Processor heat dissipation method, computer equipment, storage medium and program product

Abstract

The present application discloses a heat dissipation method, computer equipment, storage medium and program product for a processor, which relates to the field of processor heat dissipation technology, including: when the target processor power consumption mode is high power consumption mode, obtaining a target temperature difference and a target processor utilization rate, mapping the target temperature difference and the target processor utilization rate to corresponding fuzzy sets using a Gaussian function, determining and determining a target value of a heat dissipation parameter combination based on the target temperature difference membership and the target processor utilization membership, and then adjusting the speed of a fan corresponding to the target processor; when the target processor power consumption mode is low power consumption mode, obtaining a first value of the target processor's current temperature setting point and a second value of the target processor's desired temperature setting point, and adjusting the wind direction of the fan corresponding to the target processor based on the first value and the second value. This solves the technical problem that a single heat dissipation strategy is difficult to meet the heat dissipation requirements of different processors, and achieves the technical effect of improving the heat dissipation reliability of the processor.

Description

Technical Field

[0001] The present application relates to the field of processor heat dissipation technology, and in particular to a processor heat dissipation method, computer equipment, storage medium, and program product. Background Art

[0002] Graphics processing units (GPUs) generate significant heat during high-speed read and write operations. This heat dissipation issue is particularly acute in densely deployed server environments with multiple GPUs operating simultaneously. Excessive GPU temperatures can not only affect GPU performance and lifespan but can also lead to data loss or hardware failure. Therefore, effective heat dissipation control is crucial to ensuring stable GPU operation.

[0003] Traditional cooling systems usually adopt a unified cooling strategy, which makes it difficult to fine-tune the cooling requirements of different GPUs, and easily leads to insufficient or excessive cooling of some GPUs. Summary of the Invention

[0004] The present application provides a processor heat dissipation method, computer equipment, storage medium and program product to solve the technical problem in related technologies that a single heat dissipation strategy is difficult to meet the heat dissipation requirements of different processors.

[0005] The present application provides a method for dissipating heat from a processor, comprising:

[0006] Get the target processor power consumption mode;

[0007] In response to the target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, where the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine the target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine the target processor utilization membership; a target value of a heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; and a speed of a fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination;

[0008] In response to the target processor power consumption mode being the low power consumption mode, a first value of the target processor current temperature setting point and a second value of the target processor desired temperature setting point are obtained, and the wind direction of the fan corresponding to the target processor is adjusted based on the first value and the second value.

[0009] The present application also provides a computer device, comprising: a memory for storing a computer program; and a processor for implementing the steps of the data processing method in the following embodiments when executing the computer program.

[0010] Get the target processor power consumption mode;

[0011] In response to the target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, where the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine the target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine the target processor utilization membership; a target value of a heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; and a speed of a fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination;

[0012] In response to the target processor power consumption mode being the low power consumption mode, a first value of the target processor current temperature setting point and a second value of the target processor desired temperature setting point are obtained, and the wind direction of the fan corresponding to the target processor is adjusted based on the first value and the second value.

[0013] The present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of the data processing method in the following embodiments are implemented.

[0014] Get the target processor power consumption mode;

[0015] In response to the target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, where the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine the target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine the target processor utilization membership; a target value of a heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; and a speed of a fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination;

[0016] In response to the target processor power consumption mode being the low power consumption mode, a first value of the target processor current temperature setting point and a second value of the target processor desired temperature setting point are obtained, and the wind direction of the fan corresponding to the target processor is adjusted based on the first value and the second value.

[0017] The present application also provides a computer program product, including a computer program, which implements the steps of the data processing method in the following embodiments when the computer program is executed by a processor.

[0018] Get the target processor power consumption mode;

[0019] In response to the target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, where the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine the target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine the target processor utilization membership; a target value of a heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; and a speed of a fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination;

[0020] In response to the target processor power consumption mode being the low power consumption mode, a first value of the target processor current temperature setting point and a second value of the target processor desired temperature setting point are obtained, and the wind direction of the fan corresponding to the target processor is adjusted based on the first value and the second value.

[0021] The heat dissipation method for the processor provided by this application distinguishes different target processor power consumption modes and sets different processor heat dissipation strategies.

[0022] In response to the target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, wherein the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine the target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine the target processor utilization membership; the target value of the heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; the speed of the fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination; in response to the target processor power consumption mode being a low power consumption mode, a first value of the current temperature setting point of the target processor and a second value of the expected temperature setting point of the target processor are obtained, and the wind direction of the fan corresponding to the target processor is adjusted based on the first value and the second value. Therefore, the speed of the fan corresponding to the processor or the wind direction of the fan corresponding to the processor can be selectively adjusted according to the heat dissipation requirements of the processor in different power consumption modes, so that refined control of the heat dissipation of the processor can be achieved while taking into account energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0024] Figure 1A schematic flow chart of a heat dissipation method for a processor provided in one embodiment of the present application;

[0025] Figure 2 A schematic diagram of the structure of a heat dissipation system for a processor provided in one embodiment of the present application;

[0026] Figure 3 A schematic structural diagram of a heat dissipation system for a processor provided in another embodiment of the present application;

[0027] Figure 4 A schematic structural diagram of a heat dissipation device for a processor provided in one embodiment of the present application;

[0028] Figure 5 This is a diagram of the internal structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0030] It should be noted that, in the description of this application, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. The terms "first," "second," etc., in this application are used to distinguish similar objects, and are not used to describe a particular order or sequence.

[0031] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0032] like Figure 1 As shown, an embodiment of the present application provides a method for cooling a processor, the method specifically comprising the following steps:

[0033] Step 101: Obtain the target processor power consumption mode.

[0034] Step 102: In response to the target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, where the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine the target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine the target processor utilization membership; a target value of a heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; and the speed of a fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination.

[0035] The target processor of the present application may specifically be a graphics processing unit (GPU), which is a hardware chip specially designed to efficiently process graphics rendering, parallel computing, and machine learning tasks.

[0036] The target processor's power consumption status can be queried by directly reading the processor's MSR registers, obtaining it through the baseboard management controller's IPMI commands, or reading it through the Windows WMI interface. The target processor's power consumption status obtained by different methods is classified as high power consumption mode or low power consumption mode based on a unified standard.

[0037] In one embodiment, in high power consumption mode, a target temperature difference and a target processor utilization can be obtained. The target temperature difference is the difference between the actual target processor temperature and the desired target processor temperature. The target processor utilization can be calculated by calculating the percentage of time the GPU spends executing non-idle tasks. For example, if 800 milliseconds of the past second were spent executing computations, the utilization is 80%.

[0038] In this application, multiple temperature difference fuzzy sets can be pre-defined, each corresponding to a temperature difference range. The temperature difference ranges of different temperature difference fuzzy sets in this application can intersect. In other words, a target temperature difference can belong to two temperature difference fuzzy sets simultaneously. After obtaining the target temperature difference, a Gaussian function can be used to map the target temperature difference to the target temperature difference fuzzy set, thereby determining the target temperature difference's membership.

[0039] Specifically, multiple temperature difference fuzzy sets and temperature difference ranges corresponding to the multiple temperature difference fuzzy sets are obtained; the temperature difference fuzzy set corresponding to the temperature difference range matching the target temperature difference is selected as the target temperature difference fuzzy set; the target temperature difference range corresponding to the target temperature difference fuzzy set is obtained, and the center point value of the target temperature difference range and the target temperature difference standard deviation of the multiple temperature differences in the target temperature difference range are determined; the target temperature difference membership is calculated based on the center point value of the target temperature difference range, the target temperature difference standard deviation and the target temperature difference membership calculation formula.

[0040] The center point value of the target temperature range is the average value of all temperature differences within the target temperature range. The square of the difference between every two temperature differences within the target temperature range is calculated, and the average of these squared differences is used to obtain the variance. The square root of the variance is taken to obtain the target temperature standard deviation. The target temperature standard deviation measures the degree of dispersion of the temperature differences around the mean. The center point value of the target temperature range and the target temperature standard deviation are input into the target temperature membership calculation formula to calculate the target temperature membership.

