Start and stop read clock for synchronous memory
The programmable RCK mechanism in GDDR memory addresses inefficiencies in DDR DRAM by dynamically controlling RCK modes, enhancing efficiency and power management, especially for discrete GPUs.
Patent Information
- Application Number
- JP2024534114
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2022-06-27
- Filing Date
- 2022-11-30
- Publication Date
- 2024-12-05
- Estimated Expiration
- 2042-11-30
AI Technical Summary
Existing DDR DRAM technologies, such as GDDR6, require further improvements in memory efficiency and power management, particularly in scenarios with infrequent read operations, leading to unnecessary power consumption and signal interference.
The implementation of a programmable read clock (RCK) mechanism in GDDR memory that allows for 'read-only', 'always-on', and 'disabled' modes, controlled by mode register settings, enabling dynamic adjustment of RCK behavior based on command types and conditions.
This approach enhances memory efficiency by reducing power consumption during infrequent read operations, improving signal integrity, and optimizing bus utilization, particularly beneficial for discrete GPUs with high memory bandwidth demands.
Smart Images

Figure 0007595228000001_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] DDR DRAM is synchronous because it operates according to a free-running clock signal that synchronizes the issuance of commands from the host processor to the memory and thus the exchange of data between the host processor and the memory. The DDR DRAM can be used to synchronize commands according to a clock signal and generate a read data strobe signal. For example, the DDR DRAM receives write data using a center-aligned data strobe signal known as "DQS" provided by the host processor, and the memory captures data on both the rising and falling edges of the DQS. Similarly, the DDR DRAM provides read data synchronously with the edge-aligned DQS where the DDR DRAM provides the DQS signal. During a read cycle, the host processor internally delays the DQS signal to align it with the center portion of the DQ signal, typically by an amount determined at start-up by performing data eye training. Some DDR DRAMs, such as Graphics DDR, version 6 (GDDR6) DRAMs, receive both a main clock signal and a separate write clock signal, and programmably generate a read data strobe signal.
[0003] However, while these enhancements have improved the speed of DDR memory used for main memory in computer systems, further improvements are desired. [Brief description of the drawings]
[0004] [Figure 1] 1 is a block diagram illustrating a data processing system according to some embodiments. [Diagram 2] 2 is a block diagram of the GDDR DRAM of FIG. 1 according to some embodiments. [Diagram 3] 3 is a table showing mode register settings for receive clock modes of the memory of FIG. 2. [Figure 4]3 is a flowchart useful for understanding the operation of the memory of FIG. 2, according to some embodiments. [Diagram 5] 3 is a timing diagram illustrating receive clock timing characteristics of the memory of FIG. 2 in accordance with some embodiments. [Figure 6] 3 is a timing diagram illustrating further characteristics of the receive clock timing of the memory of FIG. 2 in accordance with some embodiments. [Figure 7] 3 is a timing diagram illustrating further characteristics of the receive clock timing of the memory of FIG. 2 in accordance with some embodiments. [Figure 8] 2 is a block diagram of a portion of the memory controller of FIG. 1 according to some embodiments. [Figure 9] FIG. 4 is a flow diagram of a process for operating a read clock according to some embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0005] 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.
[0006] The memory controller includes a command queue, an arbiter, and a read clock control circuit. The command queue includes a number of entries for holding incoming memory access commands. The arbiter is for selecting a command from the command queue for dispatch to the memory. The read clock control circuit is for monitoring the read command selected for dispatch to the memory and, in response to a specified condition, instructing the memory to change the state of the read clock from a first state in which the read clock remains active following a postamble period of the read command to a second state in which the read clock becomes inactive following the postamble period.
[0007] The method includes, in a memory controller, selecting memory access commands and dispatching them to a memory, the method monitors read commands selected for dispatch to the memory, and in response to a specified condition, instructs the memory to change a state of a read clock from a first state in which the read clock remains active following a postamble period of the read command to a second state in which the read clock becomes inactive following the postamble period.
