A method for controlling high temperature of GPU memory

Abstract

This invention provides a method for controlling high-temperature GPU memory, comprising: periodically reading the refresh rate of the LPDDR memory chips within the GPU board using software or a combination of software and hardware; determining whether the LPDDR memory chips are operating in a high-temperature environment based on the refresh rate; if operating in a high-temperature environment, performing read / write tests on the LPDDR memory chips to obtain the test results and the operating status of the LPDDR memory chips; and taking corresponding strategies to reduce the operating temperature of the LPDDR memory chips based on their different operating states. This invention detects whether the memory chips are operating in a high-temperature environment by detecting their own refresh rate, and enables LPDDR to provide early warning and self-adjustment in the face of high-temperature environments, thereby improving the overall stability and maintainability of the GPU board.

Description

Technical Field

[0001] This invention relates to the field of GPU memory testing, and more specifically, to a method for controlling high temperature of GPU memory. Background Technology

[0002] See Figure 1 The GPU board consists of three major IP subsystems: GPU module, DC display module, and LPDDR storage module. The GPU core is responsible for rendering graphics data, the DC display is responsible for outputting the rendered image data to the terminal display screen, and the LPDDR storage module is responsible for storing the rendered graphics data.

[0003] Double Data Rate Synchronous Dynamic Random Access Memory (DDR) is a common computer memory technology. Low Power Double Data Rate SDRAM (LPDDR) is a memory technology specifically designed for mobile electronic products. Compared to DDR, LPDDR offers lower power consumption, higher performance, and smaller size. Therefore, GPUs currently use LPDDR more often as their video memory to reduce board power consumption. The stability of LPDDR is crucial, and a key factor affecting LPDDR stability is its operating temperature.

[0004] Current solutions typically integrate a temperature sensor inside the GPU board to detect the board's operating temperature. However, this approach has a drawback: the temperature measured is the ambient temperature of the board, not the operating temperature of the LPDDR memory chips themselves. The two are not equivalent, making it impossible to accurately characterize the state of the LPDDR memory chips. Generally, the operating temperature of LPDDR memory chips is higher than this. Furthermore, it's impractical to install a temperature sensor directly on the LPDDR memory chips. Therefore, existing solutions cannot accurately determine whether LPDDR memory chips are operating in a high-temperature environment. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a method for controlling high temperature of GPU memory, overcoming the inability of the prior art to obtain the high temperature situation of memory chips during operation.

[0006] According to a first aspect of the present invention, a method for controlling high temperature of GPU memory is provided, comprising:

[0007] The system allocates a portion of dynamic memory for use as video memory addresses;

[0008] The refresh rate of the LPDDR memory chips inside the GPU board is periodically read using software or a combination of software and hardware, and the LPDDR memory chips are determined to be in a high-temperature operating environment based on the refresh rate of the LPDDR memory chips.

[0009] Under the condition that the LPDDR memory chip is in a high-temperature working environment, read and write tests are performed on the LPDDR memory chip using software to obtain read and write test results.

[0010] Based on the read / write test results, the working status of the LPDDR memory chip is determined;

[0011] Based on the different operating states of the LPDDR memory chips, corresponding strategies are adopted to reduce the operating ambient temperature of the LPDDR memory chips.

[0012] Based on the above technical solution, the present invention can also be improved as follows.

[0013] Optionally, the refresh rate of the LPDDR memory chips inside the GPU board is periodically read using software, and the LPDDR memory chips are determined to be operating in a high-temperature environment based on the refresh rate of the LPDDR memory chips, including:

[0014] A separate thread is created in kernel mode. The thread sends an MRR command through the controller DMC at preset intervals to read the value of the MR4 register in the LPDDR memory chip. This value represents the refresh rate inside the LPDDR memory chip.

[0015] Based on software-based determination, when the refresh rate inside the LPDDR memory chip reaches a preset refresh rate threshold multiple times consecutively, the LPDDR memory chip is in a high-temperature operating environment; otherwise, the LPDDR memory chip is in a non-high-temperature operating environment.