[0041] Exemplarily, the multiple temperature difference fuzzy sets of the present application may include a first temperature difference fuzzy set, a second temperature difference fuzzy set, a third temperature difference fuzzy set, a fourth temperature difference fuzzy set, and a fifth temperature difference fuzzy set, and the target temperature difference fuzzy set is at least one of the multiple temperature difference fuzzy sets.

[0042] The target temperature difference membership calculation formula set in this application can be as follows:

[0043] ;

[0044] Wherein, μ represents the target temperature difference membership, X1 represents the target temperature difference belonging to the first temperature difference fuzzy set, Y1 represents the center point value of the first temperature difference range corresponding to the first temperature difference fuzzy set, and Z1 represents the first temperature difference standard deviation of the multiple temperature differences in the first temperature difference range; X2 represents the target temperature difference belonging to the second temperature difference fuzzy set, Y2 represents the center point value of the second temperature difference range corresponding to the second temperature difference fuzzy set, and Z2 represents the second temperature difference standard deviation of the multiple temperature differences in the second temperature difference range; X3 represents the target temperature difference belonging to the third temperature difference fuzzy set, Y3 represents the center point value of the third temperature difference range corresponding to the third temperature difference fuzzy set, and Z3 represents the third temperature difference standard deviation of the multiple temperature differences in the third temperature difference range; X4 represents the target temperature difference belonging to the fourth temperature difference fuzzy set, Y4 represents the center point value of the fourth temperature difference range corresponding to the fourth temperature difference fuzzy set, and Z4 represents the fourth temperature difference standard deviation of the multiple temperature differences in the fourth temperature difference range; X5 represents the target temperature difference belonging to the fifth temperature difference fuzzy set, Y5 represents the center point value of the fifth temperature difference range corresponding to the fifth temperature difference fuzzy set, and Z5 represents the fifth temperature difference standard deviation of multiple temperature differences in the fifth temperature difference range.

[0045] Furthermore, the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function, and determining the target processor utilization membership includes: obtaining multiple processor utilization fuzzy sets and utilization ranges corresponding to the multiple processor utilization fuzzy sets; selecting a processor utilization fuzzy set corresponding to the utilization range matching the target processor utilization as the target processor utilization fuzzy set; obtaining a target utilization range corresponding to the target processor utilization fuzzy set, determining the center point value of the target utilization range, and the target utilization standard deviation of multiple utilizations in the target utilization range; and calculating the target processor utilization membership based on the center point value of the target utilization range, the target utilization standard deviation, and the target utilization membership calculation formula.

[0046] This application also pre-sets multiple processor utilization fuzzy sets, each processor utilization fuzzy set corresponds to a processor utilization range, and the temperature difference ranges of different processor utilization sets in this application may also have intersections. The center point value of the target utilization range refers to the average value of all utilizations belonging to the target utilization range. The square of the difference between every two utilizations in the target utilization range is calculated, and the average of these square differences is obtained to obtain the variance. The square root of the variance is taken to obtain the target utilization standard deviation. The target utilization standard deviation can measure the degree of dispersion of the utilization value around the mean. The center point value of the target utilization range and the target utilization standard deviation are input into the target utilization membership calculation formula to calculate the target processor utilization membership.

[0047] Exemplarily, the multiple processor utilization fuzzy sets of the present application may include a first processor utilization fuzzy set, a second processor utilization fuzzy set, and a third processor utilization fuzzy set. The target processor utilization fuzzy set may be at least one of the multiple processor utilization fuzzy sets. The target utilization membership calculation formula set in the present application may be as follows:

[0048] ;

[0049] in, βrepresents the membership of the target processor utilization, A1 represents the target processor utilization belonging to the first processor utilization fuzzy set, B1 represents the center point value of the first utilization range corresponding to the first processor utilization fuzzy set, and C1 represents the first utilization standard deviation of the multiple utilizations in the first utilization range; A2 represents the target processor utilization belonging to the second processor utilization fuzzy set, B2 represents the center point value of the second utilization range corresponding to the second processor utilization fuzzy set, and C2 represents the second utilization standard deviation of the multiple utilizations in the second target processor utilization range; A3 represents the target processor utilization belonging to the third processor utilization fuzzy set, B3 represents the center point value of the third utilization range corresponding to the third processor utilization fuzzy set, and C3 represents the third utilization standard deviation of the multiple utilizations in the third utilization range.

[0050] In this application, the target temperature difference and target processor utilization are mapped to the target temperature difference fuzzy set and the target processor utilization fuzzy set respectively through Gaussian function, which can analyze the potential properties of the target temperature difference and the target processor utilization, and provide support for the subsequent precise adjustment of the fan speed corresponding to the target controller.

[0051] After obtaining the target temperature difference membership and the target processor utilization membership through the above steps, the target value of the heat dissipation parameter combination can be determined based on the target temperature difference membership and the target processor utilization membership, specifically including: obtaining a predefined fuzzy rule table, the predefined fuzzy rule table defines the initial heat dissipation parameter combinations corresponding to different temperature difference fuzzy sets and different processor utilization sets; selecting a target initial heat dissipation parameter combination that matches the target temperature difference fuzzy set and the target processor utilization fuzzy set from the predefined fuzzy rule table; using the target temperature difference membership and the target processor utilization membership as the weights of multiple heat dissipation parameters in the target heat dissipation parameter combination, and calculating the target values ​​of multiple heat dissipation parameters in the heat dissipation parameter combination.

[0052] Cooling parameters include the setpoint (temperature setpoint), alarm thresholds (used to determine whether an overtemperature alarm is required), and PID parameters. PID parameters specifically include ki (integral parameter), kp (proportional parameter), and kd (derivative parameter). PID parameters can be used to adjust fan speed. The proportional parameter kp determines the amount of control action proportional to the current temperature error. If the current temperature is significantly above the setpoint, the fan speed will be significantly increased to reduce the temperature, depending on the kp value. The integral parameter ki accounts for accumulated temperature errors and adjusts for long-term temperature deviations. For example, if the temperature remains slightly above the setpoint, ki will gradually increase the control action, further adjusting the fan speed. The derivative parameter kd is sensitive to the rate of temperature change. If the temperature changes rapidly, kd will adjust the fan speed accordingly to quickly respond to the temperature change and prevent excessive temperature rise or fall.

[0053] The predefined fuzzy rule table can be in the form of a PID parameter reference value matrix. For example, different columns in the first row of the rule table represent different processor utilization fuzzy sets. Different rows in the first column of the rule table represent different temperature difference fuzzy sets, and the intersections of other columns and rows in the table represent initial cooling parameter combinations consisting of integral parameters and their initial values, proportional parameters and their initial values, and differential parameters and their initial values, under the temperature difference fuzzy set corresponding to that row and the processor utilization fuzzy set corresponding to that column. Therefore, the predefined fuzzy rule table can be used to find initial cooling parameter combinations that take into account both processor utilization and temperature difference factors.

[0054] Because the target temperature difference and target processor utilization of the target processor in the present application can respectively belong to different temperature difference fuzzy sets and different processor utilization fuzzy sets at the same time, in a specific embodiment, the target temperature difference fuzzy set can include a first target temperature difference fuzzy set and a second target temperature difference fuzzy set; the target processor utilization fuzzy set can include a first target processor utilization fuzzy set and a second target processor utilization fuzzy set; in this case, using the target temperature difference membership and the target processor utilization membership as the weights of the multiple heat dissipation parameters in the target heat dissipation parameter combination, calculating the target values ​​of the multiple heat dissipation parameters in the heat dissipation parameter combination can include:

[0055] S1: Obtain a first target temperature difference membership corresponding to a target temperature difference belonging to a first target temperature difference fuzzy set and a second target temperature difference membership corresponding to a target temperature difference belonging to a second target temperature difference fuzzy set.