[0008] The data processing system includes a memory, an integrated circuit including a data processor and a memory controller coupled to the data processor and the memory. The memory controller includes a command queue, an arbiter, and a read clock control circuit. The command queue includes a number of entries for holding incoming memory access commands. The arbiter is for selecting a command from the command queue for dispatch to the memory. The read clock control circuit is for monitoring a read command selected for dispatch to the memory and, in response to a specified condition, instructing the memory to change the state of the read clock from a first state in which the read clock remains active following a postamble period of the read command to a second state in which the read clock becomes inactive following the postamble period.
[0009] According to various embodiments disclosed herein, a memory provides the ability to start and stop a read clock (RCK) that the memory provides to a memory controller based on a command provided to the memory. Further, this behavior can be programmably enabled and disabled based on the value of one or more bits in a mode register.
[0010] 1 is a block diagram illustrating a data processing system 100 in accordance with 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 graphics DDR (GDDR) memory 140.
[0011] GPU 110 is a discrete graphics processor that has 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 (PHY) 115, and a GDDR 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, 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 111 issues SIMD 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 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, and a second downstream bidirectional port. 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 the side of the circuitry away from the data processor and towards the memory. The memory controller 114 controls the timing and ordering of data transfers to and from the DDR memory 130 and the GDDR memory 140. DDR and GDDR memories support asymmetric access, i.e., accesses to open pages in the memory are faster than accesses to closed pages. The memory controller 114 stores memory access commands and processes them out of order for efficiency, for example, by prioritizing accesses to open pages and deprioritizing frequent bus turnarounds from writes to reads and vice versa, while observing certain quality of service objectives.
[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 an implemented version of the DDR memory 130, such as DDR version 5 (DDR5), and performs training operations at the direction of the memory controller 114. Similarly, the GDDR 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 GDDR memory 200. The GDDR PHY 116 meets all specified timing parameters of an implemented version of the GDDR memory 140, such as GDDR version 7 (GDDR7), and performs training operations at the direction of the memory controller 114.
[0015] The inventors have discovered that the read clock (RCK) that a memory, for example, GDDR memory 200, provides to GDDR PHY 116 can be programmed to operate in certain new and advantageous ways. According to some embodiments, the memory has a "read-only" mode. In the read-only mode, the memory provides a read command on the RCK signal and begins toggling the RCK signal during a read preamble period before data transmission of the read command, and continues toggling at least until the end of the read postamble period following the read command. The read-only mode provides the ability to reduce power consumption during workloads where read operations are or may be infrequent.
[0016] GDDR memory 200 also has an "always-run" mode, in which GDDR memory 200 provides a RCK signal continuously as long as a write clock (WCK) is received from a host, e.g., a memory controller or memory PHY in a host processor chip. The always-run mode provides the ability for the host processor PHY to remain locked, avoiding the need for resynchronization during the preamble period.
[0017] According to some embodiments, the memory further comprises a disable mode in which the memory does not provide any read clock signal.
[0018] Figure 2 is a block diagram of the GDDR memory 200 of Figure 1, according to some embodiments. The GDDR memory 200 generally includes a control circuit 210, an address path 220, a memory array and page buffer 230, a data read path 240, a set of bond pads 250, and a data write path.
[0019] The control circuit 210 includes a command decoder 211, a mode register 212, and an RCK logic and state machine 213. The command decoder 211 decodes commands received from command and address pins (not shown in FIG. 2) into one of several supported commands defined by the memory's command truth table. One type of command decoded by the command decoder 211 is a mode register set (MRS) command. The MRS command causes the command decoder to provide a setting to the indicated mode register whose address input contains the setting. The MRS command has been known in the context of DRAM for quite some time and differs between different GDDR DRAM versions. The mode register 212 stores the programmed settings and possibly outputs information about the GDDR DRAM. The RCK logic and state machine 213 has a first input connected to an output of the command decoder 211, a second input connected to a particular output of the mode register 212, and an output. As will be further explained, the RCK logic and state machine 213 processes the read clock flag. The read clock flag indicates on the fly the read clock behavior after the read postamble period, i.e. during the "interamble" period. The read clock flag can be encoded as a command signal, a separate signal, or in any other known manner.
