Heat dissipation method of server and electronic device

Abstract

The application discloses a heat dissipation method of a server and an electronic device, relates to the technical field of server heat dissipation, and comprises the following steps: obtaining a running state parameter of the server; determining a thermal risk index of the server according to the running state parameter, wherein the thermal risk index is used for indicating the probability that the server reaches a thermal peak; determining a heat pipe heat dissipation strategy corresponding to the thermal risk index, wherein the heat pipe heat dissipation power corresponding to the heat pipe heat dissipation strategy is positively correlated with the thermal risk index; and executing the heat pipe heat dissipation strategy, so that the heat stored in a heat dissipation phase change material of the server is transferred through the heat pipe, and the heat dissipation phase change material is used for absorbing the heat of the server through solid-liquid conversion.

Description

Technical Field

[0001] This application relates to the field of server heat dissipation technology, and in particular to server heat dissipation methods and electronic devices. Background Technology

[0002] With the rapid development of cloud computing, high-performance computing, and high-density computing, the power consumption and heat generation of server chips are escalating rapidly. Servers often experience sudden high loads, generating intense heat spikes within seconds, with instantaneous power reaching hundreds of watts. This causes chips to rapidly heat up to safety thresholds, necessitating efficient and fast-responding cooling systems for server heat dissipation. Currently, server cooling primarily relies on air cooling or liquid cooling at fixed thresholds, and these systems typically only activate when temperature peaks occur, resulting in ineffective cooling and impacting normal server operation. Summary of the Invention

[0003] This application provides a server heat dissipation method and electronic device to at least solve the problem that related technologies mainly rely on air cooling or liquid cooling to dissipate server heat at fixed thresholds, and are basically only activated when temperature peaks occur, resulting in poor heat dissipation and affecting the normal operation of the server.

[0004] This application provides a server heat dissipation method, including: Obtain the server's operating status parameters; Based on the operating status parameters, the thermal risk index of the server is determined, and the thermal risk index is used to indicate the probability that the server will reach a heat peak. Determine the heat pipe cooling strategy corresponding to the thermal risk index, wherein the heat pipe cooling power corresponding to the heat pipe cooling strategy is positively correlated with the thermal risk index; The heat pipe cooling strategy is implemented to transfer the heat stored in the heat-dissipating phase change material of the server through the heat pipe, the heat-dissipating phase change material being used to absorb the heat of the server through solid-liquid conversion.

[0005] This application also provides a server heat dissipation device, including: The acquisition module is used to acquire the operating status parameters of the server; The processing module is used to determine the thermal risk index of the server based on the operating status parameters, wherein the thermal risk index is used to indicate the probability that the server will reach a heat peak; The processing module is further configured to determine a heat pipe cooling strategy corresponding to the thermal risk index, wherein the heat pipe cooling power corresponding to the heat pipe cooling strategy is positively correlated with the thermal risk index. The processing module is further configured to execute the heat pipe cooling strategy, so that the heat stored in the heat dissipation phase change material of the server is transferred through the heat pipe, and the heat dissipation phase change material is used to absorb the heat of the server through solid-liquid conversion. This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the heat dissipation method of any of the above-described servers.

[0006] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above-described server heat dissipation methods.

[0007] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described server heat dissipation methods.

[0008] This application obtains the server's operating status parameters; based on these parameters, it determines the server's thermal risk index, which indicates the probability of the server reaching a heat peak; it then determines a heat pipe cooling strategy corresponding to the thermal risk index, where the heat pipe cooling power is positively correlated with the thermal risk index; and finally, it executes the heat pipe cooling strategy to transfer the heat stored in the server's heat-dissipating phase change material (PCM) through heat pipes. The PCM absorbs heat from the server through solid-liquid conversion. In this scheme, server heat dissipation is divided into passive heat dissipation via the PCM and active heat dissipation via the heat pipe cooling strategy. The corresponding heat pipe cooling strategy can be determined based on the thermal risk index calculated from the server's status parameters to ensure sufficient cooling power to transfer the heat absorbed by the PCM in a timely manner, achieving effective heat dissipation. Attached Figure Description

[0009] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart of a server heat dissipation method provided in this application embodiment. Figure 1 ; Figure 2 A structural diagram of the integrated thermal buffer device provided in the embodiments of this application; Figure 3 A flowchart of a server heat dissipation method provided in this application embodiment. Figure 2 ; Figure 4 A structural diagram of a server heat dissipation device provided in an embodiment of this application; Figure 5 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0012] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0013] With the rapid development of cloud computing, AI training, high-performance computing, and high-density computing, the power consumption and heat generation of server chips are escalating rapidly. In AI training clusters, GPUs or processors often generate severe heat spikes within seconds due to sudden high loads, with instantaneous power reaching hundreds of watts. If the heat is not dissipated in time, the chip junction temperature may exceed the safe threshold, triggering frequency reduction protection, task interruption, and in severe cases, permanent component failure, threatening the stability of the data center and data security.

[0014] Efficient and fast-response cooling systems have become a critical infrastructure for ensuring reliable server operation. Currently, mainstream server cooling solutions fall into two categories: the first is air cooling systems, which have a simple structure and rely on forced convection from fans, combined with heat sinks for heat exchange. However, due to the low thermal conductivity of air, they often use fixed temperature thresholds to trigger fan speed adjustments (e.g., increasing fan speed when the temperature exceeds a certain threshold). This often results in strong cooling only being activated after the chip has already experienced thermal overshoot, leading to severely insufficient cooling capacity in high heat flux density scenarios. Air cooling systems lack the ability to sense load dynamics and temperature rise rates, resulting in delayed response and an inability to effectively suppress heat peaks.