[0056] S2: Obtain a first target processor utilization membership corresponding to a target processor utilization belonging to the first target processor utilization fuzzy set and a second target processor utilization membership corresponding to a target processor utilization belonging to the second target processor utilization fuzzy set.

[0057] S3: comparing the first target temperature difference membership and the second target temperature difference membership, in response to the first target temperature difference membership being greater than the second target temperature difference membership, using the first target temperature difference membership as the main temperature difference weight and the second target temperature difference membership as the auxiliary temperature difference weight.

[0058] Here, by using the Gaussian function to simultaneously capture the characteristics of the target temperature difference in the first target temperature difference fuzzy set and the characteristics of the target temperature difference in the second target temperature difference fuzzy set, the characteristics of the target temperature difference in different temperature difference fuzzy sets can be multimodally perceived, avoiding the control blind spot caused by a single membership, and improving the temperature detection sensitivity. The target temperature difference membership with a large membership value, that is, a large proportion, is used as the main temperature difference weight, and conversely, the target temperature difference membership with a small membership value, that is, a small proportion, is used as the auxiliary temperature difference weight. This can give priority to responding to the control needs of the target temperature difference with a large membership value, and promptly respond to sudden large-scale cooling needs, while retaining the auxiliary temperature difference weight to prevent overshoot.

[0059] S4: According to the influence of the first target processor utilization on the processor heat dissipation and the influence of the second target processor utilization on the processor heat dissipation, respectively set a first target processor utilization weight and a second target processor utilization weight, wherein the target processor utilization weight is proportional to the influence of the target processor utilization on the processor heat dissipation.

[0060] Here, by using a Gaussian function to capture the characteristics of the target processor utilization within the first target processor utilization fuzzy set and the heat generation characteristics of the target processor utilization within the first target processor utilization fuzzy set, we can avoid the control blind spots caused by a single membership degree and improve the sensitivity of processor utilization detection. A higher weight is assigned to target processor utilization that has a greater impact on processor heat dissipation, that is, a higher load.

[0061] S5: Add the product of the first target processor utilization weight and the first target processor utilization membership and the product of the second target processor utilization weight and the second target processor utilization membership to obtain a processor utilization rule weight.

[0062] By using a weighted approach, a smooth transition of the load state is achieved, the impact of instantaneous load fluctuations is reduced, and the fan speed fluctuation can be reduced.

[0063] S6: Compare the main temperature difference weight and the processor utilization rule weight, and in response to the main temperature difference weight being greater than the processor utilization rule weight, use the processor utilization rule weight as the main rule weight.

[0064] S7: Compare the auxiliary temperature difference weight and the processor utilization rule weight, and in response to the processor utilization rule weight being greater than the auxiliary temperature difference weight, use the auxiliary temperature difference weight as the auxiliary rule weight.

[0065] Here, by comparing the main temperature difference weight, the auxiliary temperature difference weight and the processor utilization rule weight respectively, and selecting the smaller rule weight in the comparison process as the main rule weight and the auxiliary rule weight, it can be ensured that the control intensity does not exceed the credibility of any input condition.

[0066] S8: Selecting any type of heat dissipation parameter as a target heat dissipation parameter in sequence, and obtaining an initial heat dissipation parameter value corresponding to the target heat dissipation parameter.

[0067] S9: Add the product of the main rule weight and the initial heat dissipation parameter value corresponding to the target heat dissipation parameter and the product of the auxiliary rule weight and the initial heat dissipation parameter value corresponding to the target heat dissipation parameter to obtain a target value corresponding to the target heat dissipation parameter.

[0068] Finally, each heat dissipation parameter is weighted and synthesized by the main rule weight and the auxiliary rule weight to obtain the target value corresponding to the heat dissipation parameter. The target values ​​corresponding to each type of heat dissipation parameter are combined to finally obtain the target value corresponding to the target heat dissipation parameter.

[0069] For example, assuming the target initial cooling parameter combination is L = [kp = 3.0, ki = 0.1, kd = 1.0], the calculated primary rule weight is 0.6, and the calculated secondary rule weight is 0.5, then the target value for the kp type cooling parameter is = initial kp value (3.0) * primary rule weight 0.6 + initial kp value (3.0) * secondary rule weight (0.5) = 3.3. Similarly, the final target value for the cooling parameter combination is L (final) = [kp = 3.3, ki = 0.11, kd = 1.1].

[0070] In this application, the fan speed is adjusted primarily by using heat dissipation parameters such as proportional parameters, integral parameters, and differential parameters. In one embodiment, adjusting the speed of the fan corresponding to the target processor based on the target value of the heat dissipation parameter combination includes: obtaining a proportional target value of the proportional parameter, an integral target value of the integral parameter, and a differential target value of the differential parameter; inputting the proportional target value, the integral target value, and the differential target value into a target value calculation formula for the heat dissipation parameter combination to obtain the target value of the heat dissipation parameter combination; converting the target value of the heat dissipation parameter combination into a target pulse width modulation signal duty cycle; obtaining the maximum rated speed of the fan corresponding to the target processor, and calculating a target fan speed based on the maximum rated speed and the target pulse width modulation signal duty cycle; and adjusting the speed of the fan corresponding to the target processor based on the target fan speed.

[0071] The target value calculation formula for the heat dissipation parameter combination includes:

[0072] ;

[0073] The constraints are:

[0074] ;

[0075] Where u(t) is the target value of the heat dissipation parameter combination, λ is the target processor performance gain coefficient, and k u is the target processor utilization gain coefficient, U(t) is the target processor utilization, kp is the proportional target value, ki is the integral target value, kd is the differential target value, e(t) is the target temperature difference, is the error change rate, P fan (u(t)) is the power of the fan corresponding to the target processor, P max is the maximum power of the fan, e max is the maximum temperature difference.

[0076] The target processor performance gain coefficient can be set according to the performance mode of the processor. The processor performance mode can include high-performance mode, normal mode, and power-saving mode. The performance mode of the processor can be determined according to the power consumption of the processor and the load size of the processor. The performance mode gain coefficient is proportional to the level of the performance mode. For example, in the high-performance mode, the target processor performance gain coefficient is higher, and in the power-saving mode, the target processor performance gain coefficient is lower. In this way, different performance modes are distinguished to set the gain coefficient as the target value of the heat dissipation parameter combination, which can quickly respond to heat dissipation in the high-performance mode. The specific value of the target processor performance gain coefficient can be set according to actual experience. The target processor utilization gain coefficient is determined according to the size of the target processor utilization. Generally, the greater the target processor utilization, the greater the target processor utilization gain coefficient. The specific value of the target processor utilization gain coefficient can be set according to actual experience.

[0077] In this application, the target temperature difference is constrained by introducing the maximum temperature difference, and the power of the fan corresponding to the target processor is constrained by introducing the maximum fan power. At the same time, the target processor performance gain coefficient and the target processor utilization gain coefficient are used to achieve real-time adjustment of the target value of the cooling parameter combination according to the processor utilization and processor performance scenario, thereby controlling the fan cooling intensity, and limiting the fan power and the maximum temperature difference to avoid fan overheating or overconsumption.

[0078] The proportional target value, integral target value, and differential target value are input into the target value calculation formula for the cooling parameter combination, outputting the target value for the cooling parameter combination, also known as the PID control value. The PID control value is converted into a target pulse width modulation signal duty cycle. The resulting target value for the cooling parameter combination is a percentage. This value is converted into a PWM value (pulse width modulation signal duty cycle) using an 8-bit register (0-255). The resulting PWM value is a number between 0 and 255. The relationship between fan speed and PWM is typically nonlinear. The target fan speed can be obtained by looking up the corresponding relationship between fan speed and PWM, or by calculating the product of the maximum rated speed of the fan corresponding to the target processor and the target pulse width modulation signal duty cycle.

