labeled "CMD PC1", a data sequence from the MRCD to the DRAM chip for PC0 labeled "QACA[13:0]", a data sequence from the MRCD to the DRAM chip for PC0 labeled "QBCA[13:0]", a data sequence from the DRAM chip to the MDB for PC0 labeled "MDQ", another data sequence from the DRAM chip to the MDB for PC1 labeled "MDQ", and a data sequence on the memory bus data (DQ) lines labeled "DQ". Data bytes are labeled with a "D" and their ordinal number.
[0031] FIG. 8 is a timing diagram 800 illustrating several signals illustrating the operation of the HB-DIMM 500 of FIG. 5 in the 2N mode of operation. The signal names are the same as in FIG. 7. Two options for the frequency of the DCK clock are shown, labeled "DCK opt1" and "DCK opt2". For DCK opt1, the DCK frequency is half that of DCK opt2, at 3.2 GHz in this example versus 6.4 GHz. For DCK opt1, the C / A bus is clocked at double data rate with a 1:1 ratio relative to QCK, while for DCK opt2, the C / A bus is clocked at single data rate.
[0032] 9 is a table 900 containing various options available for configuring an HB-DIMM, according to various embodiments. Options are shown for using combinations of chip select (CS) and chip identifier (CID) values to access ranks of DRAM chips on the module in various configurations. Three configurations with two pseudo-channels "PC0" and "PC1" and a quad-rank configuration with no pseudo-channels are shown. The first column lists the ranks of the DRAM chips and the second column lists the pseudo-channels (PC).
[0033] In a configuration with one rank per channel (RPC) and four data lines per DRAM chip (1RPCx4), CS0 and CS1 are used to select between the pseudo channels. In a configuration with one RPC and eight data lines per DRAM chip (1RPCx8), CS0 and a CID value of 0 or 1 are used to select between the pseudo channels. In a configuration with two RPC and eight data lines per DRAM chip (2RPCx8), for rank 0, CS0 and a CID value of 0 or 1 are used to select between the pseudo channels, and for rank 1, CS1 and a CID value of 0 or 1 are used to select between the pseudo channels. In a configuration with quad rank and four data lines per DRAM chip (QRx4), CS0, CS1 and a CID value of 0 or 1 are used to select between the ranks.
[0034] FIG. 10 is a table 1000 including example clock speeds and data rates for HB-DIMMs, according to some embodiments. The illustrated configuration is shown for an 8800 MT / s (mega transfers per second) capable HB-DIMM implemented similarly to that of FIG. 5. The first column lists the signal or data rate. The second column shows the speed of a 2RPC HB-DIMM in 1N mode with a QCK to DCK ratio of 2:1 (i.e., FIG. 7). The third column shows the speed of a 2RPC HB-DIMM in 2N mode with a QCK to DCK ratio of 2:1 (i.e., DCK opt1 in FIG. 8). The fourth column shows the speed of a QR HB-DIMM in 2N mode with a QCK to DCK ratio of 1:1 (i.e., FIG. 8, DCK opt1). The configurable 2N mode allows for larger timing margins in systems where C / A timing is challenging, allowing DDR5 DRAM chips to be used in a variety of high-capacity DIMM modules that help extend the useful life of the DDR5 standard.
[0035] FIG. 11 is a table 1100 including example speed limits for various configurations of HB-DIMMs, according to various embodiments. The first row shows values for a QR LRDIMM (without HB-DIMM features) as a baseline for comparison. The other rows show HB-DIMM configurations, including a set of high-capacity or "tall" configurations, labeled on the left. The configuration labeled "Gear" includes a gear-down mode in which the HB-DIMM can be configured to force the DRAM address / command bus to use every other rising edge of the DDR5 memory bus clock. In the quad rank (QR) tall configuration, no pseudo channel is used, and the preferred embodiment employs CID bits to map addresses to ranks on the module.
