Server heat dissipation test system

Abstract

This disclosure relates to a server heat dissipation testing system. The system includes: at least one heat dissipation testing device; each device controls a drive component via a first controller, thereby controlling the heating power of a corresponding heat-generating module; multiple heat-generating modules are each equipped with a heatsink under test; at least one of the heatsinks under test is a wind-cooled heatsink; the testing device further includes an air supply component corresponding to the wind-cooled heatsink under test; the first controller controls each air supply component to draw ambient air into the air intake duct, the ambient air flows through the drive component and absorbs heat from the drive component to form a preheated airflow, which is then output to the wind-cooled heatsink under test. This implementation addresses the problems in existing technologies where an ideal cold air inflow environment leads to an overestimation of heat dissipation performance, while introducing additional thermal simulation equipment in a compact testing space disrupts the airflow field and increases costs.

Description

Technical Field

[0001] This disclosure relates to the field of testing, and more specifically, to a server heat dissipation testing system. Background Technology

[0002] In the research and development and production of high-performance AI servers, in order to verify the reliability of heat dissipation, heat dissipation testing devices are usually used to replace real high-value computing hardware for blind thermal testing before large-scale installation. This reduces hardware testing costs and avoids the risk of hardware burnout caused by heat dissipation failure.

[0003] Currently, heat dissipation testing devices are mainly used to test the heat dissipation performance of heat sinks on graphics cards. Therefore, existing heat dissipation testing devices only simulate the heat generation of GPU chips and use ideal room temperature air in the laboratory as the heat dissipation medium by default.

[0004] However, in real-world operating environments, servers have compact internal spaces and contain complex cabling, motherboards, power supply modules, and other heat-generating components. The heat generated by these components accumulates inside the chassis, resulting in the incoming air temperature blowing towards the rear-mounted heatsink being much higher than the ideal cool air temperature in the laboratory. This leads to a systematic overestimation of the thermal performance of existing devices. Furthermore, given the limited internal space of thermal testing equipment, additional thermal simulation equipment to simulate multiple heat-generating components is not only difficult to install, space-consuming, and costly, but it also disrupts the internal airflow pattern. Summary of the Invention

[0005] The purpose of this disclosure is to provide a server heat dissipation testing system to solve the problems of the existing technology where the ideal cold air inflow environment leads to an overestimation of heat dissipation performance, while the introduction of additional thermal simulation equipment in a compact test space will disrupt the airflow field and increase costs.

[0006] To achieve the above objectives, the present disclosure adopts the following technical solution:

[0007] This disclosure provides a server heat dissipation testing system, comprising: at least one heat dissipation testing device; the heat dissipation testing device includes a test housing, a first controller respectively disposed within the test housing, a plurality of heat-generating modules, and drive components connected to each of the heat-generating modules in a one-to-one correspondence; the first controller is respectively connected to the plurality of drive components to control the drive components to adjust the heat generation power of the corresponding heat-generating modules; each heat-generating module is respectively provided with a heat sink to be tested; at least one of the plurality of heat sinks to be tested includes a heat sink to be tested; the heat dissipation testing device further includes an air supply component corresponding to the heat sink to be tested; the air outlet of the air supply component faces the windward side of the heat sink to be tested; the air inlet of at least one air supply component forms an air inlet duct, and the plurality of drive components are disposed in the air inlet duct; the first controller is connected to each of the air supply components to control each air supply component to draw ambient air in the air inlet duct, so that the ambient air flows through the air inlet duct and passes through the plurality of drive components to form a preheated airflow and is output to the heat sink to be tested.

[0008] According to a preferred embodiment of the first aspect, the heating module includes: a heating plate; a heat-insulating base disposed between the heating plate and the test housing; and a heat-conducting base disposed on the side of the heating plate away from the heat-insulating base.

[0009] According to a preferred embodiment of the first aspect, a plurality of first temperature sensors are uniformly arranged between the heating plate and the heat-conducting base; the plurality of first temperature sensors are all connected to the input terminal of the first controller.

[0010] According to a preferred embodiment of the first aspect, the air supply assembly includes at least one speed-regulating fan; the first controller is communicatively connected to the speed-regulating fan to output a speed control signal to the speed-regulating fan based on temperature signals collected by a plurality of first temperature sensors on the corresponding heating plate.

[0011] According to a preferred embodiment of the first aspect, the heat dissipation testing system includes at least two heat dissipation testing devices, each including at least one first testing device and at least one second testing device; the plurality of heat-generating modules in the first testing device are arranged in alignment along the same horizontal plane within the corresponding testing housing; the plurality of heat-generating modules in the second testing device are arranged in staggered and interleaved arrangement along the same horizontal plane within the testing housing.

[0012] According to a preferred embodiment of the first aspect, at least one of the plurality of heat sinks is a liquid-cooled heat sink to be tested.

[0013] According to a preferred embodiment of the first aspect, the heat dissipation testing system further includes a refrigeration device and a circulation pipeline; the liquid-cooled radiator under test is a liquid-cooled plate, and each of the liquid-cooled plates is connected to the refrigeration device via the circulation pipeline to form a cooling circulation loop.

[0014] According to a preferred embodiment of the first aspect, the circulation pipeline includes an inlet sub-pipe and an outlet sub-pipe; the inlet end of each liquid-cooled plate is connected in parallel with the inlet sub-pipe, and the outlet end of each liquid-cooled plate is connected in parallel with the outlet sub-pipe, so as to uniformly cool the liquid-cooled plate through the refrigeration equipment.

[0015] According to a preferred embodiment of the first aspect, an adjustable wind deflector assembly is further provided inside the test housing; the adjustable wind deflector assembly is disposed between the air supply assembly and the air-cooled heat sink under test; the adjustable wind deflector assembly forms pores with adjustable porosity, the pores being used to adjust the flow rate of the preheating airflow.

[0016] According to a preferred embodiment of the first aspect, the adjustable wind deflector assembly includes: a fixed plate having a plurality of first ventilation holes arranged in an array on the fixed plate; a movable plate fitted to one side of the fixed plate having a plurality of second ventilation holes arranged in an array on the movable plate, the first ventilation holes corresponding one-to-one with the second ventilation holes; and a drive mechanism, a first controller being communicatively connected to the drive mechanism, the drive mechanism being kinetically connected to the movable plate, so as to control the drive mechanism via the first controller to drive the movable plate to slide in a direction parallel to the surface of the fixed plate; wherein, the overlapping portion of the first ventilation holes and the second ventilation holes forms the pores, and when the movable plate slides relative to the fixed plate, the porosity of the air passage pores is adjusted by changing the overlapping area of ​​the first ventilation holes and the second ventilation holes.

[0017] The beneficial effects of this disclosure are as follows:

[0018] This disclosure overcomes the limitation of existing testing devices that only simulate a single GPU heat source by placing the drive component that controls the power of the heat-generating module within the air intake duct of the air supply component. In actual testing, the ambient cool air drawn by the air supply component will inevitably flow through and cool the drive component before reaching the air cooler under test. Since the drive component itself generates considerable heat when driving the high-power heat-generating module, the ambient cool air will naturally absorb this heat during this process, thus being converted into preheated airflow with increased temperature. This not only achieves cooling of the drive component but also overcomes the technical bias in traditional testing fields that treats the drive component as a source of thermodynamic and aerodynamic interference and deliberately isolates it from the main air duct. By coordinating the drive component, air supply component, and air cooler, the heating effect of peripheral components such as the power module and front-end CPU in a real AI server on the cooling airflow is simulated. This reduces the difference in ambient air temperature between the heat dissipation testing system and the real AI server, and the reduction in ambient air temperature difference can further reduce the systematic overestimation of the heatsink performance during testing. At the same time, this disclosure resolves the technical contradiction that adding an independent preheating device within a compact housing would disrupt the airflow field and increase manufacturing costs.