[0020] The address path 220 receives multi-bit address signals and includes an input buffer 221 and an address latch 222 for each address signal, a set of row decoders 223, and a set of column decoders 224. The input buffer 221 receives and buffers a corresponding multi-bit address signal and provides a multi-bit buffered address signal in response. The address latch 222 has an input connected to the output of the input buffer 221, an output, and a clock input that receives a signal labeled "WCK". The address latch 222 latches the bits of the buffered address on a clock edge, e.g., a rising edge, and serves as a write clock during write commands as well as a main clock used to capture commands. The row decoder 223 has an input connected to the output of the address latch 222, and an output. The column decoder 224 has an input connected to the output of the address latch 222, and an output.
[0021] The memory arrays and page buffers 230 are organized into sets of individual memory arrays known as separately addressable banks. For example, the GDDR memory 200 may have a total of 16 banks. Each bank can have only one "open" page at a time, which has its contents read into the corresponding page buffer for faster read and write access. The row decoder 223 selects the row in the bank accessed during the activate command, and the contents of the indicated row are read into the page buffer and the row is ready for read and write access. The column decoder 224 selects a column of the row according to a column address.
[0022] The data read path 240 includes a read queue 241, a read latch 242, an output buffer 243, a delay locked loop (DLL) 244, and RCK and
[0023]
number
[0024]
number
[0025] Write data path 260 includes an input buffer 261, a write latch 262, and a write queue 263. Input buffer 261 has an input connected to a set of bond pads 250 labeled “DQ”, and an output. Write latch 262 has an input connected to the output of input buffer 261, and an output. Write queue 263 has an input connected to the output of write latch 262, and an output connected to the memory array and page buffer 230.
[0026] In operation, the GDDR memory 200 allows for simultaneous operations on the memory banks, and in one embodiment, the GDDR memory 200 is compatible with any of the Double Data Rate (DDR) standards issued by the Joint Electron Device Engineering Council (JEDEC), such as the emerging Graphics DDR, version 7 (GDDR7) standard. To access data, a memory accessing agent, such as the GPU 110, activates a row in the memory bank by issuing an activate (ACT) command. In response to the ACT command, data from memory cells along the selected row are stored in the corresponding page buffer. In DRAM, data reads are destructive to the contents of the memory cells, but a copy of the data is stored in the page buffer. After the memory controller 114 finishes accessing the data in the selected row of the bank, it closes the row by issuing a precharge (PRE) command (or a write or read command with automatic precharge, or a precharge all command). The PRE command causes the data in the page buffer 124 to be rewritten to that row in the selected bank, and then allows another row to be activated. These operations are conventional for DDR memories and are described in various JEDEC standard documents and will not be described further.
[0027] However, according to various embodiments disclosed herein, GDDR memory 200 includes a modified set of mode registers 212 that adds a mode register field that can be used to define the behavior of the RCK signal that memory 200 provides along with the data accessed during a read cycle, as compared to existing standards such as GDDR6. In addition, memory 200 will transmit RCK (and optionally
[0028]
number
[0029] Figure 3 shows a table 300 illustrating mode register settings for the read clock mode of the memory of Figure 2. Table 300 shows the values of different bits or bit fields of a 12-bit mode register, the 12 bits corresponding to the address signals into which the mode register is loaded. Table 300 has six columns including an OP code (operation code) column which corresponds to particular bits of the mode register, a function column which identifies the function defined by the corresponding bits, an OP code value column which specifies different values of the OP code, and a description column which identifies the meaning of the different OP code values.
[0030] Mode register bits [1:0] are labeled "RCKMODE" and identify the selected RCK mode. A value of 00b (binary) identifies a disabled mode in which RCK is not provided by memory 200. This mode is the default mode.