[0015] The second type is the traditional liquid cooling system (cold plate type or immersion type), which uses a fixed flow pump or a simple on / off valve to control the liquid cooling circulation. It lacks the ability to predict dynamic changes in load and temperature rise trends, and cannot dynamically adjust the cooling intensity according to the real-time heat load. In addition, even when the server is idle, the liquid cooling pump still runs at a constant high speed, causing unnecessary energy waste and noise pollution. When there is a sudden high load, it is difficult to avoid instantaneous temperature rise due to the lack of a buffer mechanism.

[0016] Currently, traditional heat dissipation systems generally neglect the issue of "thermal inertia." Chip heat generation has instantaneous characteristics (second-level response), while heat dissipation systems have physical response delays. Without an intermediate thermal buffer, relying solely on external cooling equipment will inevitably result in control blind spots.

[0017] In summary, to address all or part of the aforementioned technical problems, this application provides a server heat dissipation method and electronic device. The method involves acquiring server operating status parameters; determining a thermal risk index based on these parameters, where the thermal risk index indicates the probability of the server reaching a heat peak; determining a heat pipe cooling strategy corresponding to the thermal risk index, where the heat pipe cooling power is positively correlated with the thermal risk index; and executing the heat pipe cooling strategy to transfer heat stored in the server's heat-dissipating phase change material via heat pipes. The heat-dissipating phase change material absorbs heat from the server through solid-liquid conversion. In this solution, server heat dissipation is divided into passive heat dissipation via the heat-dissipating phase change material and active heat dissipation via the heat pipe cooling strategy. Furthermore, the corresponding heat pipe cooling strategy can be determined based on the thermal risk index calculated from the server's status parameters to ensure sufficient cooling power to transfer the heat absorbed by the phase change material in a timely manner, achieving effective heat dissipation.

[0018] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, Figure 1 A flowchart of a server heat dissipation method provided for embodiments of this application is shown, the method including the following steps: 101. Obtain the server's running status parameters.

[0020] In this embodiment of the application, the server may generate heat due to load during operation, so it is necessary to pay attention to the server's operation. Therefore, the server's operating status parameters can be obtained, which may include temperature, load, etc., and can reflect the current operating status of the server.

[0021] 102. Determine the server's thermal risk index based on the operating status parameters.

[0022] In this embodiment of the application, after determining the operating status parameters, the thermal risk index of the server can be calculated based on the operating status parameters. The thermal risk index can be used to indicate the probability that the server will reach a heat peak.

[0023] It's important to note that the heat generated by a server fluctuates during operation. Under heavy load, the server may generate more heat; conversely, under light load, it will generate less. Additionally, the ambient temperature can also affect the server's heat output. These factors can lead to momentary heat spikes, where temperatures reach high values, such as exceeding a certain threshold. This can impact normal server operation. Therefore, by analyzing real-time operational parameters, a server's thermal risk index can be determined to predict the probability of future heat spikes and allow for timely cooling measures to prevent them.

[0024] 103. Determine the heat pipe cooling strategy corresponding to the thermal risk index.

[0025] In this embodiment of the application, after determining the thermal risk index, a heat pipe cooling strategy corresponding to the thermal risk index can be determined. Since the thermal risk index indicates the probability that the server will reach the heat peak, the higher the thermal risk index, the more likely the server is to reach the heat peak, and therefore a higher cooling power is required for cooling. The lower the thermal risk index, the lower the probability that the server will reach the heat peak, and therefore a higher cooling power is not required for cooling. Therefore, it can be seen that the heat pipe cooling power corresponding to the heat pipe cooling strategy is positively correlated with the thermal risk index.

[0026] 104. Implement a heat pipe cooling strategy to transfer the heat stored in the heat dissipation phase change material of the server through heat pipes.

[0027] In this embodiment of the application, after the heat pipe cooling strategy is determined, the heat pipe cooling strategy can be executed to cool the server in a timely manner.

[0028] It should be noted that heat dissipation phase change materials can be installed in servers. These materials can automatically absorb the heat generated in the server through solid-liquid conversion. However, these materials only store the heat and cannot actively transfer it. Therefore, heat pipes are needed to transfer the heat stored in the heat dissipation phase change materials. This heat pipe cooling strategy can achieve different heat pipe cooling power by adjusting the power of the heat pipe for heat transfer.

[0029] In this embodiment, the server's operating status parameters are obtained; based on these parameters, a thermal risk index is determined, indicating the probability of the server reaching a heat peak; a heat pipe cooling strategy corresponding to the thermal risk index is determined, with the heat pipe cooling power positively correlated with the thermal risk index; the heat pipe cooling strategy is then executed to transfer heat stored in the server's heat-dissipating phase change material via heat pipes. This heat-dissipating phase change material absorbs heat from the server through solid-liquid conversion. In this scheme, server heat dissipation is divided into passive heat dissipation via the heat-dissipating phase change material and active heat dissipation via the heat pipe cooling strategy. The corresponding heat pipe cooling strategy can be determined based on the thermal risk index calculated from the server's status parameters to ensure sufficient cooling power to transfer the heat absorbed by the phase change material in a timely manner, achieving effective heat dissipation.

[0030] In some embodiments, the heat-dissipating phase change material can be a high-performance composite phase change material (PCM), such as paraffin-based composite graphite or metal foam. This heat-dissipating PCM is integrated inside the heat sink. When the ambient temperature reaches its phase change point, the PCM undergoes a solid-liquid phase change, absorbing a large amount of heat. When the server encounters a sudden high load (such as AI cluster training) leading to a heat load phenomenon, the PCM absorbs the excess heat, alleviating the rapid rise in server temperature and effectively suppressing the occurrence of heat peaks. The latent heat of phase change of the phase change material can be selected between 200-300 kJ / kg. This material is only used for thermal energy storage and needs to be combined with heat pipe technology for heat transfer.

[0031] It should be noted that during server operation, the primary heat source is the chips configured within the server. This means that the heat generated by the chips needs to be absorbed by the PCM (Polymer Processing Module). For this purpose, an integrated thermal buffer device can be designed, such as... Figure 2 As shown, the integrated heat buffer device may include a gradient PCM, an embedded heat pipe, and an integrated cold plate.