[0036] 12 is a table 1200 including various HB-DIMM module configurations according to various embodiments. The last two rows show module configurations enabled by the technology herein, including modules providing quad-rank (QR) capabilities. As can be seen from the various speed and capacity options shown in FIGS. 9-13, the technology herein allows flexible configurations for various modes to accommodate different application requirements.
[0037] 13 is a block diagram illustrating a module register clock driver (MRCD) 1300 according to some embodiments. The MRCD 1300 is generally embodied as an RCD chip on an HB-DIMM module and is suitable for use in the HB-DIMMs of FIGS. 3, 5, 6, 16, and other HB-DIMM module configurations herein.
[0038] The MRCD1300 includes a reference voltage input labeled "Vref", C / A bus and parity inputs labeled "DCA[6:0]_A, DPAR_A", chip select inputs labeled "DCS[1:0]_A_n", a reset input labeled "DRST_n", a differential pair of DCK inputs labeled "DCK_t" and "DCK_c", a first DRAM C / A output labeled "QACA[13:0]_A", a second DRAM C / A output labeled "QBCA[13:0]_A", and a third DRAM C / A bus labeled "QBCA[13:0]_A". The MRCD 1300 includes an RCD for a first pseudo channel labeled "Channel A" and an MRCD 1320 for a second pseudo channel labeled "Channel B". The RCD for Channel A includes a control word state machine ("State Machine 1302") having control logic and chip select logic, and a clock circuit 1310.
[0039] The state machine 1302 includes a host bus interface that receives C / A bus and parity inputs DCA[6:0]_A, DPAR_A, and chip select input DCS[1:0]_A_n. The state machine 1302 also includes an input that receives a clock signal QCK from a clock circuit 1310. The first DRAM C / A output QACA[13:0]_A and the first DRAM chip select output QACS[1:0]_A_n are connected to a first rank of DRAM chips (e.g., R0 of PC0 in FIG. 5), and the second DRAM C / A output QBCA[13:0]_A and the second DRAM chip select output QBCS[1:0]_A_n are connected to a second rank of DRAM chips (e.g., R1 of PC0 in FIG. 5). The buffer command bus outputs BCOM[2:0]_A, BCS_A_n are connected to the MDB of the HB-DIMM (for example, 504 in FIG. 5).
[0040] The clock circuit 1310 receives a reset input "DRST_n" and differential DCK inputs labeled "DCK_t" and "DCK_c" and generates a clock signal QCK that is provided to the state machine 1302 and, optionally, to another identical state machine for channel B. The clock circuit 1310 includes a differential receiver 1314 and a phase locked loop (PLL) 1316. In operation, the differential receiver 1314 receives a differential clock signal DCK from the host memory bus, which is provided to the input of the PLL 1316. The output of the PLL 1316 provides a QCK (backside) clock for operating the MRCD and DB of the HB-DIMM. As indicated by the labels, the PLL 1316 is configurable to operate in two modes, one in which the DCK and QCK clock signals have a 2:1 ratio and one in which they have a 1:1 ratio. In the 2:1 mode, the PLL 1316 functions as a clock divider. Although example DCK frequencies of 3.2 GHz and 6.4 GHz are shown, these ratios may be used in embodiments using various frequencies of DCK.
[0041] The MRCD 1320 for channel B generally includes a second state machine 1302 that also receives a Vref input, C / A bus and parity inputs "DCA[6:0]_A, DPAR_A", chip select inputs "DCS[1:0]_A_n", and a reset input labeled "DRST_n" from the host controller interface. As shown, an optional second DCK can be received as an input if the host interface supports separate DCK signals for the different pseudo channels. In such an embodiment, a second clock circuit 1310 is present for channel B and generates its own QCK signal to drive the DRAM chips for the second pseudo channel.