[0031] A value of 01b indicates read-only mode. As explained further below, in read-only mode, RCK is provided during one or more read cycles, with each read cycle including both a preamble and a postamble. When read-only mode is selected, the interamble behavior is such that successive reads are spaced apart for greater than a minimum amount of time, i.e., at least t CCD +1 RCK cycles, where t CCDis the minimum inter-command delay time. In general, during read data mode, RCK starts the preamble period before the transfer of data in a read cycle and ends the preamble period after the read cycle. In particular, it starts toggling simultaneously with data transfer for a read command (RD), a read with auto precharge command (RDA), and a read training (RDTR) command. It stops on a clear condition. In some embodiments, the clear condition includes receipt of a write command (write command (WR), write command with auto precharge (WRA) or write training (WRTR) command), receipt of an all-banks idle indicator, or entry into a powered down state.
[0032] A value of 10b indicates an always-running mode, in which the RCK runs continuously as long as the WCK used to generate the RCK is received by memory 200.
[0033] The value of 11b is reserved (RSVD) but allows the definition of new modes that provide an RCK signal to be added in the future using this mode register structure.
[0034] Mode register bit [2] defines the read clock type (RCKTYPE). A value of 0b indicates that the GDDR memory 200 provides the RCK signal as a single-ended signal, i.e.
[0035]
number
[0036]
number
[0037] Mode register bits [4:3] define the length of the static preamble period. To allow the memory controller to lock onto the preamble, each preamble period has a static period, a slow period, and a fast period. In the example preamble signals of Figures 3 and 4, static states 304 and 404 represent the static period, initial toggle states 306 and 406 represent the slow period, and final toggle states 308 and 408 represent the fast period. During the static period, the read clock signal is driven in its inactive state, i.e., RCK is driven low,
[0038]
number
[0039] Mode register bit [5] defines the RCK level setting (RCKLEVEL), which in this example provides two possible value pairs for the RCK level during the RCK preamble and RCK operation. A value of 0b defines a +3 / -3 value for the PAM driver of the RCK signal, and a value of 1b defines a +1 / -1 value. Although these value pairs are shown as examples, in various embodiments, configuration settings can be used to program any desired value pair, including defining different value pairs for each phase of the preamble.
[0040] Mode register bits [7:6] define the length of the fast preamble period. A value of 00b indicates a fast preamble period of 0 clock cycles, i.e. no fast preamble period. Values of 01b, 10b and 11b define static periods of 2, 4 and 6 cycles, respectively.
[0041] Mode register bit [8] is undefined and reserved for future use (RFU).
[0042] Mode register bits [10:9] define the length of the slow preamble period. A value of 00b indicates a slow preamble period of 0 clock cycles, i.e. no slow preamble period. Values of 01b, 10b and 11b define static periods of 1, 2 and 3 cycles. Note that while the fast and slow preamble periods are independently programmable, if the OP code bits [7:6] and [10:9] have the same value then the fast and slow preambles are the same length of time.
[0043] Mode register bit
[11] is undefined and is marked as RFU.
[0044] It should be apparent that these mode register encodings are just one possible way of encoding these values, and that other encodings are possible, e.g., instead of using a dedicated mode register, these bits could be distributed among multiple mode registers, e.g., in otherwise unused or reserved bit positions.
[0045] Figure 4 shows a flowchart 400 useful for understanding the operation of memory 200 of Figure 2, according to some embodiments. Flowchart 400 governs the RCK start and stop behavior and interamble behavior when RCKMODE is set to read-only. Flowchart 400 defines a flag known as the RCKON flag.
[0046] Flow begins at action box 410 when a first command is received. Decision box 420 determines whether the command is a read command (such as a read command (RD), a read command with auto precharge (RDA), or a read training (RDTR) command), and if so, determines the state of the RCKON flag. If the command is not a read command, or if it is a read command and the RCKON flag is clear, flow proceeds to action box 430. In action box 430, RCK stops toggling after the read postamble period.