[0032] The design employs a gradient phase transition temperature (PTH) layered PCM design, dividing the PCM into multiple regions. Each region can have a different PTH temperature, with the PTH temperature of the heat-dissipating PTH material closer to the heat-generating chip being lower than that of the material farther away. It's conceivable that when the chip starts generating heat, the heat is first transferred to the nearest PCM. Therefore, the PTH temperature of the nearest PCM can be set to a lower value, allowing it to preferentially absorb heat from the chip and respond quickly to sudden loads. Conversely, the PTH temperature of the more distant PCMs can be set to a higher value, providing sustained buffering capability.

[0033] For example, in Figure 2As can be seen, the PCM is divided into three regions. PCM region 1 is the closest to the chip. The PCM in region 1 can be made of paraffin-based composite graphite material, and its phase transition temperature can be set to 40°. That is to say, when the chip temperature exceeds 40°, the PCM in region 1 will begin to absorb heat through solid-liquid conversion. PCM region 3 is the furthest from the chip. The PCM in region 3 can be made of high thermal conductivity composite material, and its phase transition temperature can be set to 60°. That is to say, only when the chip temperature exceeds 60° will the PCM in region 3 begin to absorb heat through solid-liquid conversion. PCM region 2 is located in the middle. The PCM in region 2 can be made of metal foam material, and its phase transition temperature can be set to 50°.

[0034] By creating a temperature gradient through gradient PCM design, the buffer time is extended. The low-temperature phase change material near the chip side preferentially absorbs heat and responds quickly to sudden loads, while the high-temperature phase change materials in the middle and far ends provide continuous buffering capabilities. By selecting gradient high-performance composite PCM materials, significant temperature rise reduction, reduced heat dissipation energy consumption in low-load scenarios, and extended hardware lifespan are achieved, resulting in technical benefits.

[0035] Heat pipe technology enables rapid heat transfer to the outside of components when the PCM (Polymerized Phase Change Material) begins dissipating heat. A heat pipe consists of a sealed conduit and an internal condensate liquid. At the high-temperature end, the liquid evaporates and moves towards the low-temperature end. At the low-temperature end, the gas releases heat and condenses, rapidly transferring the heat. The condensed liquid then flows back to the evaporation end through capillary action or gravity, completing the cycle. When the PCM absorbs heat and reaches its phase change temperature, it changes from a solid to a liquid state, absorbing a large amount of heat while maintaining a relatively constant temperature. The PCM acts as a heat buffer, and subsequently, when the system load decreases, it releases the stored heat through the heat pipe. The efficient heat conduction characteristics of heat pipe technology are suitable for high-density server environments, achieving good heat dissipation performance within a limited space. In other words, the heat-dissipating phase change material absorbs heat from the server, changing from a solid to a liquid state, and then returns to a solid state after transferring the heat through the heat pipe.

[0036] In some embodiments, the evaporation end of the heat pipe can be directly embedded inside the PCM, instead of the traditional external bonding method. Specifically, the heat pipe can be a superconducting heat pipe with a diameter of 3-6 mm; the length of the evaporation end can be 15-25 mm, directly embedded inside the PCM; the spacing between the heat pipes is ≤8 mm, arranged in a dense array; thermally conductive silicone is used to fill the gap between the heat pipe and the PCM to reduce contact thermal resistance; the condensation end of the heat pipe is integrally connected to the cold plate substrate. This effectively increases the contact area with the PCM, allowing heat inside the PCM to be quickly dissipated through the heat pipe, significantly improving thermal conductivity (approximately 60%), and forming an integrated structure between the heat pipe and the PCM. Figure 2 As can be seen, heat pipes can be directly embedded with PCMs in various regions.

[0037] Alternatively, the condenser end of the heat pipe can be integrated with the cold plate substrate. The cold plate substrate can be made of alloy material, with the flow channels and the condenser end of the heat pipe arranged alternately. The surface of the cold plate can be anodized to improve corrosion resistance. Compared with traditional cold plates, this significantly increases the heat exchange area (about 60%) and reduces the volume (about 20%), improving space utilization. There is no additional interface between the heat pipe and the cold plate, reducing thermal resistance.

[0038] In some embodiments, the PCM's heat absorption from the server can be understood as passive absorption. That is, as long as the server temperature reaches the PCM's phase transition temperature, the PCM will automatically absorb heat, and the PCM will not actively choose whether to absorb heat. However, the heat pipe cooling strategy can be understood as active cooling. The heat pipe cooling strategy can adjust the cooling efficiency by adjusting the heat pipe's cooling power. In other words, after the PCM passively absorbs the server's heat, the heat pipe can actively choose how much cooling power to transfer the heat, based on the heat pipe cooling strategy corresponding to the thermal risk index calculated based on the operating state parameters mentioned above.

[0039] In some embodiments, the heat pipe cooling strategy includes the opening degree of the heat valve on the heat pipe and the rotation speed of the liquid cooling pump. Different heat valve opening degrees result in different coolant flow rates, and different liquid cooling pump rotation speeds result in different coolant circulation efficiency. Since the server heat absorbed by the heat dissipation phase change material is transferred through the flow of coolant in the heat pipe, the flow of coolant affects the server's heat dissipation. The heat pipe cooling strategy instructs increasing the heat valve opening and the liquid cooling pump rotation speed to improve the heat pipe's cooling power, enabling the heat pipe to quickly transfer heat from the heat dissipation phase change material to the external environment, ensuring cooling efficiency. If the current server temperature is relatively stable and rapid cooling is not required, the heat pipe cooling strategy instructs decreasing the heat valve opening and the liquid cooling pump rotation speed. Since the flow of coolant in the heat pipe and the operation of heat dissipation components such as the liquid cooling pump generate power consumption and noise, reducing the heat pipe's cooling power minimizes the power consumption and operating noise of heat dissipation components while ensuring normal heat dissipation. In other words, the higher the thermal risk index, the greater the opening of the thermal valve and the speed of the liquid cooling pump, allowing the coolant to flow rapidly in the heat pipe, thereby quickly transferring the server heat absorbed by the heat dissipation phase change material and achieving heat dissipation.