[0016] Optionally, based on a hardware-software hybrid approach, the refresh rate of the LPDDR memory chips inside the GPU board is periodically read, and the LPDDR memory chips are determined to be operating in a high-temperature environment based on the refresh rate of the LPDDR memory chips, including:

[0017] A hardware monitoring module is implemented on the system bus. A timer is implemented in the hardware monitoring module. When the timer expires, the hardware monitoring module automatically sends an MRR command to read the value of the MR4 register in the LPDDR memory chip. This value represents the refresh rate inside the LPDDR memory chip.

[0018] The hardware monitoring module determines that when the refresh rate inside the LPDDR memory chip reaches the preset refresh rate threshold multiple times consecutively, the LPDDR memory chip is in a high-temperature operating environment; otherwise, the LPDDR memory chip is in a non-high-temperature operating environment.

[0019] When the hardware monitoring module determines that the LPDDR memory chip is in a high-temperature operating state, it triggers an interrupt to the software. After the software receives the interrupt, it disables the interrupt.

[0020] Optionally, when the LPDDR memory chip is operating in a high-temperature environment, read / write tests are performed on the LPDDR memory chip using software to obtain read / write test results, including:

[0021] When the LPDDR memory chip is in a high-temperature operating environment, a memory block with a different address from the memory chip is allocated. The read and write accesses of the LPDDR memory chip through the system bus are obtained based on the software, and the read and write test results of the LPDDR memory chip are obtained based on the read and write accesses.

[0022] Optionally, the read / write test results of the LPDDR memory chip are obtained based on the read / write access, including:

[0023] If the read / write test result of the LPDDR memory chip is passed, it is determined that the LPDDR memory chip is in normal working condition; if the read / write test result of the LPDDR memory chip is failed, it is determined that the LPDDR memory chip is in abnormal working condition.

[0024] Optionally, based on the different operating states of the LPDDR memory chips, corresponding strategies are adopted to reduce the operating ambient temperature of the LPDDR memory chips, including:

[0025] When the LPDDR memory chip is in normal working condition, the software notifies the system to enable the dynamic frequency adjustment strategy of the LPDDR memory chip without resetting, thereby reducing the operating temperature of the LPDDR memory chip by lowering the operating frequency and reducing the refresh rate.

[0026] When the LPDDR memory chip is in an abnormal working state, the software notifies the system that the LPDDR memory chip is abnormal and needs to be reinitialized.

[0027] This invention provides a method for controlling high temperature of GPU memory. It uses the refresh rate of the LPDDR chip itself to infer whether the memory chip is operating in a high-temperature environment, and realizes early warning and self-adjustment of LPDDR in the face of high temperature environment, thereby improving the overall stability and maintainability of GPU board. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the GPU board structure;

[0029] Figure 2 This is a flowchart of a GPU memory high-temperature control method according to an embodiment of the present invention;

[0030] Figure 3 This is a flowchart illustrating a software-based method for controlling high-temperature GPU memory, as described in an embodiment of the present invention.

[0031] Figure 4 This is a flowchart illustrating a method for controlling high-temperature GPU memory based on a hardware-software hybrid approach, as described in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined with each other to form feasible technical solutions. Such combinations are not constrained by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0033] See Figure 1 This is a schematic diagram of the structure of a GPU board. The GPU board mainly includes the GPU core, LPDDR video memory chips and the Display module. Video memory is the data used by the GPU to render images and is also the address where the display module obtains image data. As the transfer station for data interaction between the GPU and the display, the stability of video memory is crucial. A major factor in the stability of video memory is its operating temperature.

[0034] Therefore, this invention provides a flowchart of a method for controlling high temperature of GPU memory, see below. Figure 2 The main steps in controlling high temperatures of GPU memory are as follows:

[0035] Step 1: The system allocates a portion of dynamic memory as video memory addresses.

[0036] Step 2: Periodically read the refresh rate of the LPDDR memory chips inside the GPU board using software or a combination of software and hardware, and determine whether the LPDDR memory chips are in a high-temperature working environment based on the refresh rate inside the LPDDR memory chips.