[0047] If the command is a read command and the RCKON state variable is set to 1, flow proceeds to action box 440. In action box 440, memory 200 continues to toggle the RCK signal after the postamble period of the read command. From this point on, the state of RCKON becomes don't care. Flow proceeds to decision box 450 to determine whether a clear condition has been received. In some embodiments, the clear condition includes one or more of receiving an explicit read clock stop command by memory 200, receiving a write command by memory 200 (e.g., one or more of a write command (WR), a write with auto precharge command (WRA), or a write read training (WRTR) command), receiving a mode register set command by memory 200, detecting an all-banks idle state by memory 200, and detecting a powered down state of memory 200. If a clear condition is not received, flow returns to decision box 450. If a clear condition is received, flow proceeds to action box 460. In action box 460 , state variable RCKON is cleared to 0, RCK stops toggling after the read postamble, and flow returns to decision box 420 .
[0048] 5 illustrates a timing diagram 500 illustrating characteristics of the receive clock timing of the memory 200 of FIG. 2, in accordance with some embodiments. In the timing diagram 500, the horizontal axis represents time in picoseconds (ps) and the vertical axis represents the amplitude of various signals in volts. Shown along the vertical axis are three signals or signal groups of interest: a command signal, an RCK signal, and a data signal. The dashed lines indicate low-to-high and high-to-low transitions of the RCK signal, corresponding to various time points.
[0049] In the example shown in timing diagram 500, mode register 300 is programmed with RCKMODE=Read Only, RCKTYPE=Single-Ended, RCKPRE_Static=4, RCKPRE_LS=1, and RCKPRE_HS=2. Timing diagram 500 illustrates the issuance of a read command, labeled "RD," on the second RCK transition with the RCKON attribute set to 1. Due to read latency, memory 200 does not provide read data until the 25th clock cycle. Therefore, prior to this RCK cycle, memory 200 provides the preamble defined in table 300.
[0050] In this example, the burst length is 16, and the memory 200 stores t CCDMIN+1 can accept other commands in CCDMIN+7 It becomes necessary to define the interamble behavior rather than issuing it until the last data transmission of the first cycle. As seen here, the interamble is a combination of continuous toggling RCK after the last data transmission of the first cycle, followed by a low period of the slow preamble of the second ready cycle, followed by a fast toggling of the preamble of the fast part.
[0051] 6 is a timing diagram 600 illustrating further characteristics of the receive clock timing of memory 200 of FIG. 2, in accordance with some embodiments. In timing diagram 600, the horizontal axis represents time in picoseconds (ps) and the vertical axis represents the amplitude of various signals in volts. As discussed above, the command signal, the RCK signal, and the data signal are shown along the vertical axis. The dashed lines indicate low-to-high and high-to-low transitions of the RCK signal, denoted as "t1" through "t 68 " corresponds to the point in time specified.
[0052] In the example shown in timing diagram 600, mode register 300 is programmed with RCKMODE=read only, RCKTYPE=single ended, RCKPRE_Static=4, RCKPRE_LS=1, and RCKPRE_HS=2. Timing diagram 600 illustrates the issuance of a read command RD at t2 where RCKON=0, and RCKON has not been previously set from a previous clear condition. In this case, the RD command causes memory 200 to issue a preamble as defined in the mode register, perform a read burst cycle with RCK toggle, and follow the read burst cycle with a postamble. In this case, the postamble includes a trailing static portion of 2 clock cycles to terminate the postamble period. Thus, when RCKMODE=read only, the host processor will issue RCK and
[0053]
number
[0054] 7 is a timing diagram 700 illustrating further characteristics of the receive clock timing of memory 200 of FIG. 2, according to some embodiments. In timing diagram 700, the horizontal axis represents time in picoseconds (ps) and the vertical axis represents the amplitude of various signals in volts. As discussed above, the command signal, the RCK signal, and the data signal are shown along the vertical axis. The dashed lines indicate low-to-high and high-to-low transitions of the RCK signal, corresponding to various time points.