[0040] In some embodiments, since the coolant needs to flow within the heat pipe, multiple thermal valves can be configured for the heat pipe to allow for more flexible adjustment of the heat pipe cooling strategy. Each thermal valve can adjust the coolant flow through different opening degrees, and the valve core of each thermal valve can independently control its opening degree. The valve core employs electromagnetic drive and nonlinear opening control. Similar to the gradient PCM design principle, gradient control can also be implemented for the opening degree of each thermal valve. Thermal valves closer to the heat-generating chip require priority in heat dissipation to alleviate emergency temperature rises. Therefore, the valve core weight of thermal valves closer to the heat-generating chip is greater than that of thermal valves farther away from the heat-generating chip. Thus, the opening degree of each thermal valve can be calculated based on different valve core weights and the thermal risk index. The liquid cooling pump speed can also be determined based on the thermal risk index and the current speed.

[0041] In some embodiments, the opening degree of each thermal valve can be expressed by the following formula:

[0042] in, This refers to the opening degree of the i-th thermal valve, with a value ranging from 0 to 1; This is the maximum opening coefficient, with a value ranging from 0.9 to 1; This is the heat risk index. This refers to the valve core weight, which is related to the distance between the thermal valve and the heating element. Thermal valves closer to the heating element have a higher valve core weight. For example: (near)=1.2, (mid)=1.0, (far) = 0.8; n is a non-linear exponent, ranging from 1.5 to 2.0.

[0043] As can be seen from the above formula, when the thermal risk index is low, the thermal valve opening increases slowly to avoid frequent adjustments; when the thermal risk index is high, the thermal valve opening increases rapidly to ensure sufficient heat dissipation capacity; the thermal valve near the chip side opens first to ensure priority heat dissipation in the high-temperature area.

[0044] In addition, a microcontroller can be set up for each thermal valve to receive the thermal risk index, calculate the optimal opening degree of each thermal valve in real time, and drive the thermal valve to operate. It provides feedback on the current status of the thermal valve (position, fault detection), which can control the latency to <50ms (a reduction of about 80% compared to an external controller). It supports multi-machine collaboration, can achieve server-level linkage through CAN bus, has fault self-diagnosis capabilities, and improves system reliability.

[0045] In some embodiments, the liquid cooling pump speed can also be adjusted according to the thermal risk index, which can be achieved using the following formula:

[0046] in, This is the speed of the liquid cooling pump; Minimum rotational speed; This is the maximum speed; This is the thermal risk index, and m is the pump speed regulation index, with a value range of 1.2-1.8.

[0047] As can be seen from the above formula, when the thermal risk index is low, reducing the speed results in significant energy savings, while when the thermal risk index is high, rapidly increasing the speed ensures sufficient cooling capacity, smooth adjustment, and avoids pressure shocks caused by sudden changes in pump speed.

[0048] In this embodiment, the heat dissipation phase change material can automatically absorb the heat generated in the server through solid-liquid conversion. Based on the thermal risk index, the specific values ​​of the heat pipe's thermal valve opening and the liquid cooling pump speed are calculated to achieve heat dissipation with different heat pipe cooling powers. This combines the passive heat dissipation of the heat dissipation phase change material with the active heat dissipation of the heat pipe strategy, achieving effective heat dissipation of the server. In addition, gradient settings are applied to both the heat dissipation phase change material and the thermal valve to ensure that the heat dissipation phase change material and thermal valve near the heat-generating chip can respond promptly and effectively to rapid temperature rise, further improving the reliability of server heat dissipation.

[0049] like Figure 3 As shown, Figure 3 Another flowchart of a server heat dissipation method provided for embodiments of this application, the method may include the following steps: 301. Obtain the server's running status parameters.

[0050] In some embodiments, when the operating status parameters include real-time load, the operating status parameters of the server are obtained, which may specifically include: obtaining the real-time power of the server; calculating a first power difference between the real-time power and the idle power of the server, and a second power difference between the idle power and the maximum power of the server; and determining the real-time load of the server based on the first power difference and the second power difference.

[0051] It should be noted that server load is a key indicator for measuring server operation and performance. It quantifies the amount of tasks or resource usage a server handles within a specific time period, reflecting the server's busyness and potential performance bottlenecks. Therefore, when calculating load, it can be based on the server's real-time power consumption, taking into account the server's power consumption even when idle, as well as its maximum power limit. Thus, the server's real-time load can be expressed as:

[0052] Where L represents the server's real-time load. This refers to the server's real-time power. This is the server's maximum power. This represents the server's idle power.

[0053] In this embodiment of the application, in order to measure the server load, the server load can be characterized by the proportion of the server's real-time power within the power range formed by the server's idle power and maximum power. This can improve the accuracy of the server load and more effectively reflect the server's operating status.

[0054] In some embodiments, operating status parameters may include at least: real-time temperature, real-time load, and heating rate. The real-time temperature represents the current temperature of the server. The real-time load can be calculated using the power parameters described above. The heating rate can be understood as the rate at which the server temperature rises. Generally, server temperature changes relatively gradually; however, if there is a sudden and significant increase in load, the server temperature may rise rapidly in a short period, which can affect the normal operation of the server. Therefore, it is necessary to pay attention not only to the server's real-time temperature but also to its heating rate. When operating status parameters include real-time temperature, real-time load, and heating rate, a weighted sum can be performed according to the weights of each operating status parameter to obtain a thermal risk index, which will be described separately below.

[0055] 302. Determine the temperature risk index based on real-time temperature and the first weight.

[0056] In this embodiment, the real-time temperature can be the temperature of the server at the current moment, which can be detected by a temperature sensor installed inside the server. The first weight can be a weight set for the real-time temperature as an operating status parameter. Therefore, the temperature risk index can be determined based on the real-time temperature and the first weight.