[0079] In actual applications, servers may have mixed GPU configurations, or they may have multiple GPUs (e.g., GPUs with multiple GPUs) or be configured with multiple GPUs independently. Because different GPU types vary in power consumption and heat dissipation characteristics, heat dissipation management in mixed environments is more complex. Therefore, in this application, the heat dissipation parameters of multiple target processors can be combined to select the heat dissipation parameters of the target processor with the highest heat dissipation requirements as the heat dissipation standard for the entire system. This ensures that even under the most stringent heat dissipation requirements, all processors remain within an appropriate operating temperature range, thereby ensuring system stability and reliability. For example, assume that in a mixed GPU environment, there are two GPUs, A and B. GPU A has a maximum temperature of 60°C and a warning temperature of 70°C under high load, while GPU B has a maximum temperature of 55°C and a warning temperature of 65°C. In this case, the system will use the heat dissipation requirements of GPU B as a benchmark to develop a corresponding heat dissipation strategy, ensuring that the GPU temperature does not exceed 65°C under any circumstances, thereby damaging the GPU.

[0080] In one embodiment, when the target value of the heat dissipation parameter combination is used as the heat dissipation requirement consideration standard, multiple target pulse width modulation signal duty cycles corresponding to multiple target processors can be obtained, and the target pulse width modulation signal duty cycle with the largest value is selected from the multiple target pulse width modulation signal duty cycles as the reference target pulse width modulation signal duty cycle; the target fan speed is calculated based on the reference target pulse width modulation signal duty cycle and the maximum rated speed, and the speeds of the multiple fans corresponding to the multiple target processors are adjusted based on the target fan speed.

[0081] This application uses cooling parameter speed regulation to determine the maximum cooling requirement of the GPU, specifically the setpoint, ki, kp, kd and other parameters, to control the fan's PID speed. The application continuously monitors the GPU temperature and dynamically adjusts the fan speed based on the target temperature difference corresponding to the target processor, combined with cooling parameters such as ki, kp, and kd, to ensure that the GPU operates within the appropriate temperature range. At the same time, the application also monitors the alarm threshold and issues an alarm message immediately if the temperature exceeds the alarm threshold.

[0082] In one embodiment, the heat dissipation method further includes:

[0083] S10: Acquire key parameters of the target processor and the initial physical address of the target processor, and determine whether the target processor has been replaced based on the key parameters and the initial physical address of the target processor.

[0084] Specifically, the processor inserted into the current initial physical address is used as a verification processor; key parameters of the verification processor and key parameters of the target processor are obtained, and the key parameters include processor manufacturer ID, processor ID, processor sub-manufacturer ID, and sub-processor ID; it is determined whether the key parameters of the verification processor are consistent with the key parameters of the target processor; in response to the key parameters of the verification processor being consistent with the key parameters of the target processor, it is determined that the target processor has not been replaced; in response to the key parameters of the verification processor being inconsistent with the key parameters of the target processor, it is determined that the target processor has been replaced.

[0085] In actual use, GPU replacement is inevitable. This application utilizes the baseboard management controller (BMC) to exchange information with the GPU, reading key parameters such as the GPU's vendor ID (processor manufacturer ID), device ID (processor ID), sub-vendor ID (processor sub-vendor ID), and sub-device ID (sub-processor ID) in real time. By analyzing these parameters, it is possible to accurately determine the currently connected GPU model and whether the GPU has been replaced.

[0086] S20: In response to the target processor having been replaced, determining the type of the replacement processor inserted at the current initial physical address.

[0087] Specifically, a replacement processor access link is determined based on the current initial physical address; a correspondence between a preset processor protocol address and a processor type is obtained; any processor protocol address is selected in turn from the correspondence between the preset processor protocol address and the processor type as the replacement processor protocol address; the replacement processor is accessed through the replacement processor access link and the replacement processor protocol address to obtain a replacement processor access status; in response to the replacement processor access status being an access success, the processor type that matches the replacement processor protocol address is used as the type of the replacement processor.

[0088] If it is determined that the target processor originally inserted at the initial physical address has been replaced, it is necessary to first determine the type of processor that was replaced. Given the current initial physical address, the access link to the replacement processor inserted at the current physical address can be determined based on the initial physical address.

[0089] See also Figure 2In this application, the BMC's I2C interface is connected to the I2C switch (expander) on the motherboard. After the BMC starts, the I2C interface is initialized and configured first, and appropriate parameters such as communication rate and data format are set to ensure that it can communicate normally with the I2C switch on the motherboard. The reason for choosing the I2C switch is that it can realize the expansion function of the I2C bus, allowing the BMC I2C to connect to more devices. This connection method builds the basic physical link for data transmission of the entire system, ensuring that the BMC can communicate with subsequently mounted devices. Different channels of the I2C switch on the motherboard (multiple channels are shown in the figure) are connected to each GPU respectively. Each GPU backplane acts as an independent branch and obtains I2C communication capabilities from the motherboard I2C switch. This design enables the system to flexibly manage multiple GPUs and can easily add new GPU connections when the number of GPUs needs to be expanded.

[0090] A Baseboard Management Controller (BMC) is a management controller designed specifically for servers and other hardware devices. As part of the Intelligent Platform Management Interface (IPMI), it allows system administrators to monitor and control the server's hardware status through an independent management network. Typically integrated on the server's motherboard, the BMC has its own processor, memory, and storage, running independent management software. In a server environment, the BMC plays a critical role in effectively monitoring the operating status of the GPU.

[0091] I2C (Inter-Integrated Circuit) is a bidirectional two-wire serial bus standard used to connect microcontrollers to other integrated circuits (ICs) on the same circuit board.

[0092] In this application, the BMC sequentially accesses the I2C (Inter-Integrated Circuit) bus system based on pre-defined hardware link logic. During this process, particular attention is paid to the riser card under the I2C link, as it integrates an I2C switch that manages multiple channels. The BMC sequentially iterates through each channel, aiming to establish communication with the GPU device connected to the I2C link. During each I2C link access, the BMC needs to determine the type of GPU connected. To do this, the BMC utilizes pre-configured addresses and access protocols for various GPU types. These protocols are specifically tailored for different GPU types (e.g., Class A, Class B, Class C, etc.). The BMC attempts to access each possible address and protocol combination. Specifically, starting with the first address, it sequentially sends request signals to the device using each GPU protocol. If, on a particular I2C link, communication with the device using the Class A GPU protocol is successful and a response signal is received from the device, the BMC can determine that the GPU connected to that link is a Class A GPU.

[0093] In one feasible implementation, the GPU protocol addresses of two manufacturers may conflict, and an error may occur in determining whether the GPU is in place. When the protocol addresses conflict, key parameters such as the GPU's vender ID (processor manufacturer ID), device ID (processor ID), sub vender ID (processor sub-manufacturer ID), and sub device ID (sub-processor ID) can be obtained based on the protocol address to distinguish GPUs from different manufacturers. In this way, the type of GPU can be identified more accurately.

[0094] In this application, the GPU adaptation conf file can be loaded into the memory so that the relevant information of GPUs from different manufacturers can be quickly found later. The GPU-related information here can include the temperature storage location and calculation method. The temperature storage location here refers to the different register addresses. Each GPU has a different location for reading temperature or power, so it is necessary to record the temperature storage location corresponding to the GPU to quickly find the relevant information of the GPU. According to the instructions in the adaptation file, a command to obtain temperature information is sent to the GPU, the raw temperature data is read, and it is converted into the actual temperature value according to the calculation method. For the basic information of the GPU, such as key parameter information such as venderID, the physical location information is written to the assert.json file (a transfer file) for subsequent retrieval.