[0055] In timing diagram 700, memory 200 receives a RD command with the RCKON attribute set to 1 at a time prior to the time shown in timing diagram 700. Memory 200 provides a preamble for the RCK signal in response to the RD command. However, after the data transfer is complete, memory 200 continues to toggle the RCK signal, which in the example of timing diagram 700 forms an extended interamble period. As shown in FIG. 7, memory 200 receives a WR command (WR, WRA, or WRTR) before a subsequent read command is received. In this case, RCK logic and state machine 213 decodes the write command and stops toggling the RCK signal after the write command latency period.
[0056] By providing the ability to start and stop RCK toggling, memory 200 provides a read clock signal that is a hybrid of a strobe and a clock signal. Memory 200 also provides a mechanism for the memory to inhibit itself from continuously outputting RCK after a clear condition. This capability allows a DLL in the memory controller to remain locked during a streak of read commands but stop toggling to save power in response to a clear condition. The memory controller reorders commands to improve memory bus usage efficiency, grouping commands of the same type to reduce the frequency of bus turnaround from read to write and from write to read. For example, efficiency of bus usage is particularly important for discrete GPUs, which users often configure to push the limits of performance. This capability provides several advantages.
[0057] First, it allows memory controllers to operate more efficiently by simplifying designs that would otherwise be required due to interamble complexity. In particular, interamble calculations can be simplified or eliminated.
[0058] Second, it preserves the signal integrity of the signal that toggles when RCK is not used. For example, this mechanism allows the user to avoid continuous generation of RCK in response to receiving a write command. Thus, the write cycle is more robust and there is less signal interference and crosstalk, providing more timing margin for capturing data in memory.
[0059] Third, by suppressing the switching of the RCK signal during streaks of write accesses, it saves switching power of high speed signal switching of external signals that are typically driven on a printed circuit board when not needed.
[0060] Fourth, the host processor can set the RCKON flag used in memory in a variety of ways, providing flexibility in implementation. For example, the host processor can issue a mode register set command to program an unused or vendor-specific RCKON bit in the mode register. It can issue an explicit RCKON command or include the RCKON attribute in the RD command encoding. It can also activate a new dedicated signal line.
[0061] FIG. 8 is a block diagram of a portion 800 of the memory controller 114 of FIG. 1 according to some embodiments. The illustrated portion 800 is suitable for use in a memory controller for GDDR memory and other similar DRAM types with a controllable read clock. The portion 800 includes a command queue 810, an arbiter 820 with an associated read clock control circuit 825, and a multiplexer 840. The command queue 810 stores incoming memory access requests, which may be read or write cycles to any bank in the memory, generated by any of the memory access agents in the data processing system 100. In the example of FIG. 8, the command queue 810 has a total of 32 entries for all banks and ranks, with each entry storing the access address, a tag indicating its relative age, and other metadata. Arbiter 820 is bidirectionally connected to each entry in command queue 810 to read its attributes, has an additional input for receiving the access when it is dispatched, has outputs for providing protocol commands such as ACT and PRE commands, and has a control output for selecting one of the 32 entries to send to interface 212. Multiplexer 840 has 32 inputs connected to each entry in command queue 810, an additional input connected to the output of arbiter 820, a control input connected to the output of arbiter 820, and an output for providing the dispatched access to the memory PHY.
[0062] The read clock control circuit 825 has a bidirectional connection to the arbiter 820 and several inputs connected to respective outputs of the arbiter 820. The read clock control circuit 825 includes a number of command entries 828 for tracking the currently selected command and at least one command available for selection (labeled "n" and "n-1", respectively), and read clock control logic for implementing the tracking functionality as described below. In various embodiments, the read clock control circuit 815 is typically implemented with the command selection logic of the arbiter for tracking the available commands.
[0063] In operation, the read clock control circuitry 825 receives an indicator from the arbiter 820 when a read command has been selected for dispatch to the memory interface queue, and checks specified conditions that are generally used to control the behavior of the RCK, as described further below with respect to Figure 9. In embodiments in which the read clock is controlled by a flag or tag associated with the read command, the read clock control circuitry 425 updates the tag in the command queue 410 to set the specified value of the tag and specify the mode of the read clock.