[0019] It should be understood that the above general description and the following detailed description are exemplary only and do not limit this disclosure. Attached Figure Description

[0020] The accompanying drawings, which are incorporated herein and form part of this specification, illustrate one or more embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure and to enable those skilled in the art to make and use the present disclosure.

[0021] Figure 1 A schematic diagram of the structure of a server heat dissipation testing system according to a first embodiment of the present disclosure is shown.

[0022] Figure 2 A schematic diagram of the structure of a server heat dissipation testing system according to a second embodiment of the present disclosure is shown.

[0023] Figure 3 A schematic diagram of the structure of a server heat dissipation testing system according to a third embodiment of this disclosure is shown.

[0024] Figure 4 A schematic diagram of the heating module of this disclosure is shown.

[0025] Figure 5 A schematic diagram of the temperature sensor installation structure of this disclosure is shown.

[0026] Figure 6 A schematic diagram of the connection between the first controller and the temperature sensor of this disclosure is shown.

[0027] Figure 7A schematic diagram of the structure of the driving component of this disclosure is shown.

[0028] Figure 8 A schematic diagram of the structure of a server heat dissipation testing system according to the fourth embodiment of this disclosure is shown.

[0029] Figure 9 A schematic diagram of the structure of a server heat dissipation testing system according to the fifth embodiment of this disclosure is shown.

[0030] Figure 10 A schematic diagram of the server heat dissipation testing system according to the sixth embodiment of this disclosure is shown.

[0031] Figure 11 A schematic diagram of the structure of a liquid-cooled refrigeration device according to a first embodiment of the present disclosure is shown.

[0032] Figure 12 A schematic diagram of the structure of a liquid-cooled refrigeration device according to a second embodiment of the present disclosure is shown.

[0033] Figure 13 A schematic diagram of the server heat dissipation testing system according to the sixth embodiment of this disclosure is shown.

[0034] Figure 14 A structural schematic diagram of the adjustable wind deflector assembly of this disclosure is shown.

[0035] Figure 15 A schematic diagram of the base structure of this disclosure is shown.

[0036] Figure 16 A schematic diagram of the test housing of this disclosure is shown.

[0037] Figure 17 A schematic diagram of the cascaded structure of the heat dissipation testing devices disclosed herein is shown. Detailed Implementation

[0038] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments may be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, the description of these embodiments is intended to make this disclosure more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0039] With the explosive growth of artificial intelligence technology, the demand for training and inference of large language models has driven the evolution of server hardware architecture. High-end AI servers, represented by Hyperscale Graphics eXtension (HGX) or Open Compute Project Accelerator Module (OCP Accelerator Module, OAM), have GPU chips with power consumption exceeding 700W to 1000W, and exhibit a step-like thermal shock characteristic, transitioning from standby to full load within milliseconds. To avoid the risk of burnout from directly using high-value computing hardware for extreme thermal testing, the industry commonly employs thermal testing devices that use heat dissipation modules instead of actual GPUs for blind testing of heat sinks.

[0040] In related technologies, the design concept of heat dissipation testing devices is as follows:

[0041] In the standard thermal testing phase, since the standard thermal solutions of related technologies can fully meet the temperature control requirements of traditional computing devices such as CPUs, the thermal bottleneck of high-end servers has completely shifted to the ultra-high power consumption GPU clusters. Therefore, the thermal testing devices in related technologies do not need to test the cooling devices installed on the CPU. In addition, the testing devices of related technologies only test the GPU. Therefore, in the design concept of related technologies, peripheral heat-generating devices such as CPUs, motherboards, power supply modules, or hard drives in AI servers are usually ignored, and a single heat source model is adopted; that is, the testing focus is only on the GPU node with the most extreme heat generation and the most demanding requirements for the performance of the heatsink. This method usually directs the ideal ambient cool air directly to the heatsink under test. The premise of this testing method is that the external ambient cool air can blow directly onto the GPU heatsink under test without interference.

[0042] However, in order to meet the extremely high computing power density, the internal space of real AI servers is greatly compressed, and the front of the GPU is usually densely packed with high-power power modules, complex wiring, and motherboard components. In actual operation, before the external cool air reaches the GPU heatsink, it will inevitably pass over these front-end heat-generating components, absorb heat, and be converted into preheated airflow with increased temperature.

[0043] This results in the air temperature blowing onto the heatsink under test in the testing equipment being significantly lower than the actual inflow air temperature inside the actual chassis. Attempting to add separate heating devices within a compact test housing to specifically simulate the thermal radiation of front-end components would not only significantly increase the manufacturing cost of the testing equipment, but these additional devices would also directly encroach on the narrow airflow space, disrupting the original airflow pattern. For example, adding bulky separate heating devices within a compact test housing to specifically simulate front-end thermal radiation would not only significantly increase manufacturing costs, but these additional devices would also become physical obstacles within the main airflow duct. Aerodynamically, this encroachment on narrow airflow ducts would cause a flow obstruction effect, disrupting the airflow pattern.

[0044] The disconnect between the ideal testing environment and real-world harsh operating conditions and the gaseous heat flow causes existing devices to systematically overestimate the actual heat dissipation capacity of the heatsinks. When these heatsinks that perform well in tests are actually deployed in high-density server chassis, they often cannot handle the superimposed thermal load from the preheating airflow, ultimately leading to excessively high GPU temperatures, causing throttling of computing power, or even hardware overheating and crashes, rendering the evaluation results of the testing devices worthless for engineering guidance.

[0045] Based on this, one embodiment of this disclosure provides a server heat dissipation testing system, including:

[0046] At least one heat dissipation testing device;

[0047] Any of the heat dissipation testing devices includes a test housing, a first controller respectively disposed within the test housing, a plurality of heat-generating modules, and a drive component connected to each of the heat-generating modules in a one-to-one correspondence.

[0048] The first controller is connected to multiple drive components respectively, so as to control the drive components to adjust the heating power of the corresponding heating modules through the first controller;

[0049] Each of the aforementioned heat-generating modules is provided with a heat sink to be tested; at least one of the multiple heat sinks to be tested is an air-cooled heat sink to be tested.

[0050] The heat dissipation testing device further includes an air supply component corresponding to the air-cooled radiator under test; the air outlet of the air supply component faces the windward side of the air-cooled radiator under test; at least one of the air supply components has an air inlet forming an air inlet duct, and multiple drive components are disposed in the air inlet duct.

[0051] The first controller is connected to each of the air supply components, so that the first controller controls each of the air supply components to draw ambient air from the air inlet duct, so that the ambient air flows through the air inlet duct and passes through multiple drive components to form a preheated airflow and is output to the air-cooled heat sink under test.

[0052] In the thermal and wind tunnel testing concepts of related technologies, the absolute controlled variable method is usually adopted. That is, only a single heat source, the device used to simulate the heat generation of the graphics card, is allowed in the test system, while parasitic heat sources such as driver components are treated as interfering heat sources. In order to ensure that the test of the heatsink under test is not interfered with, related technologies usually isolate the driver components from the environment in which the heatsink under test is located. This disclosure overcomes the technical bias in the traditional testing field that treats the driver components as a source of thermodynamic and aerodynamic interference and deliberately isolates them from the main airflow. Instead, it takes the opposite approach and places the driver components directly in the air intake airflow. This unconventional design not only does not cause negative interference to the test, but also transforms the parasitic heat of the driver components into a preheating source that simulates the heat generation of the front-end components of a real AI server. Without adding additional dummy load or disrupting the overall airflow smoothness of the airflow, it achieves a high degree of reproduction of the harsh flow field and thermal field environment of a real high-density chassis.