[0037] Step 3: Under the condition that the LPDDR memory chip is in a high-temperature working environment, the LPDDR memory chip is subjected to read and write tests in software mode to obtain the read and write test results.

[0038] Step 4: Determine the working status of the LPDDR memory chip based on the read / write test results;

[0039] Step 5: Based on the different working states of the LPDDR memory chips, adopt corresponding strategies to reduce the working environment temperature of the LPDDR memory chips.

[0040] Understandably, data in memory is stored in capacitors. Due to leakage current in capacitors, data will gradually be lost, so periodic refresh is needed to maintain data integrity. This invention uses the refresh rate of LPDDR memory chips to determine if the chip's temperature exceeds a threshold, thus detecting potential chip malfunctions. This invention employs two methods: 1. A purely software approach: Without interfering with GPU memory access, the kernel adds a thread that accesses the MR register (the internal register of the memory chip) in the LPDDR chip at regular intervals (e.g., 3 seconds) to obtain the refresh rate. When multiple refresh rate values ​​are at their highest, it can be predicted that the memory may be operating at high temperatures, potentially leading to access errors or data retention issues. 2. A software and hardware integrated approach: The approach in method 1 is implemented in hardware. A timer is designed on the bus. When the timer expires, the hardware automatically sends a command to obtain the chip's refresh rate. Multiple highest refresh rate values ​​trigger an interrupt. The software monitors this interrupt and notifies the GPU of a core malfunction upon receiving it. The two methods described above can proactively detect whether there is a risk of overheating in the video memory chips. Of course, even at the highest refresh rate, the video memory is not necessarily inaccessible. In this case, combining the video memory verification code (i.e., read / write video memory test) can determine whether the video memory is still working normally. Of course, to ensure stability, reducing the operating frequency of the video memory is one of the common approaches when it overheats.

[0041] In this invention, embodiments employ both purely software-based and hardware-software combined methods to detect and handle abnormal high-temperature environments in LPDDR memory chips. See also... Figure 3The flowchart below shows the process of detecting and handling abnormal high-temperature environments of LPDDR memory chips using a purely software-based approach.

[0042] (1) The system allocates a portion of the dynamic memory as video memory address. During normal operation, the GPU will have read and write access behaviors of video memory address.

[0043] (2) Without interfering with GPU operation, a separate thread can be started in kernel mode. Every certain period of time (e.g., 3 seconds), the software thread sends an MRR command (MR register read operation command of the DDR protocol standard) through the controller DMC to read the value of OP[2:0] in the MR4 register of the LPDDR memory chip. This value represents the refresh rate inside the LPDDR memory chip. The higher the temperature, the higher the refresh rate of the LPDDR memory chip.

[0044] (3) When the refresh rate of the LPDDR memory chip remains at the highest level for an extended period, it indicates that the LPDDR memory chip is currently operating in a high-temperature environment, potentially exceeding the high-temperature threshold. Therefore, the operating temperature of the LPDDR memory chip can be predicted by observing multiple refresh rate settings. For example, if the refresh rate of the LPDDR memory chip is consistently at the highest level (e.g., three times), it can be suspected that the LPDDR memory chip may be experiencing stability issues (including data read / write and data retention) due to excessively high ambient temperatures. Of course, if the refresh rate is not consistently at the highest level multiple times, the influence of excessively high temperatures can be ruled out.

[0045] (4) If the refresh rate of the LPDDR memory chip repeatedly reaches the highest level, the LPDDR memory chip may be working normally or abnormally. Therefore, read / write tests of the LPDDR memory chip are necessary. When performing read / write tests on the LPDDR memory chip, allocate another block of memory (distinct from the memory address) and use the system bus to access the LPDDR chip for read / write operations to determine its current working status.

[0046] (5) If the read and write tests of the LPDDR memory chips pass, it means that the LPDDR memory chips are in normal working condition. Although the LPDDR memory chips are in normal working condition at this time, in order to prevent the LPDDR memory chips from having possible stability problems due to prolonged operation in a high-temperature environment, such as read and write errors, data retention errors, etc., the system can be notified by software to enable the LPDDR's DVFS dynamic frequency adjustment strategy without resetting. By reducing the frequency point, the access bandwidth of the LPDDR memory chips is reduced, which indirectly affects the operating temperature of the LPDDR memory chips. Under normal circumstances, the refresh rate will also decrease accordingly.