[0057] In some embodiments, a temperature risk index is determined based on real-time temperature and a first weight, which may specifically include: obtaining the maximum safe temperature of the server; and determining the temperature risk index based on the ratio between the real-time temperature and the maximum safe temperature and the first weight.

[0058] It should be noted that the thermal risk index needs to be calculated from a temperature perspective. In other words, the current thermal risk needs to be assessed from a temperature perspective. Therefore, it is necessary to determine the risk situation of the current temperature value. If there is only a real-time temperature value, it is impossible to make a judgment. Therefore, the maximum safe temperature of the server can be determined. The maximum safe temperature can be the maximum temperature that the server can withstand during normal operation. If the temperature is exceeded, it may cause damage to the server and its components. Then, the ratio between the real-time temperature and the maximum safe temperature is weighted to obtain the temperature risk index.

[0059] 303. Determine the load risk index based on real-time load and the second weight.

[0060] In this embodiment of the application, the second weight can be a weight set for the operating status parameter of real-time load. After the real-time load is determined, since the real-time load itself is the proportion of the current power, the real-time load can be directly weighted to determine the load risk index.

[0061] 304. Determine the temperature rise risk index based on the temperature rise rate and the third weight.

[0062] In this embodiment, the heating rate can be the rate at which the server's temperature rises, and the third weight can be a weight set for the heating rate as an operating status parameter. Therefore, the heating risk index can be determined based on the heating rate and the third weight.

[0063] It should be noted that the first weight, second weight, and third weight mentioned above are weights set for each running state parameter. The weight values ​​can be set by the user or according to the current scenario requirements. The weight values ​​can also be adjusted according to the actual scenario requirements. However, each weight is a value greater than or equal to 0 and less than or equal to 1, and it is necessary to ensure that the sum of the first weight, second weight, and third weight is 1.

[0064] 305. Determine the thermal risk index based on the temperature risk index, load risk index, and temperature rise risk index.

[0065] In this embodiment of the application, after determining the temperature risk index, load risk index and temperature rise risk index respectively, the temperature risk index, load risk index and temperature rise risk index can be added together to obtain the server's thermal risk index. The value range of the thermal risk index is greater than 0 and less than 1. The closer the thermal risk index is to 1, the higher the thermal risk, and the more likely the server is to experience a heat peak.

[0066] In some embodiments, the current temperature rise of the server can reflect the real-time operation of the server. However, the thermal risk index can be understood as a prediction of a future heat peak. Therefore, when calculating the thermal risk index, the future temperature rise rate can also be predicted. Thus, the temperature rise rate can include the current temperature rise rate and the predicted temperature rise rate.

[0067] Therefore, a heating risk index is determined based on the heating rate and a third weight. Specifically, this may include: obtaining the server's maximum heating rate during a historical period; determining a first heating index based on the ratio between the current heating rate and the maximum heating rate and a first heating weight; and determining a second heating index based on the ratio between the predicted heating rate and the maximum heating rate and a second heating weight.

[0068] It should be noted that the maximum heating rate can be the maximum heating rate monitored during the server's normal operation in the past. Exceeding this heating rate may cause damage to the server and its components due to excessively rapid heating. This maximum heating rate can be used as a benchmark value to normalize the current heating rate and the predicted heating rate. In this way, based on the maximum heating rate, weights can be assigned to the normalized current heating rate and the predicted heating rate respectively. The first heating weight and the second heating weight can also be determined according to the actual scenario. In this case, for calculating the thermal risk index, four variables can be included: real-time temperature, real-time load, current heating rate, and predicted heating rate. Then, the sum of the first weight corresponding to real-time temperature, the second weight corresponding to real-time load, the first heating weight corresponding to current heating rate, and the second heating weight corresponding to predicted heating rate must be 1.

[0069] Based on this, the thermal risk index is determined according to the temperature risk index, load risk index, and temperature rise risk index. Specifically, the thermal risk index can be determined based on the temperature risk index, load risk index, first temperature rise index, and second temperature rise index.

[0070] It should be noted that when the heating rate includes the current heating rate and the predicted heating rate, there are a total of four changes, which correspond to four risk indices: temperature risk index, load risk index, first heating index and second heating index. Therefore, the final thermal risk index can be determined based on the temperature risk index, load risk index, first heating index and second heating index.

[0071] In some embodiments, in high-density servers, multiple chips such as CPUs, GPUs, memory, and accelerator cards are closely arranged, resulting in significant thermal coupling effects. When one chip generates heat, the heat is transferred to adjacent chips through PCB heat conduction, air convection, and shared heat sinks, causing the temperature of adjacent chips to passively rise (thermal crosstalk), which may trigger unnecessary frequency reduction or performance loss. Therefore, when calculating the thermal risk index, the thermal coupling effect between chips can be further considered. A multi-chip thermal coupling coefficient can be determined and weighted, and this multi-chip thermal coupling coefficient can be a constant.

[0072] In summary, this heat risk index can be expressed by the following formula:

[0073] in, This is the heat risk index. , For real-time temperature, Where L is the maximum safe temperature and L is the real-time load. , Given the current rate of temperature increase, For maximum heating rate, , To predict the rate of temperature rise, This represents the thermal coupling coefficient of a multi-chip system. , , , and All are weights. .

[0074] In some embodiments, there are multiple weights in the above formula, and the value of each weight can be dynamically adjusted according to the working scenario. Specifically, a scenario feature vector can be generated based on real-time temperature, real-time load, and current heating rate to describe the current working scenario. Then, based on the scenario feature vector, parameter matrix, and bias vector, the probability distribution of each scenario is output through a normalized exponential function, thereby corresponding to the value of each weight.

[0075] The scene feature vector can be represented by the following formula:

[0076] Where S is the scene feature vector. , For real-time temperature, Maximum safe temperature; , Given the current rate of temperature increase, This represents the maximum heating rate. , Given the current load growth rate, This represents the maximum load growth rate.