[0053] Specifically, in this application, ambient cool air serves as the cooling medium, effectively carrying away the high-load waste heat from the drive components and ensuring the operational stability of the test system's hardware. This preheated airflow, carrying the heat from the drive components, also replicates the actual inflow air temperature in a real high-density AI server, where external cool air passes over the front-end power module and CPU nodes, reducing the temperature difference between the thermal testing system and a real AI server.

[0054] In addition, this application abandons the traditional intervention method of adding a large independent dummy load in a compact test housing, maintains the physical unobstructed air intake channel, avoids the flow obstruction effect and flow field disruption caused by the addition of physical obstacles, and ensures the smoothness of the test airflow.

[0055] It should be noted that each of the driving components in this embodiment has a power input terminal for connecting to an external power supply. This external power supply is not a limiting component of this disclosure; it can be a standard mains power network in a laboratory, an independently configured regulated DC power supply cabinet, or other equipment capable of providing the required power. This disclosure does not specifically limit its application. Specifically, the input terminal of each driving component is used to connect to the external power supply, and the output terminal is electrically connected to the corresponding heating module. The first controller is communicatively connected to each driving component to issue power control commands to the driving components, thereby achieving independent and dynamic adjustment of the heating power of each heating module.

[0056] It should be noted that those skilled in the art will understand that, due to individual differences and varying working principles of electronic components, the waste heat power generated by the drive component under a specific load may deviate in absolute terms from the combined heat generation of a multi-core CPU and high-power power supply module in a real server at the same node. However, the architecture design in this embodiment, which uses the drive component as a preheating source, demonstrates a high degree of rationality and adaptability in constructing thermodynamic boundaries. Specifically, the greater the heat generation power of the heating module, the greater the current output by the drive component, and the nonlinear increase in its own waste heat generation. This high computing load inevitably leads to a positively correlated physical coupling effect of high initial temperature rise, which naturally aligns with the thermodynamic changes within a real AI server, qualitatively breaking the idealized assumption of direct cold air blowing from room temperature in traditional equipment.

[0057] In a specific example, if there are too many peripheral heat-generating devices in the AI ​​server, the heat generated by peripheral heat-generating devices such as the CPU, motherboard, power supply module or hard drive is much greater than the total heat generated by all driving components. In order to eliminate the residual deviation caused by the different physical properties of the heat sources at the quantitative level, the heat dissipation test device in this embodiment also includes at least one ambient temperature sensor.

[0058] The ambient temperature sensor is disposed in the air inlet duct and is located on the airflow path between the downstream air outlet of the drive component and the air-cooled heat sink under test; the ambient temperature sensor is communicatively connected to the first controller.

[0059] The first controller is configured to acquire the actual preheating air temperature parameters collected by the ambient temperature sensor.

[0060] Following the example above, the first controller has a built-in thermodynamic compensation algorithm. The first controller is configured to: in the operating state of the test device, obtain the average test junction temperature fed back by the first temperature sensor on the i-th heating module and the actual preheating air temperature fed back by the ambient temperature sensor in the air duct through the first controller, calculate the actual temperature difference between the test junction temperature and the actual preheating air temperature; obtain the theoretical inflow air temperature of the target AI server in a real chassis environment; linearly sum the theoretical inflow air temperature and the fixed temperature difference, calculate and output the calibrated target junction temperature of the i-th heating module.

[0061] Specifically, the theoretical inflow air temperature of the target AI server in a real chassis environment is the sum of the theoretical baseline room temperature of the target AI server and the theoretical front-end temperature rise. The theoretical baseline room temperature represents the macroscopic cold air temperature of the data center where the target AI server will be actually deployed. This data is typically set according to Data Center Service Level Agreements or American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHEEE) standards, such as 25°C for a conventional air-cooled data center or 35°C for a high-density data center. The theoretical front-end temperature rise represents the preheating temperature rise caused by the airflow to heat-generating components inside the target AI server chassis, located in front of the GPU under test, such as the CPU, redundant power supply module, hard drive array, and network card, under full load conditions, for example, 12°C. This data is an inherent design attribute of the target model.

[0062] Following the example above, as a specific example, in this embodiment, during the operation of the heat dissipation testing device, the first controller controls the air supply component to operate at full speed, and the heating module to operate at full power. At this time, the waste heat generated by the drive component heats the actual preheating air temperature of the heat sink under test to 35°C. After reaching thermal equilibrium in this state, the first temperature sensor measures the test junction temperature of the heating module to be 80°C. The first controller calculates that the fixed temperature difference of the heat sink under test is 45°C, i.e., 80°C - 35°C = 45°C. This 45°C temperature difference represents the inherent temperature rise resistance of the heat sink under the current airflow and power.

[0063] The researchers input the actual operating environment parameters of the target AI server. Assuming that the machine is fully loaded with more heat-generating devices such as power supplies and hard drive arrays, the front-end heat exhaust will cause the actual intake temperature blowing towards the GPU to reach an even more severe 42°C.

[0064] The first controller directly adds the actual intake air temperature of 42℃ to the fixed temperature difference of 45℃ from the heatsink, ultimately calculating the calibrated target junction temperature of the heat-generating module inside the actual server as 87℃, i.e., 42℃ + 45℃ = 87℃. Researchers use this calibrated target junction temperature output by the first controller to determine the heat dissipation performance of the air-cooled heatsink.

[0065] In a specific example, such as Figure 1 and Figure 2 As shown, this disclosure provides a heat dissipation testing system for an AI server. The system includes at least one modularly designed heat dissipation testing device 00. Within the testing housing, multiple heat-generating modules 30 simulating computing chips are arranged in an array, along with driving components 20 connected one-to-one with each heat-generating module 30.

[0066] The first controller 10 is communicatively connected to each of the drive components 20 and can independently issue power control commands to each drive component 20, thereby achieving precise and dynamic adjustment of the heat dissipation power of the corresponding heat dissipation module 30 to simulate the heat dissipation conditions of the graphics card under different computing loads.

[0067] For computing nodes employing an air-cooled heat dissipation architecture, this device configures an air supply component on the windward side of the air-cooled heatsink 81 under test. This disclosure breaks with the conventional design of traditional testing equipment that isolates the drive circuit from the heat source. Specifically, this device extends the air inlet of the air supply component to form an air inlet duct 84, and integrates multiple drive components 20 within this air inlet duct 84. During test operation, the first controller 10 controls the operation of the air supply component to draw in cool air from the external environment. As this ambient air flows through the air inlet duct 84, it first washes over the surface of the drive components 20, absorbing the waste heat generated by the operation of the drive components 20, forming a preheated airflow with a specific temperature gradient; the preheated airflow is output from the air outlet of the air supply component and directly output to the downstream air-cooled heatsink 81 under test for heat exchange.

[0068] In one possible implementation, the heating module includes:

[0069] Heating plate;

[0070] A heat-insulating base is disposed between the heating plate and the test housing; and...

[0071] A heat-conducting base is disposed on the side of the heating plate away from the heat-insulating base.