[0042] In operation, the MRCD 1300 executes memory access commands received via the host bus through the host connector interface by routing command / address (C / A) signals to the memory chips to provide at least two independently addressable pseudo channels, with the RCD circuitry addressing each pseudo channel based on chip identifier (CID) bits derived from the C / A signals. In various embodiments, the state machine 1302 interprets and remaps selected CID bits of the C / A signals to generate output signals for accessing the DRAM chips according to the configurations described above. The state machine 1302 also generates MDB command signals for the HB-DIMM data buffers to properly provide the results of the memory access commands to each pseudo channel (in the case of a configuration having multiple pseudo channels) or each rank (in the case of a quad-rank configuration), as further described below with respect to FIG. 14. Specifically, the MRCD 1300 generates and sends to the MDB a command code indicating a pseudo channel selection for the memory access command based on the CID bits for the memory access command. The MRCD 1300 also includes a differential driver (not shown) for generating QCK clock outputs Q(OA)CK_A_t and Q(OA)CK_A_c for operating the DRAM chips of the HB-DIMM.
[0043] As shown in the timing diagrams of Figures 4, 7, and 8, the MRCD 1300 does not reorder memory access commands received over the host bus, but instead provides responses for each pseudo channel in sequence. The MRCD 1300 causes read responses for at least two pseudo channels to be transmitted over the data bus in a non-interleaved format. In some embodiments, the MRCD 1300 is programmable to operate in a first mode in which the data clock (DCK) and C / A pins of the host bus operate at a half-rate (2N) speed and the C / A pin operates in double data rate (DDR) mode, and a second mode in which the DCK and C / A pins of the host bus operate at a full-rate (1N) speed and the C / A pin of the host bus operates in single data rate (SDR) mode.
[0044] In some embodiments, the MRCD 1300 supports directed refresh management (DRFM) commands by converting a precharge per bank (PREpb) signal into a precharge per bank signal for a first pseudo channel of the multiple pseudo channels and converting a signal on a reserved for future use (RFU) bit derived from the C / A signal into a PREpb signal for a second pseudo channel of the multiple pseudo channels.
[0045] In some embodiments, the MRCD 1300 is programmable to operate in a quad-rank configuration that executes memory access commands received over the host bus by routing the C / A signals to the memory chips in quad-rank (QR) format. In such a case, the CID bits and chip select signals are mapped to logical ranks of the DRAM chips, for example, as shown in FIG. 9.
[0046] 14 is a mixed circuit and block diagram illustrating a module data buffer (MDB) 1400, according to some embodiments. The MDB 1400 is generally embodied in an MDB chip on an HB-DIMM module and is suitable for use in the HB-DIMMs of FIGS. 3, 5, 6, 16, and other HB-DIMM module configurations herein. The MDB 1400 generally includes a buffer controller 1410, a first DRAM interface 1430, a second DRAM interface 1440, and a host connector interface 1450.
[0047] The buffer controller 1410 has a first pair of differential clock inputs labeled "BCK_t" and "BCK_c," a control bus input labeled "BCOM[2:0]," a chip select input labeled "BCS_n," a reset input labeled "BRST_n," an impedance calibration output labeled "ZQCAL," and a number of outputs connected to the host connector interface 1450.
[0048] Each of the DRAM interfaces 1430 and 1440 has a differential pair of data strobe inputs / outputs labeled "MDQS0_t" and "MDQS0_c," a DRAM data bus input / output labeled MDQ[3:0], a reference voltage input, and a number of inputs / outputs connected to a host connector interface 1450. The host connector interface 1450 has a first differential pair of data strobe inputs / outputs labeled "DQS0_t" and "DQS0_c," a second differential pair of data strobe inputs / outputs labeled "DQS1_t" and "DQS1_c," and DRAM data bus inputs / outputs labeled "DQ[3:0], DQ[7:4]" (separated by lower and upper nibbles as shown to use separate data strobe signals for the upper and lower nibbles). The host connector interface 1450 also has multiple connections to a buffer controller 1410, as described further below.