[0077] Furthermore, the scene classification result is calculated as Softmax(W·S+b). Here, S is the scene feature vector, W and b are the parameter matrix and bias vector trained using historical data, respectively; Softmax is the normalized exponential function, outputting the probability distribution for each scene. This probability distribution can be used to determine the heat risk index. , and The value of .

[0078] For example, for steady-state low-load scenarios: α=0.6, β=0.2, γ=0.1, δ=0.1, ε=0.0, focusing on the absolute temperature value; for scenarios with rapidly increasing load: α=0.2, β=0.4, γ=0.3, δ=0.1, ε=0.0, focusing on load and temperature rise; for sudden high-load scenarios: α=0.1, β=0.2, γ=0.4, δ=0.2, ε=0.1, focusing on the rate of temperature rise and prediction; for multi-chip collaborative scenarios: α=0.2, β=0.2, γ=0.3, δ=0.15, ε=0.15, focusing on thermal coupling effects.

[0079] In this embodiment, the server's temperature, load, current heating rate, predicted heating rate, and thermal coupling effect between chips can be comprehensively considered and weighted to calculate the current thermal risk index. This allows the final thermal risk index to fully reflect the risk status of the server's current operating parameters, while also taking into account future heating conditions. This enables early prediction of heat peaks, addresses the response delay problem of the heat dissipation system, and significantly reduces the temperature rise during heat peaks. Consequently, more accurate heat dissipation strategies can be adopted in advance, better meeting the server's heat dissipation needs.

[0080] In some embodiments, the heating rate includes a current heating rate and a predicted heating rate. The current heating rate can be calculated based on the real-time temperature, and the predicted heating rate can be predicted based on load changes. Specifically, the current load change rate can be determined based on the real-time load; the predicted load can be determined based on the real-time load and the current load change rate; the predicted temperature can be determined based on the predicted load and the load-temperature mapping function; and the predicted heating rate can be determined based on the predicted temperature and the real-time temperature.

[0081] It should be noted that, firstly, the current load change rate can be calculated based on the current load, which can be obtained from the load at the current moment and the load at past moments. Then, combining the real-time load at the current moment and the current load change rate, the load at future moments can be predicted. This prediction time window can be set by the user, such as 5-10 seconds, to obtain the predicted load, which can be specifically expressed as: = L+kL·(dL / dt)·Δt, where, This refers to the predicted load, where L is the real-time load, kL is the load-temperature rise conversion factor (which can be calibrated using historical data, typically 0.5-1.5), (dL / dt) is the current load change rate, and Δt is the prediction time window. Since high loads lead to temperature increases, the predicted load can be converted to a predicted temperature based on a load-temperature mapping function. This function can be obtained through polynomial fitting or neural network model training, and can be specifically expressed as: =f( )·Δt+T, where, That is, predicting the temperature. That is, the predicted load, f( The load temperature mapping function is Δt, which is the prediction time window, and T is the real-time temperature. Finally, based on the predicted temperature and the real-time temperature, the predicted heating rate can be obtained based on the prediction time window.

[0082] 306. Determine the current risk level of the server based on the risk range of the heat risk index.

[0083] In this embodiment of the application, after calculating the heat risk index, the current risk level of the server can be determined based on the heat risk index. Multiple risk levels can be preset, and a risk range can be set for each risk level. After determining the heat risk index, the risk range in which the heat risk index is located can be determined, and the risk level corresponding to the risk range can be determined as the current risk level of the server. The higher the heat risk index, the higher the corresponding risk level, indicating that a heat peak is more likely to occur.

[0084] In some embodiments, before determining the current risk level of the server based on the risk range of the thermal risk index, the method may further include: detecting the magnitude relationship between the thermal risk index and a preset risk threshold; when the thermal risk index is greater than or equal to the preset risk threshold, determining that the server needs to dissipate heat, and determining the risk range of the thermal risk index.

[0085] It should be noted that if the thermal risk index is relatively low, it means that the probability of a heat peak is low, and normal heat dissipation is sufficient. There is no need to take additional heat pipe cooling strategies. Therefore, a preset risk threshold can be used for preliminary screening. Only when the thermal risk index is greater than or equal to the preset risk threshold is the server considered to need heat dissipation. Thus, the risk range of the thermal risk index can be determined. If the thermal risk index is less than the preset risk threshold, it is not necessary to determine the risk range of the thermal risk index and no further steps need to be performed.

[0086] 307. Based on the risk level, determine the heat pipe cooling strategy corresponding to the risk level.

[0087] In this embodiment of the application, after determining the risk level, a heat pipe cooling strategy corresponding to the risk level can be determined. That is, different heat pipe cooling strategies can be set in advance for each risk level. Since the heat pipe cooling strategy includes the opening degree of the heat valve and the speed of the liquid cooling pump, the higher the risk level, the larger the opening degree of the heat valve and the speed of the liquid cooling pump can be set; the lower the risk level, the smaller the opening degree of the heat valve and the speed of the liquid cooling pump can be set.

[0088] For example, a thermal risk index < 0.5 indicates low risk, and the current heat dissipation intensity is maintained; a thermal risk index ≤ 0.5 < 0.8 indicates medium risk, and heat dissipation is moderately increased; a thermal risk index ≥ 0.8 indicates high risk, and the strong cooling mode is activated, the opening degree of each valve core and the pump speed are calculated, control is executed, valve core opening commands are issued, and speed commands are issued to the liquid cooling pump.

[0089] In this embodiment of the application, multiple risk levels can be set in combination with the thermal risk index, and corresponding risk ranges can be set to accurately determine the current risk level, thereby determining an accurate and effective heat pipe cooling strategy to quickly and effectively cool the server.

[0090] 308. During the implementation of the heat pipe cooling strategy, calculate the server's thermal risk index in real time.