[0072] In a specific example, such as Figure 3 As shown, in this embodiment, the first controller 10, the heating module 30, and the drive component 20 are all disposed inside the test housing 50 to simulate the physical space constraints of a real server. A display module 40 is disposed on the side wall of the test housing 50. The display module 40 is connected to the first controller 10 and is used to display various parameters in the first controller 10 in real time. Specifically, in this embodiment, the display module 40 and the first controller 10 are integrated into one unit as a display control device integrating control and display. Furthermore, the interface on the first controller 10 for connecting to the drive component 20 is a pluggable interface, facilitating connection expansion. Specifically, multiple drive components 20 can be integrated onto one drive board (high integration) or disposed on different drive boards (modular design for easy expansion).

[0073] For example, refer to Figure 3 The test housing 50 has an installation cavity, and multiple heating modules 30 are installed at the bottom of the cavity. Additionally, refer to... Figure 4The heating module 30 adopts a multi-layer stacked structure. It includes a heating plate 32, a heat-insulating base 33, and a heat-conducting base 31. The heating plate 32 is disposed between the heat-insulating base 33 and the heat-conducting base 31. The heat-insulating base 33 is located between the heating plate 32 and the test housing 50, and is used to simulate the server motherboard to prevent heat from being directly conducted to the support surface of the test housing 50 and affecting the simulation effect. Specifically, the heat-insulating base 33 can be a PCB board or multiple ceramic heat-insulating pillars evenly arranged on the heating plate 32. The heat-conducting base 31 is disposed on the side of the heating plate 32 away from the heat-insulating base 33, and is usually made of copper or aluminum alloy, and is used to evenly conduct the heat generated by the heating plate 32 upwards. The heating plate 32 is preferably made of aluminum nitride.

[0074] In one possible implementation, a plurality of first temperature sensors are uniformly arranged between the heating plate and the heat-conducting base;

[0075] Multiple first temperature sensors are connected to the input terminal of the first controller. During testing of the heat sink under test mounted on a thermally conductive base, since there may be heat dissipation blind spots within the internal structure of the heat sink, the uniformly distributed array of first temperature sensors in this embodiment can capture the non-uniformity of temperature rise at various points on the thermally conductive surface. Even if the output of the heating plate is uniform, by comparing the differences in temperature rise rates at each sampling point, the uneven distribution of heat dissipation efficiency or localized heat dissipation failure areas of the heat sink under test in space can be accurately located. This provides quantitative data support for evaluating the dynamic suppression capability of AI server heat dissipation modules in the face of extreme transient heat surges.

[0076] In a specific example, such as Figure 5 As shown in the embodiment of this application, a plurality of first temperature sensors 25 are uniformly arranged between the heating plate 32 and the heat-conducting base 31. Figure 6 As shown, multiple first temperature sensors 25 are connected to the input terminal of the first controller 10 corresponding to their respective heating module 30. For example, the first temperature sensor 25 can be a PT100 platinum resistance thermometer or a PT1000 platinum resistance thermometer. To ensure measurement accuracy in high-interference environments, the sensor is connected to the ADC interface of the first controller 10 via a four-wire connection or a dedicated analog-to-digital converter module 212, such as a MAX31865, is used.

[0077] Following the example above, such as Figure 7 As shown, any of the drive components 20 in this embodiment includes a power output device 22, a power sampling circuit 23, and a second controller 21; wherein the first controller 10 and the second controller 21 are communicatively connected.

[0078] The input terminal of the power output device 22 is connected to the external power supply terminal 24, and the output terminal of the power output device 22 is connected to the heating module 30 corresponding to the driving component 20.

[0079] The power sampling circuit 23 includes a voltage sampling circuit 232 and a current sampling circuit 231. The second controller 21 is correspondingly provided with a first input terminal and a second input terminal. The voltage sampling circuit 232 is connected in parallel to the input terminal of the power output unit 22, and specifically acquires the input voltage in real time through its internal resistor voltage divider network. The current sampling circuit 231 is connected in series between the output terminal of the power output unit 22 and the heating module 30, and acquires the current flowing to the heating module 30 in real time through a sampling resistor or a Hall current sensor in the current sampling circuit 231. The second controller 21 obtains the instantaneous power value through the collected voltage and current, thereby providing feedback for local closed-loop control.

[0080] The output terminal of the power sampling circuit 23 is connected to the input terminal of the second controller 21 so as to send the electrical parameters to the second controller 21 through the power sampling circuit 23;

[0081] The output terminal of the second controller 21 is connected to the control terminal of the power output device 22, so as to form an isolation structure between the first controller 10 and the power output device 22 through the second controller 21, and is also used to adjust the output power of the power output device 22 in real time based on the electrical parameters.

[0082] The second controller 21 also includes a third input terminal. Multiple temperature sensors 25 are all connected to the third input terminal of the second controller 21 corresponding to their respective heating modules 30.

[0083] In one possible implementation, the air supply assembly includes at least one speed-regulating fan;

[0084] The first controller is communicatively connected to the speed-regulating fan, so that the first controller outputs a speed control signal to the speed-regulating fan based on the temperature signals collected by the first temperature sensors on the corresponding heating plate.

[0085] In a specific example, such as Figure 1 As shown, in this embodiment, the air supply assembly achieves active cooling through at least one speed-regulating fan 83. The first controller 10 is communicatively connected to the speed-regulating fan 83 via a PWM signal line or a bus interface. Meanwhile, as... Figure 5 and Figure 6 As shown, each of the heating modules 30 has multiple first temperature sensors 25 embedded on its surface or inside. These temperature sensors 25 capture the temperature changes of the heating module 30 under different power loads in real time and feed the temperature signals back to the first controller 10.

[0086] The first controller 10 incorporates an intelligent fan speed control algorithm, such as a PID control algorithm or a segmented step algorithm. During testing, the first controller 10 aggregates feedback data from each of the first temperature sensors 25 in real time, and performs weighted averaging or extreme value processing on multiple temperature signals from different locations on the same heating module 30 to obtain the characteristic temperature of the module. When the characteristic temperature rises and approaches the preset upper temperature limit, the first controller 10 automatically increases the PWM signal output to the speed-regulating fan 83 to increase the fan speed and increase the air intake.

[0087] Following the example above, the first controller can preset specific speed and temperature curves to simulate the fan speed control logic of the target AI server during actual operation. In this way, the test system can not only verify the performance of the heat sink under full-speed conditions, but also evaluate whether heat accumulation will occur in the heat sink under energy-saving mode or low-speed conditions.

[0088] Specifically, the speed-regulating fan can be, for example, a DC axial fan with a PWM speed control interface.

[0089] In one possible implementation, the heat dissipation testing system includes at least two heat dissipation testing devices, each including at least one first testing device and at least one second testing device; in the first testing device, a plurality of heat-generating modules are arranged in alignment along the same horizontal plane within the corresponding testing housing; in the second testing device, a plurality of heat-generating modules are arranged in a staggered and interleaved manner along the same horizontal plane within the testing housing.

[0090] In one specific example, the heat dissipation testing system includes multiple heat dissipation testing devices;

[0091] The plurality of heat dissipation testing devices include at least one first testing device and a plurality of second testing devices;

[0092] In any of the first test devices, multiple heating modules are arranged in alignment along the same horizontal plane within the corresponding test housing;

[0093] In the second testing device, multiple heating modules are arranged in a staggered manner along the same horizontal plane within the testing housing; wherein, in different second testing devices, the staggered offset between the heating modules is different.