[0049] In general, the MDB 1400 receives control signals via a buffered command bus to synchronize the operation of the MDB in executing memory access commands. The MDB 1400 executes additional BCOM (Buffered Command) code to identify a pseudo channel for each command and selects which of the DRAM interfaces 1430 or 1440 is used based on the BCOM code. Similarly, in a configuration that provides quad-rank capabilities for HB-DIMMs, the MDB performs device logical rank selection based on the additional BCOM code, for example as shown in FIG. 9. As indicated by the label "Additional BCOM Code for PC," the MDB 1400 receives and executes a command code indicating a pseudo channel selection for the memory access command based on the CID bits for the memory access command by the host connector interface 1450 and either of the DRAM interfaces 1430 or 1440, and connects to the DRAM chip for the indicated pseudo channel.
[0050] Each of the DRAM interfaces 1430 and 1440 connects to a DRAM chip on the HB-DIMM. In a configuration that supports pseudo channels, the DRAM interface 1430 connects to a DRAM chip for a first pseudo channel, and the second DRAM interface 1440 connects to a DRAM chip for a second pseudo channel. As shown, the host connector interface 1450 operates at a clock rate that is twice the data rate of the DRAM interfaces 1430 and 1440 and the control bus, in this example 6.4 GHz versus 3.2 GHz. In general, the HB-DIMM includes multiple MDBs 1400 that operate in response to commands from the MRCD to send and receive data to and from the DRAM chips via the host bus at the data rate of the DRAM chips.
[0051] The buffer controller 1410 includes a differential receiver 1411, multiple other receivers 1412, 1413, 1414, a PLL 1415, multiple latches 1416, 1417, 1418, a buffer control word block 1420, a command decoder and power management block 1422, a termination control block 1424, an impedance calibration block 1426, and a delay element 1428. The differential receiver 1411 receives a buffer block signal (generally BCK) from the MRCD and provides it to the PLL 1415 and the latches 1416, 1417, 1418. In this embodiment, the PLL 1415 includes a frequency multiplier that provides a clock signal for operating circuitry in the MDB 1400 at twice the frequency of BCK, as illustrated by exemplary frequencies of 3.2 GHz for BCK and 6.4 GHz for the output of the PLL 1415. Receiver 1412 receives the control bus input BCOM[2:0] and provides it to latch 1416, which forwards it to command decoder and power management block 1422. Receiver 1413 receives chip select input BCS_n and provides it to latches 1417 and 1418, which provide it to block 1422. Receiver 1414 receives reset input BRST_n and provides it to the reset input of latch 1418.
[0052] The command decoder and power management block 1422 has an input for receiving a control bus signal on input BCOM[2:0], an input for receiving a chip select signal on input BCS_n, a bidirectional connection to the buffer control word block 1420, and an output connected to a transaction controller 1451 of the host connection interface 1450. The buffer control word block 1422 has a bidirectional connection to the command decoder and power management block 1422, an output connected to a termination control block 1424, a bidirectional connection to an impedance calibration block 1426, and an output connected to the transaction control block 1451.
[0053] The termination controller 1424 generally controls the termination resistor 1460 of the host connector interface 1450. The termination controller 1424 has a first input connected to the output of the latch 1418, a second input connected to the PLL 1415, and a third input connected to the buffer control word block 1420. The impedance calibration block 1426 has a bidirectional connection to the buffer control word block 1420 and an output that provides a signal at the impedance calibration output ZQCAL.
[0054] Each of the DRAM interfaces 1430 and 1440 includes a pair of differential transmitters 1431, a differential receiver 1432, a termination resistor 1435, a receiver 1433, a transmitter 1434, a delay element 1436, and a latch 1437. Each of the differential transmitters 1431 has an input connected to a respective output transaction controller 1451, a control input connected to an output of the transaction controller 1451, and an output connected to the inputs / outputs MDQS0_t and MDQS0_c, respectively. The differential receiver 1432 has two inputs connected to the inputs / outputs MDQS0_t and MDQS0_c, respectively, an enable input connected to the transaction controller 1451, and an output connected through a delay element 1436 to a clock input of the latch 1437. The termination resistor 1435 selectively applies termination to the input / output to MDQS0_t, MDQS0_c, and MDQ[3:0]. Each of the receivers 1433 (only one of four shown) has an input connected to the input / output MDQ[3:0], an input connected to an MDQ reference voltage generator circuit 1438, and an output coupled to a read first-in-first-out buffer 1452 through a latch 1437. Each of the transmitters 1434 (only one shown) has an input connected to a write FIFO buffer 1453, an output connected to the input / output MDQ[3:0], and an enable input connected to the transaction controller 1451.