[0047] Dynamic Voltage and Frequency Scaling (DVFS) addresses the dynamic nature of the processor's workload. By pre-classifying the processor's voltage and frequency, DVFS determines the optimal voltage and frequency level for the processor's operation within the system. During runtime, the operating system's processor power management module or driver module monitors the processor's workload changes in real time and adjusts the voltage and frequency levels according to a specific scaling strategy. This reduces power consumption while ensuring normal system functionality and considering system thermal design. This approach ensures the stability of the LPDDR memory chips without affecting the normal operation of the GPU, significantly improving the stability of the GPU board under prolonged high-temperature conditions. After the LPDDR memory chips complete the scaling process, they return to normal operation, and the software thread simultaneously monitors the refresh rate again, completing the temperature monitoring loop.

[0048] (6) If the read / write test of the LPDDR memory chip fails, it indicates that the LPDDR memory chip is in an inaccessible abnormal state. Of course, compared to situation (5), this is a low-probability event. In most cases, if the refresh rate is at its highest level, the LPDDR memory chip will still work normally for a certain period of time. If a low-probability event occurs, the high temperature of the LPDDR memory chip can also be recorded, which is convenient for subsequent system maintenance and troubleshooting. At this time, if the memory is abnormal, the software needs to notify the system that the LPDDR memory chip is abnormal and needs to be reinitialized, preferably with a relatively low refresh rate.

[0049] See Figure 4This flowchart illustrates the process of detecting and handling high-temperature anomalies in LPDDR memory chips using a hardware-software hybrid approach. The principle is essentially the same as the software-based approach. It monitors the refresh rate at its highest setting, but the hardware-software hybrid solution requires a hardware monitoring module on the bus. This module monitors the LPDDR memory chip's refresh rate and triggers an interrupt immediately after the refresh rate repeatedly meets the highest condition. Upon receiving the interrupt, the software performs subsequent LPDDR read / write tests, frequency adjustment, or reset. The specific workflow is as follows:

[0050] (1) The hardware monitoring module implements a timer. When the timer expires, the hardware monitoring module automatically sends the MRR command to obtain the value of OP[2:0] in the MR4 register on the LPDDR memory chip. This value represents the refresh rate of the LPDDR memory chip.

[0051] (2) When the hardware monitoring module determines that the refresh rate of the LPDDR memory chip is at the highest level for three consecutive times, the hardware monitoring module automatically triggers an interrupt and sends it to the software. After the software receives the interrupt, it can first disable the interrupt and focus on dealing with the possible abnormal temperature. The process is the same as (4)(5)(6) in the pure software solution. After the processing is completed, the interrupt can be enabled again.

[0052] The above process reveals the differences between a pure software solution and a combined software / hardware approach: 1. Whether the acquisition and judgment of refresh rate anomalies are handled by software or hardware; 2. Whether the refresh rate of the LPDDR memory chips is based on thread monitoring or interrupt listening. Since the refresh rates of LPDDR memory chips from different manufacturers vary, the advantage of a software solution is its ability to flexibly adapt to these differences, improving project delivery efficiency. However, thread monitoring is less efficient than interrupt listening in terms of low-level memory read / write operations. Therefore, this invention can employ either a pure software solution or a combined software / hardware approach to acquire the refresh rate of the LPDDR memory chips and determine high-temperature anomalies.

[0053] This invention provides a method for controlling high temperature of GPU memory. By obtaining the refresh rate of LPDDR memory chips, it determines whether the LPDDR memory chips are in a high-temperature working environment. It proactively provides high temperature warnings for memory chips and can quickly determine whether the memory chips are abnormal under high temperature conditions and take corresponding measures. This can greatly improve the GPU's adaptability to temperature. At the same time, even if high temperature triggers memory abnormalities, it can be recorded, which greatly facilitates the later maintenance of software and the localization of high temperature stability problems in the field.