[0094] In this embodiment, multiple heat dissipation testing devices form a static testing cluster, constructing a system-level experimental architecture that includes a baseline control group (mesh alignment) and a multi-level variable group (multiple second testing devices) with different staggered biases. Researchers can simultaneously launch all testing devices under the same external environment and in the same testing task. This eliminates unavoidable time-dimensional interference variables such as environmental temperature drift that are present when testing individual devices in batches, ensuring the rigor and high accuracy of thermodynamic comparison data between different topologies and shortening the verification cycle for AI server architecture selection.

[0095] In a specific example, in the relevant technology, the AI ​​server adopts an architecture of Hyperscale Graphics eXtension (HGX) or Open Compute Project Accelerator Module (OCP Accelerator Module, OAM). Under this architecture, the graphics cards use a tiled design, rather than the vertical plug-and-play structure of conventional servers. This architecture results in higher integration of the AI ​​server and places higher demands on the heatsink performance of the tested radiator 80. Therefore, heatsink performance testing is essential. In this embodiment, such as... Figure 3 As shown, multiple heat dissipation modules 30 are arranged in an array on the same horizontal plane within the test housing to simulate the tiled structure of a server graphics card, ensuring that multiple heat sinks 80 under test are at the same horizontal height after assembly. For example, the heat dissipation modules 30 can be directly mounted on the base plate within the test housing. The array-like arrangement of the multiple heat dissipation modules 30 includes side-by-side arrangement or staggered arrangement; for example... Figure 3 This displays a staggered and interlaced arrangement.

[0096] This embodiment uses a parallel arrangement mode to simulate the graphics card topology under the specifications of ultra-large-scale graphics extension architecture or open computing project accelerator module in physical space. This method realistically restores the compact topology of high-end AI servers such as HGX / OAM, accurately simulates real airflow obstruction, and ensures that the test data is not inflated.

[0097] On the other hand, when the heatsink under test is an air-cooled heatsink, the staggered arrangement can use physical gaps to avoid some of the front heat tail flow, thereby reducing thermal crosstalk. This arrangement not only provides an experimental platform for the development of next-generation server architectures with better heat dissipation, but also, if multiple heat dissipation test devices include both parallel and staggered arrangement architectures, researchers can intuitively obtain the thermal crosstalk effects of different staggered offsets and parallel arrangement architectures by comparing the test data of multiple heat dissipation test devices. Based on the degree of thermal crosstalk, they can optimize the arrangement of the graphics card in the server chassis or optimize the structure of the heatsink under test. For example, the liquid cooling plate can be streamlined to reduce wind resistance and improve the efficiency of hot air exhaust.

[0098] Following the example above, as a further optimization, such as Figure 8 As shown, to achieve rapid switching of the heating module 30 between the different arrangement structures and dynamic adjustment of the array interleaving degree, for at least one of the multiple heat dissipation testing devices 00, multiple slide rails 85 are arranged parallel to each other on the bottom plate of the test housing. Specifically, the multiple slide rails 85 are arranged parallel to each other along the extension direction perpendicular to the air inlet duct 84. At least one slider 86 is slidably mounted on each slide rail 85, such as... Figure 9 As shown, the heating module 30 is fixedly mounted on the corresponding slider 86. This is the initial state, with the heating modules 30 arranged side-by-side. Based on this structure, as... Figure 10 As shown, the testers independently moved the positions of each slider 86 on the corresponding transverse slide rail 85 to steplessly adjust the degree of overlap between the front and rear rows of heating modules 30, that is, to adjust the amount of overlap offset, thereby changing the projection overlap rate of the heating modules 30 on the airflow front surface.

[0099] In a specific example, as a further optimization, during the static testing of the heating modules (i.e., the heating modules are in a fixed position), a positioning locking component (not shown in the figure) is also provided on the side of the slider or slide rail. After the staggered arrangement of the heating modules is adjusted to the target test position, the slider is locked onto the slide rail to prevent positional displacement under strong airflow. The introduction of this lateral slide rail adjustment mechanism allows testers to efficiently complete thermal crosstalk comparison experiments from completely parallel to completely staggered under different shading rates from 0% to 100% on the same device without the need for cumbersome disassembly of the base plate, providing a foundation for optimizing the internal structure of the server.

[0100] It should be noted that, in this embodiment, although the physical arrangement of the graphics card in a mass-produced AI server is locked after the design is completed due to the highly integrated internal space filled with various functional components, and cannot be directly adjusted in the finished chassis, the heat dissipation testing system described in this application is not a simple reproduction of the finished product, but rather a research and development experimental platform specifically for thermal exploration and solution design. By dynamically adjusting the position of the graphics card within the heat dissipation testing device, the heat dissipation response of the graphics card under different spatial topologies is simulated. This design allows researchers to use this device to deduce and verify the optimal chip spacing and stagger ratio during the architecture definition phase of the next-generation server, thereby providing experimental data support for the subsequent refined layout planning of components within the actual chassis, airflow reservation, and customized improvements to the heat sink structure.

[0101] In a specific example, during dynamic testing of the heating module (i.e., the position of the heating module changes dynamically), the first controller is electrically connected to the drive motor (not shown in the figure), such as a stepper motor or servo motor, built into each of the slide rails. This enables automated adjustment of the position of each slider on each slide rail. Multiple sliders on the same slide rail can be connected to form a slider group, with each slider fixedly connected. The output of the drive motor is connected to one of the sliders, and by controlling the drive motor, the position of the multiple sliders on the corresponding slide rail is synchronously controlled. During the heat dissipation performance test of the heat sink under test, the first controller dynamically adjusts the lateral position of each slider on the slide rail based on the real-time temperature distribution signals fed back by multiple first temperature sensors on each heat sink under test, thereby changing the staggered offset between the heating modules in real time.

[0102] Specifically, the first controller has a preset optimization algorithm, and the first controller is used to perform the following operations during the heat dissipation performance test:

[0103] Along the direction of the preheating airflow, the average temperature difference between two adjacent rows of the heat sinks under test is monitored in real time. ;

[0104] The drive motors in each row are controlled to synchronously drive the corresponding sliders to slide laterally along the slide rail, so as to change the staggered offset between each heat sink under test;

[0105] During the sliding process, the average temperature difference is recorded in real time by the first controller. The relationship curve between the interleaving bias and the aforementioned amount;

[0106] The local minimum point in the relationship curve is obtained as the minimum judgment point of thermal crosstalk, and the position coordinates of each slider at this time are obtained.

[0107] The dynamic adjustment function in this embodiment not only greatly reduces downtime caused by manual position adjustments, but more importantly, it can accurately capture the disturbance patterns of minute spatial displacements on the global thermal flow. The optimal staggered offset obtained by the developers can be directly converted into the physical spacing design standard for graphics cards in AI server motherboard wiring schemes. This allows for optimization of graphics card physical spacing in the early stages of product design, reducing the potential risk of thermal runaway. Specifically, in this embodiment, when the heating modules on each slide rail are perfectly aligned longitudinally, they form a parallel arrangement; when the heating modules on adjacent slide rails have a lateral displacement difference, they form a staggered arrangement.

[0108] It should be further explained that the staggered offset amount mentioned in this embodiment is defined as the absolute distance between the geometric center lines of two adjacent rows of heating modules along the extension direction of the slide rail. This staggered offset amount is inversely proportional to the overlap rate of the projections of the front and rear rows of heating modules on the airflow front surface. Specifically, its boundary conditions are as follows: when the staggered offset amount is zero, the projections of the front and rear rows of heating modules completely overlap along the airflow direction, with a projection overlap rate of 100%. At this time, a completely parallel arrangement is formed, and the airflow front surface of the rear row of heating modules is completely within the high-heat wake region discharged by the front row of heating modules. When the staggered offset amount gradually increases until it is equal to or greater than the lateral physical width of a single heating module, the overlapping area of ​​the projections of the front and rear rows of heating modules along the airflow direction is zero, that is, the projection overlap rate is 0%. At this time, a completely staggered arrangement is formed, and the rear row of heating modules can completely avoid the obstruction of the front airflow, maximizing the capture of unheated ambient cold air.