[0064] The command queue 810 stores incoming accesses and assigns a tag indicating the command's relative age. The arbiter 820 decides whether to schedule any pending accesses in the command queue 810 for dispatch to the memory interface queue 214 based on a set of policies such as timing eligibility, age, fairness, and activity type. The arbiter 820 includes a page table, not shown in FIG. 8, to indicate the open pages in each bank and rank of the memory system. In general, the arbiter 820 can increase the efficiency of the memory system bus by scheduling a streak of commands that includes multiple accesses to the same row together while delaying older accesses to different rows in the same bank. The read clock mode control is particularly advantageous during a streak of read commands that may include sending multiple read commands to multiple open rows in various memory banks.
[0065] FIG. 9 illustrates a flow diagram 900 of a process for operating a read clock, according to some embodiments. It should be noted that the illustrated process is merely one suitable way to operate a controllable read clock. Also, the read clock may remain in a "read only" state, where the clock is deactivated following each postamble. However, such a state is often not suitable for scenarios where an arbiter selects multiple read commands in a sequence, often referred to as a streak. In such cases, the postamble period of the read clock may overlap with the preamble period of the next read clock. Furthermore, in scenarios where read commands are not separated by write commands, but by ACT or other commands, it may be advantageous to allow the read clock to remain active after the postamble period. The illustrated process provides a technique for controlling the read clock from a memory controller to achieve such functionality.
[0066] At block 902, the process begins when the arbiter 820 selects a read command for dispatch to a memory that includes a controllable read clock as described above. At block 904, the read clock is set to a first state that remains active following a postamble after the read command is executed. This can be done with an explicit command or by indicating a command by a read clock flag that accompanies the read command sent at block 902. In this embodiment, the process assumes that the first selected read command is part of a streak or another situation in which it is beneficial to keep the read clock activated between read commands. In other embodiments, the process can first check whether the read command is expected to be part of a streak, for example, by checking whether other read commands are available to be selected by the arbiter for an open page in the memory.
[0067] At block 906, the process continues selecting commands for dispatch to the memory. The read clock monitoring circuit monitors the commands selected for dispatch to determine whether to change the state of the read clock. If the specified condition is met, as shown at block 908, the process proceeds to block 910, where it instructs the memory to change the read clock to a second state. If the specified condition at block 908 is not met, the process continues selecting commands at block 906 and leaves the read clock in the first state.
[0068] In some embodiments, the read clock control circuitry instructs the memory to change the state of the read clock by setting a read clock flag associated with the current read command to a specified value. Note that the read clock flag may be included with each read command. In such an embodiment, the read clock flag value for the first state may be repeated with each command selected in block 906. Then, if it is desired that the read clock state be changed, in block 910, the read clock flag value for the current command is changed to a new specified value indicating a second state. In other embodiments, in blocks 904 and 910, a separate command may be issued to the memory to change the read clock state.
[0069] Referring to block 908, in some embodiments, the specified condition is that a gap of at least a predetermined size occurs between the current read command dispatched to the memory and a subsequent read command dispatched to the memory. For example, a write command separating two read commands may satisfy the specified condition. In such a case, the read clock is deactivated following the preamble period to avoid causing signal interference to the write data being sent to the memory. However, if a single command that is not a write command separates two read commands, in some embodiments, the specified condition may not be satisfied. In some embodiments, the specified condition is that at least a specified number of write commands occur between the current read command dispatched to the memory and a subsequent read command dispatched to the memory. In some embodiments, the specified condition is that the current read command is the last command in a streak of read commands selected by the arbiter. Additionally, although a "specified condition" is referred to, this may include a combination of conditions or two or more alternative conditions.
[0070] Following the command to change the read clock to a second state in block 910, the process proceeds to block 912, where the memory receives and implements the command to change the read clock state to inactive following the postamble period of the current read command.