[0055] The host connector interface 1450 includes a transaction controller 1451, a read FIFO buffer 1542, a write FIFO buffer 1453, a pair of differential transmitters 1455, a differential receiver 1456, a termination resistor 1460, a transmitter 1457, a receiver 1458, and a latch 1454. Each of the differential transmitters 1455 has an input connected to a respective output transaction controller 1451, a control input connected to an output of the transaction controller 1451, and an output connected to input / outputs DQSn_t and DQSn_c, respectively (where "n" denotes 0 or 1, indicating that separate DQS signals are provided to the lower and upper nibbles of the data bytes processed by the MDB 1400). To simplify the drawing, only one transmitter / receiver for DQS is shown. The differential receiver 1456 has two inputs connected to the inputs / outputs MDQS0_t and MDQS0_c, respectively, an enable input connected to the transaction controller 1451, and an output connected to the clock input of the latch 1454. The termination resistors 1460 selectively apply termination to the inputs / outputs DQSn_t, MDQSn_c, DQ[3:0], DQ[7:4]. The receivers 1458 (only one of eight shown) each have an input connected to the inputs / outputs DQ[3:0], DQ[7:4], an input connected to an MDQ reference voltage generator circuit 1459, and an output coupled to the write FIFO buffer 1452 through the latch 1454. The transmitters 1457 (only one shown) each have an input connected to a read FIFO buffer 1453, an output connected to the input / output MDQ[3:0], and an enable input connected to the transaction controller 1451.
[0056] As shown, the DRAM interfaces 1430 and 144 operate at half the frequency of DCK. In operation, commands from the MRCD on the HB-DIMM are received via inputs BCOM[2:0] and decoded by the command decoder and power management block 1422. The command decoder interprets the received command and accesses the buffer control word block 1420 to provide the appropriate control word to the transaction controller 1451. In response to the buffer control word, the transaction controller 1451 configures the various transmitters, receivers and buffers to receive read data via either the DRAM interfaces 1430 and 1440 and transmit it via the host connector interface at twice the DRAM data rate, or receive write data via the host connector interface and write it to the DRAM via either the DRAM interfaces 1430 or 1440 at half the data rate at which it was received. The command decoder and power management circuit 1422 implements decoding logic for mapping the CID and CS bits to the DRAM chips, for example as shown in FIG. 9, to implement the pseudo channel or quad-rank configuration as described above.
[0057] 15 is a flowchart 1500 illustrating a process for operating an HB-DIMM, according to some embodiments. The illustrated process is suitable for use with HB-DIMM configurations provided herein that employ pseudo channels.
[0058] At block 1502, the process includes receiving a memory access command at a register clock driver (RCD) over a memory interface bus. At block 1504, based on a chip identifier (CID) bit of a C / A signal for the command, the RCD maps the memory access command to a selected memory chip of a plurality of memory chips that provide at least two pseudo channels. The RCD routes a command / address (C / A) signal for the memory access command to the selected memory chip.
[0059] In block 1506, the RCD prepares and sends a data buffer command code (BCOM) that indicates a pseudo channel associated with the selected memory access command over which the command is to be executed. Generally, block 1506 may be executed simultaneously with block 1504.
[0060] The data buffer is configured according to the data buffer command code to receive the response data from the memory, at block 1508. The received data is then driven onto the host memory interface bus on the appropriate data pins for the specified pseudo channel, at block 1510. Although a sequence involving a read from memory is shown, the same process applies in the reverse direction for a write command.