[0054] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0055] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0058] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0060] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for controlling high temperature of GPU memory, characterized in that, include: The system allocates a portion of dynamic memory for use as video memory addresses; The refresh rate of the LPDDR memory chips inside the GPU board is periodically read using software or a combination of software and hardware, and the LPDDR memory chips are determined to be in a high-temperature operating environment based on the refresh rate of the LPDDR memory chips. Under the condition that the LPDDR memory chip is in a high-temperature working environment, read and write tests are performed on the LPDDR memory chip using software to obtain read and write test results. Based on the read / write test results, the working status of the LPDDR memory chip is determined; Based on the different operating states of the LPDDR memory chips, corresponding strategies are adopted to reduce the operating ambient temperature of the LPDDR memory chips.

2. The GPU memory high-temperature control method according to claim 1, characterized in that, The system periodically reads the refresh rate of the LPDDR memory chips inside the GPU board using software, and determines whether the LPDDR memory chips are operating in a high-temperature environment based on the refresh rate of the LPDDR memory chips, including: A separate thread is created in kernel mode. The thread sends an MRR command through the controller DMC at preset intervals to read the value of the MR4 register in the LPDDR memory chip. This value represents the refresh rate inside the LPDDR memory chip. Based on software-based determination, when the refresh rate inside the LPDDR memory chip reaches a preset refresh rate threshold multiple times consecutively, the LPDDR memory chip is in a high-temperature operating environment; otherwise, the LPDDR memory chip is in a non-high-temperature operating environment.

3. The GPU memory high-temperature control method according to claim 1, characterized in that, The system periodically reads the refresh rate of the LPDDR memory chips inside the GPU board using a combination of hardware and software methods, and determines whether the LPDDR memory chips are operating in a high-temperature environment based on the refresh rate of the LPDDR memory chips, including: A hardware monitoring module is implemented on the system bus. A timer is implemented in the hardware monitoring module. When the timer expires, the hardware monitoring module automatically sends an MRR command to read the value of the MR4 register in the LPDDR memory chip. This value represents the refresh rate inside the LPDDR memory chip. The hardware monitoring module determines that when the refresh rate inside the LPDDR memory chip reaches the preset refresh rate threshold multiple times consecutively, the LPDDR memory chip is in a high-temperature operating environment; otherwise, the LPDDR memory chip is in a non-high-temperature operating environment. When the hardware monitoring module determines that the LPDDR memory chip is in a high-temperature operating state, it triggers an interrupt to the software. After the software receives the interrupt, it disables the interrupt.

4. The GPU memory high-temperature control method according to claim 1, characterized in that, Under high-temperature operating conditions, read / write tests were performed on the LPDDR memory chips using software to obtain the read / write test results, including: When the LPDDR memory chip is in a high-temperature operating environment, a memory block with a different address from the memory chip is allocated. The read and write accesses of the LPDDR memory chip through the system bus are obtained based on the software, and the read and write test results of the LPDDR memory chip are obtained based on the read and write accesses.

5. The GPU memory high-temperature control method according to claim 4, characterized in that, The read / write test results of the LPDDR memory chip obtained based on the read / write access include: If the read / write test result of the LPDDR memory chip is passed, it is determined that the LPDDR memory chip is in normal working condition; if the read / write test result of the LPDDR memory chip is failed, it is determined that the LPDDR memory chip is in abnormal working condition.

6. The GPU memory high-temperature control method according to claim 5, characterized in that, Based on the different operating states of the LPDDR memory chips, corresponding strategies are adopted to reduce the operating ambient temperature of the LPDDR memory chips, including: When the LPDDR memory chip is in normal working condition, the software notifies the system to enable the dynamic frequency adjustment strategy of the LPDDR memory chip without resetting, thereby reducing the operating temperature of the LPDDR memory chip by lowering the operating frequency and reducing the refresh rate. When the LPDDR memory chip is in an abnormal working state, the software notifies the system that the LPDDR memory chip is abnormal and needs to be reinitialized.

Images