[0095] The BMC of this application supports web import and export of conf files through the redfish interface. By exporting this file, the administrator can easily modify the cooling parameters in the external environment. These parameters cover key information such as fan speed adjustment strategy and temperature threshold setting to adapt to different hardware configurations and cooling requirements. The modified parameters can be easily imported into the BMC through the web interface. Once the import is successful, the BMC will adjust the system's cooling strategy power according to the new cooling parameters. This method greatly improves the flexibility and convenience of cooling management. Administrators can quickly respond to hardware changes or optimize cooling effects without entering the complex underlying system settings.

[0096] S30: Determine a replacement processor expected temperature of the replacement processor based on the replacement processor type.

[0097] S40: Adjusting the rotation speed of the fan corresponding to the replaced processor based on the expected temperature of the replaced processor.

[0098] Once the BMC determines the GPU type under the access link, for example, a Class A GPU, it registers the device type information of the Class A GPU with a dedicated GPU monitoring thread. This monitoring thread continuously monitors the operating status of the GPUs throughout the server system. Subsequently, the monitoring thread periodically sends information requests to the GPUs based on the protocol corresponding to the Class A GPUs. For example, this monitoring thread obtains GPU temperature information through this protocol to determine whether the GPU is operating within the normal temperature range. It also obtains key information such as GPU utilization and power consumption, enabling comprehensive, real-time monitoring of the GPU's operating status and ensuring stable GPU operation in the server system. The BMC's cooling .conf file configuration file configures energy-saving policy parameters, such as the fan speed reduction ratio when the GPU is not present. System administrators can adjust these parameters based on actual needs. For example, in environments with high heat dissipation requirements, the fan speed reduction ratio can be appropriately reduced.

[0099] In one specific example, the BMC continuously monitors the system hardware status, including the presence of the GPU. In motherboard design, the GPU slot has specific pins that relay presence information to the BMC. When the GPU is properly inserted into the slot, these pins form a specific electrical signal connection, which the BMC recognizes and confirms the GPU's presence. However, when the GPU is not in place, the pin connection status changes, and the BMC detects the signal change, thus determining that the GPU is not in place. Once the BMC detects that the GPU is not in place, it executes the corresponding action based on a pre-set energy-saving strategy. Because the GPU is a high-heat component in the system, when it is not in place, the cooling demand at that location is significantly reduced. Therefore, the BMC sends a command to the fan speed control circuit. The fan control circuit uses a PWM (pulse width modulation) signal to control the fan speed, reducing the fan speed at that location, thereby reducing the fan motor's power consumption and achieving energy savings.

[0100] A GPU presence detection program is written into the BMC firmware. This program periodically reads the status of the GPU slot presence detection pin. The pin level status is read every three seconds. If the pin level is detected three times in a row, indicating that the GPU is not present, the GPU is confirmed to be absent. When the BMC detects that the GPU is absent, the program calculates the PWM signal parameters corresponding to the new fan speed based on the preset energy-saving strategy. The program specifies that the fan speed is reduced by 50% when the GPU is absent. The program then calculates the PWM duty cycle corresponding to this 50% speed and sends this PWM signal to the fan control circuit.

[0101] In one embodiment, if the processor inserted at the initial physical address at the current sampling moment is a replacement processor and the processor inserted at the initial physical address at the previous sampling moment is a target processor to be replaced, determining an expected temperature of the replacement processor based on the type of the replacement processor; and adjusting the speed of a fan corresponding to the replacement processor based on the expected temperature of the replacement processor includes:

[0102] S100: Obtaining from the system file the proportional target value, the integral target value, the differential target value, and the target pulse width modulation signal duty cycle corresponding to the target processor replaced at the last sampling moment.

[0103] S110: Collect the temperature of the processor being replaced at the current sampling moment and the temperature of the target processor being replaced at the previous sampling moment, perform a subtraction operation with the temperature of the processor being replaced at the current sampling moment as the minuend and the temperature of the target processor being replaced at the previous sampling moment as the subtrahend, and obtain a first temperature change.

[0104] S120: Multiply the first temperature change by the proportional target value corresponding to the target processor replaced at the last sampling moment to obtain a proportional change.

[0105] Exemplarily, kp_part (proportional change) = proportional target value corresponding to the target processor replaced at the previous sampling moment * (temperature of the replaced processor at the current sampling moment - temperature of the target processor replaced at the previous sampling moment). The proportional adjustment part is calculated by the difference between the temperature of the replaced processor at the current sampling moment and the temperature of the target processor replaced at the previous sampling moment, combined with the proportional target value corresponding to the target processor replaced at the previous sampling moment.

[0106] S130: Obtain the expected temperature of the replacement processor, perform a subtraction operation with the temperature of the replacement processor at the current sampling moment as the minuend and the expected temperature of the replacement processor as the subtrahend, and obtain a second temperature change.

[0107] S140: Multiply the second temperature change by the integral target value corresponding to the target processor replaced at the last sampling moment to obtain an integral change.

[0108] ki_part (integral change) = integral target value corresponding to the target processor replaced at the previous sampling time * (temperature of the replaced processor at the current sampling time - expected temperature of the replaced processor). The integral adjustment part is calculated using the deviation between the temperature of the replaced processor at the current sampling time and the expected temperature of the replaced processor, combined with the integral target value corresponding to the target processor replaced at the previous sampling time, to eliminate static error.

[0109] S150: Collect the temperature of the target processor that was replaced at the previous sampling moment, where the previous sampling moment is the previous sampling moment of the previous sampling moment. Use the sum of the temperature of the processor replaced at the current sampling moment and the temperature of the target processor replaced at the previous sampling moment as the minuend, and perform a subtraction operation with double the temperature of the target processor replaced at the previous sampling moment as the subtrahend to obtain a third temperature change.

[0110] S160: Multiply the third temperature change by the differential target value corresponding to the target processor replaced at the previous sampling moment to obtain a differential change.

[0111] Exemplarily, kd_part (differential change) = proportional target value corresponding to the replaced target processor * temperature of the replaced processor at the current sampling moment - 2 * temperature of the replaced target processor at the previous sampling moment + temperature of the replaced target processor at the previous sampling moment). The differential adjustment part is calculated based on the temperature changes at the current, previous, and previous sampling moments, combined with the proportional target value corresponding to the replaced target processor, to reflect the rate of temperature change.

[0112] S170: Calculate the target pulse width modulation signal duty cycle, proportional change, integral change, and differential change sum corresponding to the target processor replaced at the previous sampling moment to obtain the target pulse width modulation signal duty cycle corresponding to the replaced processor.

[0113] S180: Adjusting the rotation speed of the fan corresponding to the replaced processor based on the target pulse width modulation signal duty cycle corresponding to the replaced processor.

[0114] Finally, the current PWM output for control is calculated by combining the target pulse width modulation signal duty cycle, proportional change, integral change, and differential change corresponding to the target processor that was replaced at the previous sampling moment. This allows for precise dynamic adjustment of the fan speed corresponding to the processor.

[0115] In one embodiment, in response to obtaining a heat dissipation update instruction or the system being in a startup phase, the current system ambient temperature is obtained; from a preset correspondence between the system ambient temperature and the pulse width modulation signal duty cycle, a pulse width modulation signal duty cycle that matches the current system ambient temperature is searched; and based on the pulse width modulation signal duty cycle that matches the current system ambient temperature, the speed of the fan corresponding to the target processor is adjusted.

[0116] This application addresses the need for additional safeguards in certain situations, such as when the cooling process restarts or when the BMC is not fully booted. To this end, an RTOS (real-time operating system) is introduced to assist with cooling control. An RTOS offers rapid response and precise control capabilities. It reads the inlet temperature (system ambient temperature) in real time. Based on this temperature data, the RTOS can directly control fan speed, ensuring that the system maintains basic cooling capabilities even when the BMC is not functioning properly. For example, during the initial system startup, the BMC requires time to initialize and load various drivers. During this time, the RTOS monitors the inlet temperature and adjusts the fan speed to prevent GPU overheating due to delayed cooling. Similarly, if the cooling process restarts for some reason, the RTOS can continue to provide stable cooling support during the transition period until the BMC reassumes cooling control.