[0091] In the embodiments of this application, the heat pipe cooling strategy is equivalent to accelerating heat dissipation, which may reduce the thermal risk level of the server. In other words, the probability of a heat peak will be reduced, so it is not necessary to continuously execute the heat pipe cooling strategy. Therefore, the thermal risk index of the server can be calculated in real time.

[0092] 309. When the detected thermal risk index is less than the preset risk threshold, stop executing the heat pipe cooling strategy.

[0093] In this embodiment of the application, a corresponding preset risk threshold can be set for the thermal risk index. If the thermal risk index is greater than or equal to the preset risk threshold, it indicates that the current thermal risk level of the server is high and requires rapid heat dissipation. If the thermal risk index is less than the preset risk threshold, it indicates that the thermal risk level of the server is within the normal range and the server can perform normal heat dissipation without the need to adopt an additional heat pipe cooling strategy. Therefore, the heat pipe cooling strategy can be stopped.

[0094] In this embodiment of the application, during the execution of the heat pipe cooling strategy, the change of the thermal risk index can be monitored in real time. If the thermal risk index has been reduced to less than the threshold, then the heat pipe cooling strategy needs to be executed again. This can effectively reduce the power consumption of the heat dissipation components while ensuring normal heat dissipation of the server.

[0095] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0096] like Figure 4 As shown, embodiments of this application also provide a server heat dissipation device, which may include: Module 401 is used to obtain the server's running status parameters; The processing module 402 is used to determine the thermal risk index of the server based on the operating status parameters. The thermal risk index is used to indicate the probability that the server will reach its thermal peak. The processing module 402 is also used to determine the heat pipe heat dissipation strategy corresponding to the thermal risk index, wherein the heat pipe heat dissipation power corresponding to the heat pipe heat dissipation strategy is positively correlated with the thermal risk index. The processing module 402 is also used to execute a heat pipe cooling strategy, so that the heat stored in the heat dissipation phase change material of the server is transferred through the heat pipe, and the heat dissipation phase change material is used to absorb the heat of the server through solid-liquid conversion.

[0097] In some embodiments, the acquisition module 401 is specifically used to acquire the real-time power of the server; The processing module 402 is specifically used to calculate the first power difference between the real-time power and the idle power of the server, and the second power difference between the idle power and the maximum power of the server. The processing module 402 is specifically used to determine the real-time load of the server based on the first power difference and the second power difference.

[0098] In some embodiments, the processing module 402 is specifically configured to determine a temperature risk index based on the real-time temperature and a first weight. The processing module 402 is specifically used to determine the load risk index based on the real-time load and the second weight. The processing module 402 is specifically used to determine the heating risk index based on the heating rate and the third weight; The processing module 402 is specifically used to determine the thermal risk index based on the temperature risk index, the load risk index, and the temperature rise risk index. The sum of the first weight, the second weight, and the third weight is 1.

[0099] In some embodiments, the heating rate includes: the current heating rate and the predicted heating rate. The acquisition module 401 is specifically used to acquire the maximum heating rate of the server in a historical period. Processing module 402 is specifically used to determine the first heating index based on the ratio between the current heating rate and the maximum heating rate and the first heating weight; The processing module 402 is specifically used to determine the second heating index based on the ratio between the predicted heating rate and the maximum heating rate and the second heating weight; The processing module 402 is specifically used to determine the thermal risk index based on the temperature risk index, the load risk index, the first temperature rise index, and the second temperature rise index. The sum of the first weight, the second weight, the first heating weight, and the second heating weight is 1.

[0100] In some embodiments, the processing module 402 is specifically configured to determine the current load change rate based on the real-time load; The processing module 402 is specifically used to determine the predicted load based on the real-time load and the current load change rate; Processing module 402 is specifically used to determine the predicted temperature based on the predicted load and load temperature mapping function; The processing module 402 is specifically used to determine the predicted heating rate based on the predicted temperature and the real-time temperature.

[0101] In some embodiments, the acquisition module 401 is specifically used to acquire the maximum safe temperature of the server; The processing module 402 is specifically used to determine the temperature risk index based on the ratio between the real-time temperature and the maximum safe temperature and the first weight.

[0102] In some embodiments, the heat-dissipating phase change material absorbs heat from the server to change from a solid state to a liquid state, and then returns to a solid state after the heat is transferred through a heat pipe; for heat-generating chips in the server, the phase change temperature of the heat-dissipating phase change material near the heat-generating chip is lower than the phase change temperature of the heat-dissipating phase change material far from the heat-generating chip.

[0103] In some embodiments, the heat pipe cooling strategy includes the opening degree of the heat pipe's thermal valves and the rotational speed of the liquid cooling pump. The evaporation end of the heat pipe is embedded inside a heat-dissipating phase change material. The heat pipes are arranged in an array, and the heat pipes are equipped with multiple thermal valves. The valve core weight of the thermal valve closer to the heat-generating chip is greater than that of the thermal valve farther away from the heat-generating chip. The opening degree of each thermal valve is determined based on the thermal risk index and the valve core weight. The rotational speed of the liquid cooling pump is determined based on the current rotational speed of the liquid cooling pump and the thermal risk index.

[0104] In some embodiments, the processing module 402 is specifically used to calculate the thermal risk index of the server in real time during the execution of the heat pipe cooling strategy. The processing module 402 is specifically used to stop executing the heat pipe heat dissipation strategy when the detected thermal risk index is less than the preset risk threshold.

[0105] For a description of the features in the embodiment corresponding to the server heat dissipation device, please refer to the relevant description in the embodiment corresponding to the server heat dissipation method, which will not be repeated here.

[0106] like Figure 5 As shown, embodiments of this application also provide an electronic device, including a memory 501 and a processor 502. The memory 501 stores a computer program, and the processor 502 is configured to run the computer program to perform the steps in any of the above-described server heat dissipation method embodiments.

[0107] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described embodiments of the server heat dissipation method.

[0108] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0109] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described server heat dissipation method embodiments.