[0109] In one specific example, the air-cooled heat sink under test is a heat pipe fin array heat sink; it specifically includes a heat-conducting base attached to the corresponding heat-generating module, multiple composite heat pipes, and multiple parallel heat dissipation fins fixed on the composite heat pipes. Multiple heat dissipation slits are formed between the multiple parallel heat dissipation fins, and the extension direction of the heat dissipation slits is parallel to the flow direction of the preheated airflow blown out by the air supply assembly.

[0110] In one possible implementation, at least one of the plurality of heat sinks is a liquid-cooled heat sink under test.

[0111] In one specific example, multiple heat sinks under test are configured as follows:

[0112] The air-cooled heat sink and the liquid-cooled heat sink under test can be selectively and detachably installed on the same heat-generating module to alternately perform air-cooled heat dissipation performance tests and liquid-cooled heat dissipation performance tests on the same heat-generating module.

[0113] This detachable configuration allows R&D personnel to directly replace the air-cooled radiator under test with a liquid-cooled plate in its original location when the heat dissipation performance fails to meet requirements. It enables seamless switching between air-cooling and liquid-cooling testing at the same physical workstation, reducing redundant asset investment required for separately building independent air-cooling and liquid-cooling testing facilities and improving testing efficiency.

[0114] It should be noted that, in this embodiment, the detachable installation specifically means that both the air-cooled and liquid-cooled heat sinks under test are fixed to the thermally conductive base of the heat-generating module using standardized fasteners or bolt assemblies. When switching test modes, the tester only needs to loosen the fasteners to remove the current heat sink without damage, and after applying a thermal interface material, such as thermal grease, quickly attach the other heat sink using the same set of fixing holes, thereby achieving rapid, in-situ substitution between air-cooled and liquid-cooled tests.

[0115] In a specific example, such as Figure 11 As shown, the heat dissipation test system also includes a cooling device 822 and a circulation pipeline 823;

[0116] The liquid-cooled heat sink 82 to be tested is a liquid-cooled plate 821. Each of the liquid-cooled plates 821 is connected to the refrigeration equipment 822 via the circulation pipe 823 to form a cooling circulation loop.

[0117] In a specific example, such as Figure 12 As shown, each of the liquid-cooled cold plates 821 and the refrigeration equipment 822 is provided with a flow meter 828 and a pressure sensor 827 on the circulation pipeline 823. The first controller 10 is connected to each pressure sensor 827 and each flow meter 828 respectively, so that the first controller 10 can monitor the leakage of the circulation pipeline 823 where the corresponding pressure sensor 827 is located based on the pressure feedback signal of the corresponding pressure sensor 827.

[0118] Each of the circulation pipes 823 is equipped with an inlet temperature sensor 825 and an outlet temperature sensor 826, which are used to collect the coolant temperature flowing into and out of the corresponding liquid-cooled radiator 82 to be tested, respectively.

[0119] The first controller 10 is connected to each of the inlet temperature sensors 825 and each of the outlet temperature sensors 826 respectively, so that the first controller 10 can calculate the real-time heat dissipation power of the liquid cooler 82 under test based on the temperature difference data collected by the inlet temperature sensors 825 and the outlet temperature sensors 826, combined with the flow data collected by the corresponding flow meter 828.

[0120] The first controller 10 is configured to perform the following operation for any of the circulation pipes 823 to determine the leakage status of the circulation pipe 823:

[0121] The pressure values ​​at the inlet and outlet ends, collected by the pressure sensor 827 installed on the circulation pipeline 823, and the inlet and outlet flow rates, collected by the flow meter 828 installed on the circulation pipeline 823, are obtained.

[0122] The pressure difference between the pressure value at the inlet end and the pressure value at the outlet end is calculated and used as the real-time pressure gradient of the circulation pipeline 823.

[0123] The flow difference between the inlet flow rate and the outlet flow rate is calculated and used as the real-time flow mismatch of the circulation pipeline 823.

[0124] The time change rate of the real-time pressure gradient of the circulation pipeline 823 is monitored. When the time change rate is greater than or equal to a preset change rate threshold and the real-time flow mismatch is greater than or equal to a preset flow deviation threshold, it is determined that the circulation pipeline 823 has leaked.

[0125] Following the above example, the heat dissipation test device 00 also includes a controllable circuit breaker (not shown in the figure) and a fuse (not shown in the figure) connected in series in the power input circuit of each of the drive components 20;

[0126] The first controller 10 is connected to each of the controllable circuit breakers.

[0127] The first controller 10 is configured to send a trip signal to the corresponding controllable circuit breaker to cut off the power supply circuit of the drive component 20 when it is determined that any loop 823 has a leak.

[0128] The fuse is connected in series between the controllable circuit breaker and the drive assembly 20, and is used to physically melt and break when the current in the power supply circuit exceeds a preset safety threshold.

[0129] In one possible implementation, the circulation pipeline includes an inlet sub-pipe and an outlet sub-pipe;

[0130] The liquid inlet end of each liquid-cooled plate is connected in parallel with the liquid inlet sub-pipe, and the liquid outlet end of each liquid-cooled plate is connected in parallel with the liquid outlet sub-pipe, so as to perform uniform flow cooling on the liquid-cooled plate through the refrigeration equipment.

[0131] In a specific example, such as Figure 11 As shown, to achieve independent and uniform temperature control of multiple heating modules, a multi-parallel cooling architecture is constructed between the refrigeration device 822 and each liquid-cooled plate 821. Specifically, the refrigeration device 822 is equipped with multiple liquid supply ports and multiple liquid return ports; the liquid inlet end of each liquid-cooled plate 821 is connected to each liquid supply port of the refrigeration device 822 via an independent liquid inlet sub-pipe; the liquid outlet end of each liquid-cooled plate 821 is connected to each liquid return port of the refrigeration device 822 via an outlet sub-pipe, thereby forming multiple independent and parallel cooling circulation loops. This avoids fluid interference between the liquid-cooled plates and ensures that each test node can obtain coolant with a consistent initial temperature and independently adjustable flow rate.

[0132] It should be noted that, due to the practical engineering situation where some refrigeration equipment 822 has only one liquid supply port and one liquid return port, or the number of interfaces is less than the number of liquid-cooled cold plates 821 under test, this embodiment also provides a fluid distribution scheme based on a water distributor structure. Specifically, the circulation pipeline also includes a water distributor and a water collector; the liquid supply port of the refrigeration equipment 822 is connected to the input end of the water distributor through a main pipe, and the water distributor has multiple output ends, each of which is connected to the liquid inlet end of each of the liquid-cooled cold plates 821 through parallel inlet sub-pipes. Correspondingly, the liquid outlet end of each of the liquid-cooled cold plates 821 is connected to multiple input ends of the water collector through parallel outlet sub-pipes, and the output end of the water collector is connected back to the liquid return port of the refrigeration equipment 822 through a return main pipe.