[0117] When the GPU is plugged in or replaced, the file storing assert.json will change. The BMC can determine whether the GPU has been replaced by comparing the differences between the files before and after the GPU is plugged in or replaced. If the BMC detects that the GPU has been replaced, it will recalculate the required cooling parameters based on the new situation and restart the cooling process. The significance of this operation is that it allows customers to ensure that the cooling is always in the optimal state without affecting the business. Within two seconds of the cooling process being restarted, the real-time operating system (RTOS) will take over the cooling work. The RTOS will adjust the fan speed according to the server inlet temperature to ensure that the cooling can proceed normally. In this way, the cooling can be ensured to be always in the optimal state without affecting the business.

[0118] Step 103: In response to the target processor power consumption mode being the low power consumption mode, obtain a first value of the target processor's current temperature setting point and a second value of the target processor's desired temperature setting point, and adjust the wind direction of the fan corresponding to the target processor based on the first value and the second value.

[0119] Specifically, adjusting the wind direction of the fan corresponding to the target processor based on the first value and the second value includes: performing a subtraction operation with the second value as the minuend and the first value as the subtrahend to obtain the temperature difference of the temperature setting point; calculating the target ratio of the temperature difference of the temperature setting point to the second value; multiplying the target ratio by the maximum wind direction adjustment angle to obtain the target adjustment direction of the fan corresponding to the target processor; and adjusting the angle of the fan plate of the fan corresponding to the target processor based on the target adjustment direction.

[0120] The current temperature set point of the target processor here refers to the current setpoint of the target processor, and the expected temperature set point of the target processor here refers to the setpoint that the target processor needs to reach. Assume that the first value of the current temperature set point is 70 degrees, and the second value of the expected temperature set point of the target processor is 80 degrees. At this time, the target adjustment direction angle of the fan corresponding to the target processor should be (80-70) divided by 80, and then multiplied by 90 degrees (multiplied by the maximum wind direction adjustment angle). The target adjustment direction angle here refers to adjusting the angle of the fan plate corresponding to each fan, thereby adjusting the wind direction of the fan.

[0121] In one embodiment, adjusting the wind direction of a fan corresponding to a target processor includes: determining the physical locations of multiple processors to be adjusted in the system whose wind directions are to be adjusted; obtaining heat dissipation requirements of the multiple processors to be adjusted, where the heat dissipation requirement is the difference between the current temperature of the processors to be adjusted and the expected temperature of the processors to be adjusted; grouping the multiple processors to be adjusted based on the heat dissipation requirements of the multiple processors to be adjusted and the physical locations of the multiple processors to be adjusted to obtain at least one heat concentration area group; and adjusting the wind directions of the fans corresponding to the processors to be adjusted in the heat concentration area group and the wind directions of the fans corresponding to the processors to be adjusted adjacent to the heat concentration area group to achieve heat dissipation of the processors to be adjusted in the heat concentration area group.

[0122] Multiple processors to be adjusted refer to processors that are about to undergo wind direction adjustment. The physical locations of the multiple processors to be adjusted are obtained, and the processors to be adjusted that have high heat dissipation requirements and are physically close to each other are added to the heat concentration area group. Wind direction adjustment is performed preferentially on high-temperature areas or areas with high heat dissipation requirements. In addition to adjusting the wind direction of the fans corresponding to the processors to be adjusted in the heat concentration area group, the wind direction of the fans corresponding to the processors to be adjusted adjacent to the heat concentration area group can also be directed toward the heat concentration area group. In this way, wind direction adjustment resources can be concentrated to prioritize processing areas with high requirements, and local processor heat dissipation can be optimized while considering energy consumption.

[0123] In one embodiment, the fan plates corresponding to the fans of the processors to be regulated in the heat-concentrated area group can be configured to expand to form independent air ducts, forcing cool air to flow through the heat sink fins. Meanwhile, the fan plates corresponding to the fans of the processors to be regulated in the non-heat-concentrated area group can remain flush with the chassis cover to avoid excessive turbulence and pressure loss. This allows for localized cooling of processors with high heat dissipation requirements and close physical proximity when different types of GPUs are mixed. Separate fan plates for GPUs can also be used to isolate the CPU and GPU air ducts. Furthermore, each fan plate is smaller and compatible with 1U / 2U server chassis.

[0124] In one embodiment, the heat dissipation method of the processor provided in the present application is applied to a heat dissipation system of the processor, such as Figure 3As shown, the system includes a server chassis, multiple processors, a baseboard management controller (BMC), multiple fans, and multiple fan boards. Each processor corresponds to a fan, and each fan has a fan board that fits onto the server chassis cover. All fan boards can be connected to the BMC via a control bus (such as I2C / PWM). The BMC dynamically adjusts the angle of the heat dissipation baffle on each fan board, thereby adjusting the airflow direction of the fan corresponding to the processor. The fan board can adopt a hinged blade design. In the default state, it fits onto the chassis cover (reducing wind resistance). In one embodiment, when the fan board's direction is triggered, a micro-stepping motor can be used to expand it to a specific angle (adjustable from 30° to 90°), forming a physical air deflector.

[0125] The embodiment of the present application provides a heat dissipation device for a processor, and the heat dissipation device for the processor is specifically as follows: Figure 4 As shown, the heat dissipation device of the processor includes: an acquisition module 20 and an adjustment module 21.

[0126] The acquisition module 20 is configured to acquire the power consumption mode of the target processor.

[0127] The adjustment module 21 is used to obtain a target temperature difference and a target processor utilization in response to the target processor power consumption mode being a high power consumption mode, where the target temperature difference is the difference between the actual temperature of the target processor and the expected temperature of the target processor; use a Gaussian function to map the target temperature difference to a target temperature difference fuzzy set to determine the target temperature difference membership; use a Gaussian function to map the target processor utilization to a target processor utilization fuzzy set to determine the target processor utilization membership; determine a target value of a heat dissipation parameter combination based on the target temperature difference membership and the target processor utilization membership; adjust the speed of a fan corresponding to the target processor based on the target value of the heat dissipation parameter combination; in response to the target processor power consumption mode being a low power consumption mode, obtain a first value of a current temperature setting point of the target processor and a second value of an expected temperature setting point of the target processor, and adjust the wind direction of the fan corresponding to the target processor based on the first value and the second value.

[0128] like Figure 5 As shown, an embodiment of the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any of the above-mentioned processor heat dissipation method embodiments.

[0129] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any of the above-mentioned embodiments of the heat dissipation method for a processor when running.

[0130] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0131] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be judged to be beyond the scope of this application.

[0132] The above is a detailed introduction to the heat dissipation method for a processor provided by this application. This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only intended to help understand the method and core ideas of this application. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, various improvements and modifications can be made to this application, and these improvements and modifications also fall within the scope of protection of this application.

Claims

1. A heat dissipation method for a processor, characterized in that: The heat dissipation method comprises: Get the target processor power consumption mode; In response to a target processor power consumption mode being a high power consumption mode, a target temperature difference and a target processor utilization are obtained, wherein the target temperature difference is a difference between an actual temperature of the target processor and an expected temperature of the target processor; the target temperature difference is mapped to a target temperature difference fuzzy set using a Gaussian function to determine a target temperature difference membership; the target processor utilization is mapped to a target processor utilization fuzzy set using a Gaussian function to determine a target processor utilization membership; a target value of a heat dissipation parameter combination is determined based on the target temperature difference membership and the target processor utilization membership; and a speed of a fan corresponding to the target processor is adjusted based on the target value of the heat dissipation parameter combination; In response to the target processor power consumption mode being the low power consumption mode, a first value of the target processor current temperature setting point and a second value of the target processor desired temperature setting point are obtained, and the wind direction of the fan corresponding to the target processor is adjusted based on the first value and the second value.