[0061] In general, a process of operating an HB-DIMM according to embodiments described herein may include receiving and executing multiple memory access commands, where the RCD circuit does not reorder the multiple memory access commands from the order of the commands received from the host bus. As described above with respect to Figures 7 and 8, the process in some embodiments may include transmitting read responses for at least two pseudo channels over the memory interface bus in a non-interleaved format.
[0062] According to some embodiments, the method may include entering a training mode with a different mapping of ranks. Known HB-DIMM designs have a problem in that they cannot properly train the four ranks of a quad-rank module due to hardware constraints, since current hardware is limited to encompass only two ranks. According to some embodiments, a mode is added to the RCD and DB circuits of various embodiments herein, whereby during training, each of the two host-side CS inputs is assigned to map to one of the two backside ranks to avoid logical rank limitations. During training, the process can change modes to go to all four ranks. In normal mode, the process employs CS / CID signals to address the ranks.
[0063] Figure 16 is a block diagram illustrating an HB-DIMM 1600, according to some embodiments. In general, the HB-DIMM 1600 is configured in a two ranks per pseudo channel configuration like the configuration of Figure 5, but includes two RCDs instead of a single RCD. In particular, the HB-DIMM 1600 has a primary RCD 1602 and a secondary RCD 1604.
[0064] In general, the proposed HB-DIMM configurations suffer from problems related to signal loading, which is related to capacity expansion and the associated double loading on all module signals. To help solve this problem, a second RCD buffer is added. This adds more loading to the external host-facing interface, but the additional loading is likely to be tolerable. An additional problem addressed by the implementation of FIG. 16 is the connectivity between the RCD buffer and the DB. The DB is controlled by the RCD. However, with two RCDs, the HB-DIMM module determines which controls the data buffer by designating the primary and secondary RCD. Both RCDs are connected to the host interface and therefore know which commands are performed by the other RCD. This allows the primary RCD to control both its own and the secondary RCD's DB while the secondary RCD is communicating with the DRAM chips for PC1. Another way to solve this problem is to add an auxiliary communication bus between the RCDs. Such a configuration, indicated by the optional bidirectional connection labeled "DB adjustment", requires more resources than the first solution.
[0065] The RCD circuit of the HB-DIMM 1600 employs a primary RCD 1602 for executing commands to a first set of DRAM chips for a first pseudo channel (PC0) of the multiple pseudo channels, and a secondary RCD 1604 for executing commands to a second set of DRAM chips for a second pseudo channel of the multiple pseudo channels. The primary RCD 1602 controls a data buffer chip (MDB) to instruct it to transmit data for the first and second pseudo channels.
[0066] An integrated circuit or integrated circuits including the reference voltage generating circuit described herein, or any portion thereof, may be described or represented by a computer accessible data structure in the form of a database or other data structure that may be read and used by a program to directly or indirectly manufacture the integrated circuit. For example, the data structure may be a behavioral level description or a register transfer level (RTL) description of the hardware functionality in a high level design language (HDL) such as Verilog or VHDL. The description may be read by a synthesis tool that may synthesize the description to generate a netlist that includes a list of gates from a synthesis library. The netlist includes a set of gates that also represent the functionality of the hardware that includes the integrated circuit. The netlist may then be placed and routed to generate a data set that describes the geometric shapes to be applied to a mask. The mask may then be used in various semiconductor manufacturing processes to manufacture the integrated circuit. Alternatively, the database on the computer accessible storage medium may be a netlist (with or without a synthesis library) or a data set, or Graphic Data System (GDS) II data, if desired.
[0067] Although specific embodiments have been described, various modifications to these embodiments will be apparent to those skilled in the art. For example, the specific timing of commands as shown in Figures 4, 7, and 8 may be changed. Furthermore, various features herein may be used in various subcombinations. The RCD may be embodied in a circuit as a single RCD chip or a pair of cooperating RCDs. An HB-DIMM may be constructed with an RCD circuit and a DB circuit that are programmable to change modes between the quad-rank and pseudo-channel configurations herein, or may be constructed to perform only one or the other.