[0110] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above-described server heat dissipation method embodiments.

[0111] Any of the components, modules, units, parts, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuitry), manual processing, or any combination thereof. Alternatively or additionally, any functionality described herein can be executed at least in part by one or more hardware logic components, such as, but not limited to, a central processing unit (CPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an application-specific standard product (ASSP), a system-on-a-chip (SoC), a complex programmable logic device (CPLD), a microprocessor (MCU), etc. The terms "system," "computing device," or "apparatus" as used herein encompass various means, devices, and machines for processing data, including, for example, one or more programmable processors, computers, SoCs, or combinations thereof. The apparatus may also include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or one or more combinations thereof. The aforementioned computer program (also known as a program, software, software application, app, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for a computing environment.

[0112] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0113] The above provides a detailed description of a server heat dissipation method and electronic device provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for heat dissipation of a server, characterized in that, The method includes: Obtain the server's operating status parameters, which include at least: real-time temperature, real-time load, and heating rate, wherein the heating rate includes: current heating rate and predicted heating rate; Based on the real-time temperature, the real-time load, and the current heating rate, a scene feature vector is generated. The calculation formula for the scene feature vector is as follows: Where S is the scene feature vector. , For real-time temperature, Maximum safe temperature; , Given the current rate of temperature increase, This represents the maximum heating rate. , Given the current load growth rate, This represents the maximum load growth rate. Based on the scene feature vector, parameter matrix, and bias vector, the probability distribution of each working scene is obtained through a normalized exponential function. According to the probability distribution of each working scenario, the weights of the operating status parameters corresponding to the current working scenario are determined. The weights include: a first weight corresponding to the real-time temperature, a second weight corresponding to the real-time load, a first heating weight corresponding to the current heating rate, a second heating weight corresponding to the predicted heating rate, and a thermal coupling weight corresponding to the multi-chip thermal coupling coefficient. The multi-chip thermal coupling coefficient is used to indicate the thermal coupling effect between the chips in the server. The sum of the first weight, the second weight, the first heating weight, the second heating weight, and the thermal coupling weight is 1. The temperature risk index is determined based on the real-time temperature and the first weight. The load risk index is determined based on the real-time load and the second weight. Obtain the maximum temperature rise rate of the server within a historical time period; The first heating index is determined based on the ratio between the current heating rate and the maximum heating rate and the first heating weight; The second heating index is determined based on the ratio between the predicted heating rate and the maximum heating rate, and the second heating weight. The thermal risk index is determined by summing the temperature risk index, the load risk index, the first temperature rise index, the second temperature rise index, and the multi-chip thermal coupling index. The thermal risk index indicates the probability that the server will reach a heat peak. The multi-chip thermal coupling index is determined based on the multi-chip thermal coupling coefficient and the thermal coupling weight. The formula for calculating the thermal risk index is as follows: in, The heat risk index, , For real-time temperature, Where L is the maximum safe temperature and L is the real-time load. , Given the current rate of temperature increase, For maximum heating rate, , To predict the rate of temperature rise, The thermal coupling coefficient for multi-chip applications; , , , and All are weights. ; Determine the heat pipe cooling strategy corresponding to the thermal risk index, wherein the heat pipe cooling power corresponding to the heat pipe cooling strategy is positively correlated with the thermal risk index; The heat pipe cooling strategy is implemented to transfer the heat stored in the heat-dissipating phase change material of the server through the heat pipe, the heat-dissipating phase change material being used to absorb the heat of the server through solid-liquid conversion.

2. The method according to claim 1, characterized in that, The operating status parameters include: real-time load. Obtaining the server's operating status parameters includes: Obtain the real-time power of the server; Calculate the first power difference between the real-time power and the idle power of the server, and the second power difference between the idle power and the maximum power of the server. The real-time load of the server is determined based on the first power difference and the second power difference.

3. The method according to claim 1, characterized in that, The process of obtaining the server's operating status parameters includes: Determine the current load change rate based on the real-time load; The predicted load is determined based on the real-time load and the current load change rate. The predicted temperature is determined based on the predicted load and load temperature mapping function. The predicted heating rate is determined based on the predicted temperature and the real-time temperature.

4. The method according to claim 1, characterized in that, The step of determining the temperature risk index based on the real-time temperature and the first weight includes: Obtain the maximum safe temperature of the server; The temperature risk index is determined based on the ratio between the real-time temperature and the maximum safe temperature, and the first weight.

5. The method according to claim 1, characterized in that, The heat dissipation phase change material absorbs heat from the server to change from a solid to a liquid state, and then returns to a solid state after the heat is transferred through the heat pipe; for the heat-generating chip in the server, the phase change temperature of the heat dissipation phase change material closer to the heat-generating chip is lower than the phase change temperature of the heat dissipation phase change material farther away from the heat-generating chip.

6. The method according to claim 5, characterized in that, The heat pipe cooling strategy includes the opening degree of the heat pipe's thermal valves and the rotation speed of the liquid cooling pump. The evaporation end of the heat pipe is embedded inside the heat dissipation phase change material. The heat pipes are arranged in an array, and each heat pipe is equipped with multiple thermal valves. The valve core weight of the thermal valve closer to the heat-generating chip is greater than that of the thermal valve farther away from the heat-generating chip. The opening degree of each thermal valve is determined based on the thermal risk index and the valve core weight. The rotation speed of the liquid cooling pump is determined based on the current rotation speed of the liquid cooling pump and the thermal risk index.

7. The method according to claim 1, characterized in that, The execution of the heat pipe cooling strategy, which transfers the heat stored in the heat-dissipating phase change material of the server through the heat pipe, includes: During the execution of the heat pipe cooling strategy, the thermal risk index of the server is calculated in real time. When the thermal risk index is detected to be less than a preset risk threshold, the heat pipe cooling strategy is stopped.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the heat dissipation method for a server as described in any one of claims 1 to 7 when executing the computer program.

Images