2. The heat dissipation method for a processor according to claim 1, wherein: Mapping the target temperature difference to the target temperature difference fuzzy set using a Gaussian function and determining the target temperature difference membership comprises: Obtain multiple temperature difference fuzzy sets and temperature difference ranges corresponding to the multiple temperature difference fuzzy sets; Select the temperature difference fuzzy set corresponding to the temperature difference range matching the target temperature difference as the target temperature difference fuzzy set; Obtaining a target temperature difference range corresponding to the target temperature difference fuzzy set, determining a center point value of the target temperature difference range, and target temperature difference standard deviations of multiple temperature differences in the target temperature difference range; The target temperature difference membership is calculated based on the center point value of the target temperature difference range, the target temperature difference standard deviation, and a target temperature difference membership calculation formula.

3. The heat dissipation method for a processor according to claim 2, wherein: The target temperature difference fuzzy set includes at least one of the first temperature difference fuzzy set, the second temperature difference fuzzy set, the third temperature difference fuzzy set, the fourth temperature difference fuzzy set, and the fifth temperature difference fuzzy set. The target temperature difference membership calculation formula include: ； Wherein, μ represents the target temperature difference membership, X1 represents the target temperature difference belonging to the first temperature difference fuzzy set, Y1 represents the center point value of the first temperature difference range corresponding to the first temperature difference fuzzy set, and Z1 represents the first temperature difference standard deviation of multiple temperature differences in the first temperature difference range; X2 represents the target temperature difference belonging to the second temperature difference fuzzy set, Y2 represents the center point value of the second temperature difference range corresponding to the second temperature difference fuzzy set, and Z2 represents the second temperature difference standard deviation of multiple temperature differences in the second temperature difference range; X3 represents the target temperature difference belonging to the third temperature difference fuzzy set, Y3 represents the center point value of the third temperature difference range corresponding to the third temperature difference fuzzy set, and Z3 represents the third temperature difference standard deviation of the multiple temperature differences in the third temperature difference range; X4 represents the target temperature difference belonging to the fourth temperature difference fuzzy set, Y4 represents the center point value of the fourth temperature difference range corresponding to the fourth temperature difference fuzzy set, and Z4 represents the fourth temperature difference standard deviation of the multiple temperature differences in the fourth temperature difference range; X5 represents the target temperature difference belonging to the fifth temperature difference fuzzy set, Y5 represents the center point value of the fifth temperature difference range corresponding to the fifth temperature difference fuzzy set, and Z5 represents the fifth temperature difference standard deviation of multiple temperature differences in the fifth temperature difference range.

4. The heat dissipation method for a processor according to claim 1, wherein: Mapping the target processor utilization to the target processor utilization fuzzy set by using a Gaussian function and determining the target processor utilization membership comprises: Obtaining multiple processor utilization fuzzy sets and utilization ranges corresponding to the multiple processor utilization fuzzy sets; Selecting a processor utilization fuzzy set corresponding to a utilization range matching the target processor utilization as the target processor utilization fuzzy set; Obtaining a target utilization range corresponding to the target processor utilization fuzzy set, determining a center point value of the target utilization range, and target utilization standard deviations of multiple utilizations in the target utilization range; The target processor utilization membership is calculated based on the center point value of the target utilization range, the target utilization standard deviation, and a target utilization membership calculation formula.

5. The heat dissipation method for a processor according to claim 4, wherein: The target processor utilization fuzzy set is at least one of the first processor utilization fuzzy set, the second processor utilization fuzzy set, and the third processor utilization fuzzy set. The target utilization membership calculation formula is: include: ； in, β represents the target processor utilization membership, A1 represents the target processor utilization belonging to the first processor utilization fuzzy set, B1 represents the center point value of the first utilization range corresponding to the first processor utilization fuzzy set, and C1 represents the first utilization standard deviation of multiple utilizations in the first utilization range; A2 represents the target processor utilization that belongs to the second processor utilization fuzzy set, B2 represents the center point value of the second utilization range corresponding to the second processor utilization fuzzy set, and C2 represents the second utilization standard deviation of multiple utilizations in the second target processor utilization range; A3 represents the target processor utilization belonging to the third processor utilization fuzzy set, B3 represents the center point value of the third utilization range corresponding to the third processor utilization fuzzy set, and C3 represents the third utilization standard deviation of multiple utilizations in the third utilization range.

6. The heat dissipation method for a processor according to claim 1, wherein: Determining the target value of the heat dissipation parameter combination based on the target temperature difference membership and the target processor utilization membership includes: Obtaining a predefined fuzzy rule table, wherein the predefined fuzzy rule table defines initial heat dissipation parameter combinations corresponding to different temperature difference fuzzy sets and different processor utilization sets; Selecting a target initial cooling parameter combination that matches the target temperature difference fuzzy set and the target processor utilization fuzzy set from a predefined fuzzy rule table; The target temperature difference membership and the target processor utilization membership are used as weights of multiple heat dissipation parameters in the target heat dissipation parameter combination, and target values ​​of the multiple heat dissipation parameters in the heat dissipation parameter combination are calculated.

7. The heat dissipation method for a processor according to claim 6, wherein: The target temperature difference fuzzy set includes a first target temperature difference fuzzy set and a second target temperature difference fuzzy set; the target processor utilization fuzzy set includes a first target processor utilization fuzzy set and a second target processor utilization fuzzy set; The step of calculating target values ​​of the plurality of heat dissipation parameters in the heat dissipation parameter combination by using the target temperature difference membership and the target processor utilization membership as weights of the plurality of heat dissipation parameters in the target heat dissipation parameter combination includes: Obtaining a first target temperature difference membership corresponding to a target temperature difference belonging to the first target temperature difference fuzzy set and a second target temperature difference membership corresponding to a target temperature difference belonging to the second target temperature difference fuzzy set; Obtaining a first target processor utilization membership corresponding to a target processor utilization belonging to a first target processor utilization fuzzy set and a second target processor utilization membership corresponding to a target processor utilization belonging to a second target processor utilization fuzzy set; comparing a first target temperature difference membership and a second target temperature difference membership, and in response to the first target temperature difference membership being greater than the second target temperature difference membership, using the first target temperature difference membership as a primary temperature difference weight and using the second target temperature difference membership as a secondary temperature difference weight; According to the influence of the first target processor utilization on the processor heat dissipation and the influence of the second target processor utilization on the processor heat dissipation, respectively setting a first target processor utilization weight and a second target processor utilization weight, wherein the target processor utilization weight is proportional to the influence of the target processor utilization on the processor heat dissipation; Adding the product of the first target processor utilization weight and the first target processor utilization membership and the product of the second target processor utilization weight and the second target processor utilization membership to obtain a processor utilization rule weight; comparing the main temperature difference weight and the processor utilization rule weight, and in response to the main temperature difference weight being greater than the processor utilization rule weight, using the processor utilization rule weight as the main rule weight; comparing the auxiliary temperature difference weight and the processor utilization rule weight, and in response to the processor utilization rule weight being greater than the auxiliary temperature difference weight, using the auxiliary temperature difference weight as the auxiliary rule weight; Selecting any type of heat dissipation parameter in turn as a target heat dissipation parameter, and obtaining an initial heat dissipation parameter value corresponding to the target heat dissipation parameter; The product of the main rule weight and the initial heat dissipation parameter value corresponding to the target heat dissipation parameter and the product of the auxiliary rule weight and the initial heat dissipation parameter value corresponding to the target heat dissipation parameter are added together to obtain a target value corresponding to the target heat dissipation parameter.

8. The heat dissipation method for a processor according to claim 1, wherein: The heat dissipation parameter combination includes a plurality of heat dissipation parameters, wherein the plurality of heat dissipation parameters include a proportional parameter, an integral parameter, and a differential parameter. Adjusting the rotation speed of the fan corresponding to the target processor based on the target value of the heat dissipation parameter combination includes: Obtaining the proportional target value of the proportional parameter, the integral target value of the integral parameter, and the differential target value of the differential parameter; Inputting the proportional target value, the integral target value, and the differential target value into a target value calculation formula of a heat dissipation parameter combination to obtain a target value of the heat dissipation parameter combination; Converting the target value of the heat dissipation parameter combination into a target pulse width modulation signal duty cycle; Obtaining a maximum rated speed of a fan corresponding to a target processor, and calculating a target fan speed based on the maximum rated speed and a target pulse width modulation signal duty cycle; The rotation speed of the fan corresponding to the target processor is adjusted based on the target fan rotation speed.