[0133] In one possible implementation, an adjustable wind deflector assembly is also provided inside the test housing;

[0134] The adjustable wind deflector is disposed between the air supply component and the air-cooled radiator under test;

[0135] The adjustable baffle assembly has pores with adjustable porosity. These pores are used to regulate the flow rate of the preheated airflow to simulate the increased wind resistance caused by dust accumulation inside the server and the degradation of the heat dissipation performance of the air-cooled radiator under test. In this embodiment, to realistically reproduce the complex flow field deterioration and degradation of the AI ​​server throughout its entire lifecycle, an adjustable baffle assembly is detachably inserted into the main air intake duct located between the air supply assembly and the air-cooled radiator under test within the test housing.

[0136] In one possible implementation, the adjustable wind deflector assembly includes:

[0137] A fixing plate, wherein a plurality of first ventilation holes are arrayed on the fixing plate;

[0138] A movable plate is fitted to one side of the fixed plate, and the movable plate has a plurality of second ventilation holes arranged in an array, with the first ventilation hole corresponding to the second ventilation hole one by one;

[0139] The first controller is communicatively connected to the drive mechanism, and the drive mechanism is drive-connected to the movable plate, so that the first controller controls the drive mechanism to drive the movable plate to slide in a direction parallel to the surface of the fixed plate.

[0140] The overlapping portion of the first ventilation hole and the second ventilation hole forms the pore. When the movable plate slides relative to the fixed plate, the porosity of the air passage pore is adjusted by changing the overlapping area of ​​the first ventilation hole and the second ventilation hole.

[0141] In one specific embodiment, the adjustable wind deflector assembly includes two baffle plates that are attached to each other and arranged parallel to each other. Each baffle plate is densely covered with an array of ventilation holes. One baffle plate is fixed inside the test housing, while the other baffle plate can slide laterally or longitudinally relative to the fixed block. Based on this structure, the tester or the first controller can steplessly change the overlap area of ​​the ventilation holes on the two baffle plates by driving the sliding displacement of the movable baffle plate, thereby achieving linear adjustment of the entire adjustable wind deflector assembly.

[0142] The introduction of the adjustable wind deflector assembly provides the system with the ability to simulate harsh operating conditions.

[0143] like Figure 13 and Figure 14 As shown, to achieve precise and dynamic control of the preheating airflow and resistance, an adjustable baffle assembly 87 is detachably inserted into the air inlet duct 84 in this embodiment. Specifically, the adjustable baffle assembly 87 includes a fixed plate 871 and a movable plate 872 mounted on the base 873. The fixed plate 871 and the movable plate 872 are parallel to each other and fitted together, and both have ventilation holes arrayed on their surfaces. Figure 14 and Figure 15 As shown, the bottom of the movable plate 872 is slidably embedded in the guide groove of the base 873. The first controller 10, through a communication-connected drive mechanism (such as a stepper motor or linear module, not shown in the figure), can drive the movable plate 872 to move laterally relative to the fixed plate 871 along the extension direction of the base 873, thereby steplessly changing the overlapping area of ​​the ventilation holes on the two plates, thus achieving mechanical adjustment of the air inlet porosity, i.e., the effective air permeable cross-sectional area. Porosity refers to the ratio of the total area of ​​the actual projected overlap of the ventilation holes on the movable plate 872 and the ventilation holes on the fixed plate 871 in the direction parallel to the airflow direction to the total physical area of ​​the overall windward surface of the adjustable windbreak assembly 87.

[0144] Following the example above, and relying on the structure described above, the first controller 10 in this embodiment is configured to execute an aging resistance simulation mode, which is used to accelerate the reproduction of the flow field deterioration process caused by dust accumulation after long-term operation of the server in a laboratory environment.

[0145] The specific operating mechanism is as follows: After the test is started, the first controller 10 sends a progressive drive command to the drive mechanism, controlling the movable plate 872 to slide slowly according to a preset time decay function or a preset step size. This sliding action continuously or progressively reduces the overlap area of ​​the ventilation holes between the fixed plate 871 and the movable plate 872 to simulate the working condition where the front panel dust screen or air duct is progressively blocked by particulate matter over time. Specifically, the preset time decay function can be an exponential decay curve.

[0146] Following the example above, the first controller 10 reads the temperature signals from multiple first temperature sensors 25 attached to the corresponding heat-generating module 30 in real time. When the average real-time temperature or the maximum real-time temperature value of the multiple first temperature sensors 25 is greater than or equal to a preset thermal failure temperature threshold, or when the rate of change of the real-time temperature changes abruptly and exceeds a preset temperature rise rate threshold, the first controller 10 captures and records the current time parameter. This time parameter is then calibrated as the thermal failure time node of the air-cooled radiator 81 under the extreme condition of progressively decreasing airflow. This node data can provide valuable quantitative indicators for data center operation and maintenance personnel to formulate server filter cleaning or maintenance cycles.

[0147] Following the example above, the first controller 10 is also configured to execute a physical obstruction simulation mode to simulate complex physical obstructions inside chassis of different shapes, such as wind shadow effects caused by hard drive cages and messy wiring.

[0148] The specific operating mechanism is as follows: During the test preparation phase, the R&D personnel input or the first controller 10 automatically obtains the preset wind resistance parameters corresponding to the target chassis shape or the internal wiring layout. Subsequently, the first controller 10 maps the wind resistance parameters into mechanical displacement and sends a positioning command to the drive mechanism, driving the movable plate 872 to slide to the corresponding target offset position and maintain static locking.

[0149] At this point, the specific porosity formed by the fixed plate 871 and the movable plate 872 aerodynamically matches the wind pressure drop height generated by the target chassis shape or the internal wiring layout. This design eliminates the need for separate mold making and prototyping for each new chassis. By simply calibrating and locking the adjustable baffle assembly 87 through the first controller 10, the internal flow field of various high-density chassis can be realistically reproduced on an open test bench, improving the versatility of the test platform and the accuracy of hardware evaluation.

[0150] In a specific example, the overall external structure of the test housing 50 is as follows: Figure 16 As shown. It should be noted that, Figure 16 The purpose is to demonstrate the complete physical enclosure configuration of this heat dissipation testing device in actual operation.

[0151] Specifically, the lateral width of the test housing 50 is adapted to standard server specifications, and the height can be flexibly designed as 4U, 8U, or other standard U-position heights depending on the internal component density. This embodiment does not limit the specific dimensions of the test housing. Furthermore, the shape of the test housing and the number of heat-generating modules can be flexibly adjusted according to the slot specifications of the AI ​​server under test. The interior of the test housing 50 is spatially divided into three interconnected but functionally decoupled areas:

[0152] It should be noted that the top, bottom plate and side walls of the test housing 50 together form a fully enclosed micro wind tunnel. Except for the air inlet reserved on the front face of the test housing 50 for the intake of ambient air and the air outlet reserved on the rear face for the exhaust of waste heat air after heat exchange, all other joints such as the cover plate and side walls are equipped with high-elasticity sealing strips or anti-static foam. This is intended to force all airflow to flow only along a single longitudinal axis from front to back, and to prevent disorderly dissipation of airflow from the side or top.

[0153] To further improve the absolute accuracy of thermal testing, a low thermal conductivity insulating layer, such as nano-aerogel or flame-retardant insulating cotton, is attached to the inner side of the test housing 50. This structure physically creates a quasi-insulating thermodynamic boundary, minimizing the disordered heat conduction loss of the high-heat array inside the system through the housing wall to the laboratory environment.

[0154] In one possible implementation, the heat dissipation testing device is multiple;

[0155] The first controllers of the multiple heat dissipation testing devices are interconnected to enable cascade testing of the multiple heat dissipation testing devices.