[0068] Therefore, the appended claims are intended to cover all modifications of the disclosed embodiments that fall within the scope of the disclosed embodiments.
Claims
1. 1. A high-bandwidth dual in-line memory module (HB-DIMM) configured for coupling to a host bus, comprising: A plurality of memory chips; a plurality of data buffer chips coupled to respective sets of the plurality of memory chips and operable to transmit data of the plurality of memory chips over a host bus at a data rate twice that of the data rate of the plurality of memory chips; a register clock driver (RCD) circuit including a host bus interface for coupling to the host bus and a memory interface coupled to the plurality of memory chips, the RCD circuit being capable of executing memory access commands received over the host bus by routing command / address (C / A) signals to the plurality of memory chips to provide at least two independently addressable pseudo channels, the RCD circuit addressing each pseudo channel based on chip identifier (CID) bits derived from the C / A signals. High Bandwidth Dual In-line Memory Module (HB-DIMM).
2. the RCD circuit does not reorder memory access commands received over the host bus; The HB-DIMM of claim 1.
3. the RCD circuit causes read responses of the at least two pseudo channels to be transmitted over the host bus in a non-interleaved format. The HB-DIMM of claim 1.
4. the RCD circuitry is capable of executing memory access commands received via the host bus by routing C / A signals to the plurality of memory chips in a quad-rank (QR) format; the plurality of data buffer chips are capable of transmitting data from the plurality of memory chips over the host bus at a data rate of the plurality of memory chips in response to commands from the RCD circuit; The HB-DIMM of claim 1.
5. the RCD circuit is programmable to operate in a first mode in which a data clock (DCK) pin and a C / A pin of the host bus operate at a half-rate (2N) speed and the C / A pin operates in a double data rate (DDR) mode, and in a second mode in which the DCK pin of the host bus operates at a full-rate (1N) speed and the C / A pin of the host bus operates in a single data rate (SDR) mode; The HB-DIMM of claim 1.
6. the RCD circuit generates a command code indicating a pseudo channel selection of the memory access command based on the CID bit of the memory access command, and transmits the command code to at least one of the data buffer chips; The HB-DIMM of claim 1.
7. the RCD circuitry includes a first RCD for distributing commands to a first set of the plurality of memory chips and a second RCD for implementing commands to a second set of the plurality of memory chips, either of the first RCD or the second RCD acting as a primary RCD for controlling the data buffer chips to transmit data for both sets of the plurality of memory chips over the host bus; The HB-DIMM of claim 1.
8. the RCD circuit supports directed refresh management (DRFM) commands by converting a precharge per bank (PREpb) signal to a precharge per bank signal of a first one of the pseudo channels and converting a signal on a reserved for future use (RFU) bit derived from the C / A signal to a PREpb signal of a second one of the pseudo channels; The HB-DIMM of claim 1.
9. 1. A method of operating a high bandwidth dual in-line memory module (HB-DIMM), comprising: receiving a memory access command via a memory interface bus in a register clock driver (RCD) circuit; routing a command / address (C / A) signal of the memory access command to selected ones of a plurality of memory chips to provide at least two independently addressable pseudo channels in the RCD circuitry, the RCD circuitry addressing each pseudo channel based on a chip identifier (CID) bit of the C / A signal; preparing, in the RCD circuit, a data buffer command code indicating a pseudo channel associated with a selected memory access command through which the memory access command is to be executed; method.
10. receiving and executing a plurality of memory access commands; the RCD circuit does not reorder the memory access commands from an order in which the commands were received from the memory interface bus; 10. The method of claim 9.
11. transmitting read responses of the at least two pseudo channels over the memory interface bus in a non-interleaved format.