9. The heat dissipation method for a processor according to claim 8, wherein: The target value calculation formula of the heat dissipation parameter combination includes: ； The constraints are: ； Where u(t) is the target value of the heat dissipation parameter combination, λ is the target processor performance gain coefficient, and k u is the target processor utilization gain coefficient, U(t) is the target processor utilization, kp is the proportional target value, ki is the integral target value, kd is the differential target value, e(t) is the target temperature difference, is the error change rate, P fan (u(t)) is the power of the fan corresponding to the target processor, P max is the maximum power of the fan, e max is the maximum temperature difference.

10. The heat dissipation method for a processor according to claim 8, wherein: There are multiple target processors, and obtaining the maximum rated speed of the fan corresponding to the target processor, and calculating the target fan speed according to the maximum rated speed and the target pulse width modulation signal duty cycle further includes: Acquire multiple target pulse width modulation signal duty cycles corresponding to multiple target processors, and select a target pulse width modulation signal duty cycle with the largest value from the multiple target pulse width modulation signal duty cycles as a reference target pulse width modulation signal duty cycle; A target fan speed is calculated based on the reference target pulse width modulation signal duty cycle and the maximum rated speed, and the speeds of multiple fans corresponding to multiple target processors are adjusted based on the target fan speed.

11. The heat dissipation method for a processor according to claim 1, wherein: The heat dissipation method further includes: Acquiring key parameters of a target processor and an initial physical address of the target processor, and determining whether the target processor has been replaced based on the key parameters of the target processor and the initial physical address; In response to the target processor having undergone a replacement operation, determining a type of replacement processor inserted at the current initial physical address; determining a replacement processor expected temperature for the replacement processor based on the replacement processor type; The rotation speed of the fan corresponding to the replaced processor is adjusted based on the expected temperature of the replaced processor.

12. The heat dissipation method for a processor according to claim 11, wherein: The acquiring key parameters of the target processor and the initial physical address of the target processor, and determining whether the target processor has been replaced based on the key parameters of the target processor and the initial physical address includes: The processor at the current initial physical address is inserted as a check processor; Obtain key parameters of the verification processor and the target processor, including processor manufacturer ID, processor ID, processor sub-manufacturer ID, and sub-processor ID; Determining whether the verification processor key parameters are consistent with the target processor key parameters; In response to the verification processor key parameter being consistent with the target processor key parameter, determining that the target processor has not been replaced; In response to the verification processor key parameter being inconsistent with the target processor key parameter, it is determined that the target processor has been replaced.

13. The heat dissipation method for a processor according to claim 11, wherein: Determining the replacement processor type inserted at the current initial physical address includes: determining a replacement processor access link based on the current initial physical address; Obtain the correspondence between the preset processor protocol address and the processor type; Selecting any processor protocol address from the preset correspondence between the processor protocol address and the processor type in sequence as the replacement processor protocol address; Accessing the replacement processor via the replacement processor access link and the replacement processor protocol address, Get the replacement processor access status; In response to the replacement processor access status being access success, a processor type matching the replacement processor protocol address is used as the type of the replacement processor.

14. The heat dissipation method for a processor according to claim 11, wherein: If the processor initially physically inserted at the current sampling moment is a replacement processor and the processor initially physically inserted at the previous sampling moment is a target processor to be replaced, determining the replacement processor expected temperature of the replacement processor based on the type of the replacement processor; Adjusting the rotation speed of the fan corresponding to the replacement processor based on the expected temperature of the replacement processor includes: Obtain from the system file the proportional target value, integral target value, differential target value, and target pulse width modulation signal duty cycle corresponding to the target processor replaced at the previous sampling moment; collecting the temperature of the processor being replaced at the current sampling moment and the temperature of the target processor being replaced at the previous sampling moment, performing a subtraction operation using the temperature of the processor being replaced at the current sampling moment as a minuend and the temperature of the target processor being replaced at the previous sampling moment as a subtrahend to obtain a first temperature change; Multiplying the first temperature change by the proportional target value corresponding to the target processor replaced at the previous sampling moment to obtain a proportional change; Obtaining an expected temperature of the replacement processor, performing a subtraction operation with the temperature of the replacement processor at the current sampling moment as the minuend and the expected temperature of the replacement processor as the subtrahend to obtain a second temperature change; Multiplying the second temperature change by the integral target value corresponding to the target processor replaced at the previous sampling moment to obtain an integral change; collecting the temperature of the target processor that was replaced at a previous sampling moment, where the previous sampling moment is the sampling moment before the previous sampling moment, performing a subtraction operation using the sum of the temperature of the replaced processor at the current sampling moment and the temperature of the target processor that was replaced at the previous sampling moment as a minuend, and using double the temperature of the target processor that was replaced at the previous sampling moment as a subtrahend to obtain a third temperature change; Multiplying the third temperature change by the differential target value corresponding to the target processor replaced at the previous sampling moment to obtain a differential change; Calculating the sum of the target pulse width modulation signal duty cycle, proportional change, integral change, and differential change corresponding to the target processor replaced at the previous sampling moment to obtain the target pulse width modulation signal duty cycle corresponding to the replaced processor; The rotation speed of the fan corresponding to the replaced processor is adjusted based on the target pulse width modulation signal duty cycle corresponding to the replaced processor.

15. The heat dissipation method for a processor according to claim 1, wherein: The heat dissipation method further includes: In response to obtaining a heat dissipation update instruction or the system being in a startup phase, obtaining a current system ambient temperature; From the preset correspondence between the system ambient temperature and the pulse width modulation signal duty cycle, searching for the pulse width modulation signal duty cycle that matches the current system ambient temperature; The rotation speed of the fan corresponding to the target processor is adjusted based on the duty cycle of the pulse width modulation signal that matches the current system ambient temperature.

16. The heat dissipation method for a processor according to claim 1, wherein: The adjusting the wind direction of the fan corresponding to the target processor based on the first value and the second value includes: performing a subtraction operation using the second value as the minuend and the first value as the subtrahend to obtain a temperature difference between the temperature set points; Calculating a target ratio of a temperature difference between a temperature set point and a second value; Multiplying the target ratio by the maximum wind direction adjustment angle to obtain a target adjustment direction of the fan corresponding to the target processor; The angle of the fan plate of the fan corresponding to the target processor is adjusted based on the target adjustment direction.

17. The heat dissipation method for a processor according to claim 1, wherein: The adjusting the wind direction of the fan corresponding to the target processor includes: Determining the physical locations of multiple processors to be adjusted in the system for adjusting wind directions; Obtaining heat dissipation requirements of multiple processors to be adjusted, where the heat dissipation requirement is a difference between a current temperature of the processor to be adjusted and an expected temperature of the processor to be adjusted; Grouping the plurality of processors to be adjusted based on the heat dissipation requirements of the plurality of processors to be adjusted and the physical locations of the plurality of processors to be adjusted to obtain at least one heat concentration area group; The wind directions of the fans corresponding to the processors to be adjusted in the heat concentration area group and the wind directions of the fans corresponding to the processors to be adjusted adjacent to the heat concentration area group are adjusted to achieve heat dissipation for the processors to be adjusted in the heat concentration area group.

18. A computer device, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the heat dissipation method for a processor as claimed in any one of claims 1 to 17 when executing the computer program.

19. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the heat dissipation method for a processor according to any one of claims 1 to 17 are implemented.

20. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the heat dissipation method for a processor according to any one of claims 1 to 17 are implemented.

Images