[0156] In a specific example, such as Figure 17 As shown, the AI ​​server heat dissipation testing system also includes a communication gateway 70;

[0157] Each of the first controllers 10 of the aforementioned heat dissipation testing devices 00 is communicatively connected to the communication gateway 70 to enable cascading testing of multiple heat dissipation testing devices 00. Specifically, the heat dissipation testing device 00 also includes a switch 80, and the first controllers 10 are communicatively connected to the communication gateway 70 via the switch 80.

[0158] In a specific example, when large-scale rack-level testing is involved, there can be multiple thermal testing devices 00. The first controller 10 includes at least one cascading interface, and the first controllers 10 of multiple thermal testing devices 00 are interconnected via the cascading interface to form a cascaded structure. In the cascaded state, multiple devices can share a single communication gateway 70 or an external terminal 60 such as a host computer for remote collaborative control. The cascading interface can be one of a CAN bus interface, an RS485 interface, and an Ethernet interface 14, preferably an Ethernet interface 14, as shown in the reference. Figure 3 The first controller 10 also includes a network communication circuit 13, and the first microcontroller 11 is connected to the Ethernet interface 14 through the network communication circuit 13.

[0159] In another example, the first controller includes at least two cascading interfaces, in which case multiple thermal testing devices can be cascaded without the need for a switch.

[0160] For example, in a specific test case, the display module is a touch display module. The tester sets a square wave power curve with a frequency of 10Hz and a peak power of 2500W on the touch display module's operating interface. After the first controller issues the command, the second controller corresponding to the central heating module takes over control. At the rising edge of the square wave, the Joule heating of the heating plate resistance increases instantaneously due to the current surge. The second controller senses the decrease in current fluctuation through the power sampling circuit and immediately increases the duty cycle of the switching transistor in the next PWM cycle.

[0161] It should be noted that although the first and second controller models described above are preferably microcontroller units (MCUs), this is not a limitation. The first or second controller can also be implemented using field programmable gate arrays (FPGAs), digital signal processors (DSPs), MCUs, or application-specific integrated circuits.

[0162] It should be noted that the "first controller," "second controller," and the unit that performs various logic controls and data processing involved in the embodiments of this application are all manifested in physical hardware form as specific electronic components or integrated circuit chips. For example, the controller can be a physical hardware entity composed of a central processing unit (CPU), a microprocessor (MCU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or any combination thereof.

[0163] Furthermore, the various control logics mentioned in the embodiments of this application, including but not limited to "heating power regulation logic", "thermal crosstalk bias dynamic optimization algorithm", "porosity time decay driving logic" and "pipeline leakage judgment logic", are not purely mathematical calculation methods or pure computer programs that exist independently of hardware. Instead, they are solidified or stored in a non-volatile computer-readable storage medium (such as ROM, EEPROM, Flash memory and other physical memory) that is communicatively connected to the controller in the form of firmware or software instructions.

[0164] In the description of this disclosure, it should be noted that the terms "upper," "lower," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, they can be fixed connections, detachable connections, or integral connections; they can be mechanical connections or electrical connections; they can be direct connections or indirect connections through an intermediate medium; they can be internal connections between two elements. For those skilled in the art, the specific meaning of the above terms in this disclosure can be understood according to the specific circumstances.

[0165] It should also be noted that, in the description of this disclosure, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0166] Obviously, the above embodiments of this disclosure are merely examples for clearly illustrating this disclosure, and are not intended to limit the implementation of this disclosure. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all implementation methods here. Any obvious variations or modifications derived from the technical solutions of this disclosure are still within the protection scope of this disclosure.

Claims

1. A server heat dissipation testing system, characterized in that, include: At least one heat dissipation testing device; The heat dissipation testing device includes a test housing, a first controller disposed within the test housing, multiple heat-generating modules, and a drive component connected to each heat-generating module. The first controller is connected to multiple drive components respectively, so as to control the drive group to adjust the heating power of the corresponding heating module through the first controller; Each of the aforementioned heat-generating modules is provided with a heat sink to be tested; at least one of the multiple heat sinks to be tested is an air-cooled heat sink to be tested. The heat dissipation testing device further includes an air supply component corresponding to the air-cooled radiator under test; the air outlet of the air supply component faces the windward side of the air-cooled radiator under test; at least one of the air supply components has an air inlet forming an air inlet duct, and multiple drive components are disposed in the air inlet duct. The first controller is connected to each of the air supply components to control each of the air supply components to draw ambient air from the air inlet duct, so that the ambient air flows through the air inlet duct and passes through multiple drive components to form a preheated airflow and is output to the air-cooled heat sink under test.

2. The heat dissipation testing system according to claim 1, characterized in that, The heating module includes: Heating plate; A heat-insulating base is disposed between the heating plate and the test housing; and... A heat-conducting base is disposed on the side of the heating plate away from the heat-insulating base.

3. The heat dissipation testing system according to claim 2, characterized in that, Multiple first temperature sensors are uniformly arranged between the heating plate and the heat-conducting base; Multiple of the first temperature sensors are connected to the input terminal of the first controller.

4. The heat dissipation testing system according to claim 3, characterized in that, The air supply assembly includes at least one speed-regulating fan; The first controller is communicatively connected to the speed-regulating fan, so that the first controller outputs a speed control signal to the speed-regulating fan based on the temperature signals collected by the first temperature sensors on the corresponding heating plate.

5. The heat dissipation testing system according to claim 1 or 2, characterized in that, The heat dissipation testing system includes at least two heat dissipation testing devices, each including at least one first testing device and at least one second testing device; the plurality of heat-generating modules in the first testing device are arranged in alignment along the same horizontal plane within the corresponding testing housing; the plurality of heat-generating modules in the second testing device are arranged in staggered and interleaved arrangement along the same horizontal plane within the corresponding testing housing.

6. The heat dissipation testing system according to claim 1, characterized in that, At least one of the multiple heat sinks is the liquid-cooled heat sink to be tested.

7. The heat dissipation testing system according to claim 6, characterized in that, The heat dissipation testing system also includes cooling equipment and circulation piping; The liquid-cooled heat sink under test is a liquid-cooled plate, and each of the liquid-cooled plates is connected to the refrigeration equipment through the circulation pipeline to form a cooling circulation loop.

8. The heat dissipation testing system according to claim 7, characterized in that, The circulation pipeline includes an inlet sub-pipe and an outlet sub-pipe; The liquid inlet end of each liquid-cooled plate is connected in parallel with the liquid inlet sub-pipe, and the liquid outlet end of each liquid-cooled plate is connected in parallel with the liquid outlet sub-pipe.

9. The heat dissipation testing system according to claim 1, characterized in that, An adjustable wind deflector assembly is also provided inside the test housing; The adjustable wind deflector is disposed between the air supply component and the air-cooled radiator under test; The adjustable windbreak assembly has pores with adjustable porosity, which are used to regulate the flow rate of the preheating airflow.

10. The heat dissipation testing system according to claim 9, characterized in that, The adjustable wind deflector assembly includes: A fixing plate, wherein a plurality of first ventilation holes are arrayed on the fixing plate; A movable plate is fitted to one side of the fixed plate, and the movable plate has a plurality of second ventilation holes arranged in an array, with the first ventilation hole corresponding to the second ventilation hole one by one; The first controller is communicatively connected to the drive mechanism, and the drive mechanism is drive-connected to the movable plate, so that the first controller controls the drive mechanism to drive the movable plate to slide in a direction parallel to the surface of the fixed plate. The overlapping portion of the first ventilation hole and the second ventilation hole forms the pore. When the movable plate slides relative to the fixed plate, the porosity of the air passage pore is adjusted by changing the overlapping area of ​​the first ventilation hole and the second ventilation hole.

Images