10. The method of claim 9.
12. the RCD circuit is programmable to operate in a first mode in which a data clock (DCK) pin and a C / A pin of the memory interface bus operate at a half-rate (2N) speed and the C / A pin operates in a double data rate (DDR) mode, and in a second mode in which the DCK pin of the memory interface bus operates at a full-rate (1N) speed and the C / A pin of the memory interface bus operates in a single data rate (SDR) mode; 10. The method of claim 9.
13. the RCD circuitry includes a first RCD for implementing commands to a first set of the plurality of memory chips for a first pseudo channel of the pseudo channels, and a second RCD for implementing commands to a second set of the plurality of memory chips for a second pseudo channel of the pseudo channels, either of the first RCD or the second RCD acting as a primary RCD for controlling a plurality of data buffer chips to transmit data for the first pseudo channel and the second pseudo channel.
10. The method of claim 9.
14. 1. A data processing system comprising: A host bus; a memory controller coupled to the host bus for executing memory access commands over at least two independently addressable pseudo channels; a high bandwidth dual in-line memory module (HB-DIMM) coupled to the host bus; The HB-DIMM is A plurality of memory chips; a plurality of data buffer chips coupled to the plurality of memory chips and capable of transmitting data of the plurality of memory chips over a host bus at a data rate that is twice the data rate of the plurality of memory chips; a register clock driver (RCD) circuit including a host bus interface for coupling to the host bus and a memory interface coupled to the plurality of memory chips, the RCD circuit executing memory access commands received via the host bus by routing command / address (C / A) signals to the plurality of memory chips for at least two independently addressable pseudo channels, and addressing each pseudo channel based on a chip identifier (CID) bit of the C / A signal. Data processing system.
15. the RCD circuit does not reorder memory access commands received over the host bus; 15. The data processing system of claim 14.
16. the RCD circuit causes the read responses of the at least two pseudo channels to be transmitted over the data bus in a non-interleaved format.
15. The data processing system of claim 14.
17. the RCD circuit is programmable to operate in a first mode in which a data clock (DCK) pin and a C / A pin of the host bus operate at a half-rate (2N) speed and the C / A pin operates in a double data rate (DDR) mode, and in a second mode in which the DCK pin of the host bus operates at a full-rate (1N) speed and the C / A pin of the host bus operates in a single data rate (SDR) mode; 15. The data processing system of claim 14.
18. the RCD circuit generates a command code indicating a pseudo channel selection of the memory access command based on the CID bit of the memory access command, and transmits the command code to at least one of the data buffer chips; 15. The data processing system of claim 14.
19. the RCD circuitry includes a first RCD for implementing commands to a first set of the plurality of memory chips, and a second RCD for implementing commands to a second set of the plurality of memory chips, either of the first RCD or the second RCD acting as a primary RCD for controlling the data buffer chip to transmit data for both sets of the plurality of memory chips; 15. The data processing system of claim 14.
20. the memory controller is programmable to function in a first mode in which it accesses the HB-DIMM to address a pseudo channel based on the CID bits, and a second mode in which it accesses a different HB-DIMM to access a rank of memory chips based on the CID bits.
15. The data processing system of claim 14.
21. the HB-DIMM comprises a first half providing a first memory channel including a first set of the plurality of memory chips, and a second half providing a second memory channel including a second set of the plurality of memory chips; the memory controller provides a first data clock (DCK) signal to the first memory channel and a second DCK signal to the second memory channel; 15. The data processing system of claim 14.
Citation Information
Patent Citations
Identification and access of each memory device within a memory channel
JP2008544427A
Techniques to Use Chip Select Signals for a Dual In-Line Memory Module
US20170278562A1
Systems and methods for utilizing DDR4-dram chips in hybrid DDR5-dimms and for cascading DDR5-dimms
US20180225235A1
Processing multi-cycle commands in memory devices, and related methods, devices, and systems
US20210166742A1
Cited By
High-bandwidth memory module architecture
US20250279125A1