[0071] The memory described herein, or portions thereof, may be embodied in one or more integrated circuits, any of which 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 by a program and used directly or indirectly to manufacture an 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 comprises 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.
[0072] Although specific embodiments have been described, various modifications to these embodiments will be apparent to those skilled in the art. For example, the host processor can use various techniques to set the read clock on an attribute (RCKON), such as an explicit command, an unused bit in the command encoding, a mode register setting, etc. Although the RCK programming has been described in the context of GDDR memory, other types of circuits, integrated or discrete, can use a clock signal with the hybrid behavior described for the RCK signal. The RCK signal can also be used in other types of memory, including DDR and high bandwidth (HBM) memories. Additionally, various levels of granularity of preamble and postamble behavior can also be supported. Bits defining the supported RCK modes can be defined in a dedicated mode register or can be set in various bit positions in different mode registers. These bit positions may be previously unused and reserved for future use, or may be dedicated for customer-specific use.
[0073] 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 memory controller, comprising: a command queue including a plurality of entries for holding input memory access commands; an arbiter for selecting a command from the command queue for dispatch to a memory; and a read clock control circuit that monitors selected read commands for dispatch to the memory and, in response to a specified condition, instructs the memory to change the state of a read clock from a first state in which the read clock remains active after a postamble period of a read command to a second state in which the read clock becomes inactive after the postamble period. Memory controller.
2. the specified condition being that a gap of at least a predetermined size occurs between a current read command dispatched to the memory and a subsequent read command dispatched to the memory; The memory controller of claim 1 .
3. the specified condition being that at least a specified number of write commands occur between a current read command dispatched to the memory and a subsequent read command dispatched to the memory; The memory controller of claim 1 .
4. the specified condition is that the current read command is the last command in a streak of read commands selected by the arbiter; The memory controller of claim 1 .
5. the read clock control circuit instructs the memory to change the state of the read clock by setting a read clock flag associated with a current read command to a specified value; The memory controller of claim 1 .
6. the read clock control circuit instructs the memory to change the state of the read clock by sending an explicit command, separate from a read command, to the memory; The memory controller of claim 1 .
7. the read clock control circuit, in response to the specified condition not being present, causes the read clock to remain in the first state. The memory controller of claim 1 .
8. leaving the read clock in the first state includes setting a read clock flag associated with the current read command to a specified value. The memory controller of claim 7.
9. 1. A method comprising: selecting, in a memory controller, memory access commands and dispatching them to the memory; monitoring selected read commands for dispatch to the memory; instructing the memory to change a state of a read clock from a first state in which the read clock remains active after a postamble period of a read command to a second state in which the read clock becomes inactive after the postamble period. method. Instructing the memory to change the state of the read clock is based on a specified condition; the specified condition being that a gap of at least a predetermined size occurs between a current read command dispatched to the memory and a subsequent read command dispatched to the memory; 10. The method of claim 9. Instructing the memory to change the state of the read clock is based on specified conditions; the specified condition being that at least a specified number of write commands occur between a current read command dispatched to the memory and a subsequent read command dispatched to the memory; 10. The method of claim 9. Instructing the memory to change the state of the read clock is based on specified conditions; the specified condition is that the current read command is the last command in a streak of read commands selected by the memory controller; 10. The method of claim 9.
13. instructing the memory to change the state of the read clock includes setting a read clock flag associated with the current read command to a specified value.
10. The method of claim 9.
14. The instructing the memory to change the state of the read clock is based on specified conditions; responsive to the specified condition not existing, causing the read clock to remain in the first state.
10. The method of claim 9.
15. leaving the read clock in the first state includes setting a metadata flag associated with the current read command to a specified value.
15. The method of claim 14.
Citation Information
Patent Citations
Memory interface circuit and clock control method
JP2006099569A
Semiconductor memory device, method of controlling read preamble signal thereof, and data transmission system
JP2010170597A
Memory system
JP2011008779A
Extended data clock operation in memory
JP2022522012A
Multiplexed bus with multiple timing signals
US20040064660A1
