Heat dissipation structure of GPU (Graphics Processing Unit) server
By employing a combined cooling system of heat sinks and water-cooling pipes in the GPU server, along with intelligent control using temperature sensors and solenoid valves, the problem of insufficient heat dissipation during high-load operation of the GPU server is solved, achieving efficient and stable heat dissipation and extending the hardware lifespan.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Utility models(China)
- Current Assignee / Owner
- MEIZHOU BAY VOCATIONAL & TECH COLLEGE
- Filing Date
- 2025-05-16
- Publication Date
- 2026-05-05
AI Technical Summary
Insufficient heat dissipation in GPU servers during high-load operation can lead to performance degradation, shortened hardware lifespan, or even system crashes.
It adopts a dual cooling mechanism, including a combined cooling system of heat dissipation fins and water cooling pipes, combined with intelligent control of temperature sensors and solenoid valves to optimize heat dissipation efficiency.
It significantly improves the heat dissipation efficiency of GPU servers, ensuring stability and reliability under high load and extending hardware lifespan.
Smart Images

Figure CN224203658U_ABST
Abstract
Description
Technical Field
[0001] This utility model relates to the technical field of GPU servers, and more specifically, to a GPU server heat dissipation structure. Background Technology
[0002] GPU servers are fast, stable, and elastic computing services based on GPUs, applicable to various scenarios such as video encoding / decoding, deep learning, and scientific computing. They offer exceptional graphics processing and high-performance computing capabilities, providing extreme computing performance. Especially in the field of artificial intelligence (AI), GPU servers leverage their parallel computing capabilities to significantly accelerate the training and inference processes of AI models. The parallel architecture of GPUs (such as NVIDIA's CUDA cores) is particularly well-suited for handling matrix operations, accelerating backpropagation and gradient descent in neural networks. They also support high-concurrency inference tasks, such as real-time applications like facial recognition and voice interaction, including millisecond-level image recognition in autonomous driving. Training models like GPT and BERT requires the computing power of thousands of GPUs. GPUs can also accelerate policy iteration in virtual environments for applications such as robot control and game AI.
[0003] Therefore, the high performance of GPU servers places enormous demands on heat dissipation, especially under high-load AI training or inference scenarios, where GPU power consumption and heat generation are extremely high; for example, the TDP of the NVIDIA H100 GPU can reach over 700W. Insufficient heat dissipation can lead to performance throttling, shortened hardware lifespan, or even system crashes. Utility Model Content
[0004] This invention proposes a GPU server heat dissipation structure that optimizes heat dissipation efficiency through a cooling mechanism. When outside air enters the protective shell, the heat dissipation fins perform initial heat exchange and cooling on the high-temperature air. Combined with the surrounding water cooling pipes, a secondary deep cooling is performed, so that the low-temperature air, after being doubly cooled, enters the device body to participate in convection, significantly improving the overall heat dissipation efficiency to adapt to various scenarios such as high-load operation of the server.
[0005] Therefore, the technical solution adopted is as follows:
[0006] A GPU server heat dissipation structure includes an air outlet and an air inlet on both sides of the device body. An exhaust fan is installed at the air outlet, and a cooling mechanism is provided at the air outlet. The cooling mechanism includes a protective shell fixed to the device body. An air inlet is provided on the protective shell. A cooling component is fixed inside the protective shell. The cooling component includes a plurality of heat dissipation fins and water-cooling pipes passing through the heat dissipation fins. The inlet and outlet of the water-cooling pipes are respectively connected to an external circulation pipeline through a connector.
[0007] A further technical solution is that the air inlet is covered with a dustproof net, and the dustproof net is fixed to the protective shell.
[0008] A further technical solution is that a drying net is provided between the cooling mechanism and the equipment body, and the drying net covers the air inlet and is fixed to the equipment body.
[0009] A further technical solution is that a number of the heat dissipation fins are distributed at equal intervals inside the protective shell, and both ends of each heat dissipation fin penetrate the protective shell and extend to the outside of the protective shell.
[0010] A further technical solution is that the water-cooling pipes are arranged in a row of multiple pipes, and each pipe is serpentine and interspersed on several heat dissipation fins.
[0011] A further technical solution is that the external circulation pipeline includes a circulation pump, a liquid storage tank, and a radiator.
[0012] A further technical solution is that a temperature sensor is fixed on the protective shell, and both ends of the water-cooling pipe are connected to solenoid valves.
[0013] The working principle and beneficial effects of this application are as follows:
[0014] 1. The cooling mechanism optimizes the heat dissipation performance of the GPU server. When outside air enters the protective shell, the cooling components first cool the air, and then the low-temperature air enters the server body to participate in convection, which significantly improves the overall heat dissipation efficiency to adapt to various scenarios such as high-load operation of the server.
[0015] 2. The built-in heat dissipation fins first perform initial heat exchange and cooling of the high-temperature air, and then perform secondary deep cooling in conjunction with the surrounding water cooling pipes. Through this dual cooling method, the temperature of the outside air can be significantly reduced.
[0016] 3. The temperature sensor can detect the ambient temperature of the equipment and feed it back to the control system. The control system then controls the opening degree of the solenoid valve to adjust the flow rate of the coolant, adjusting the coolant flow rate to an economical mode that meets the current heat dissipation needs and reduces energy consumption. Attached Figure Description
[0017] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.
[0018] Figure 1 This is a schematic diagram of the overall structure of this application;
[0019] Figure 2 This is a schematic diagram of the overall structure of this application from another perspective;
[0020] Figure 3 This is a schematic diagram of the exploded structure of this application;
[0021] Figure 4 This is a schematic diagram of the exploded structure of the cooling mechanism in this application;
[0022] Figure 5 This is a schematic diagram of the cooling component described in this application.
[0023] In the diagram: 1. Equipment body; 2. Cooling mechanism; 21. Protective shell; 22. Air inlet; 23. Dustproof net; 24. Drying net; 25. Cooling component; 251. Heat dissipation fins; 252. Water cooling pipe; 253. Connector; 254. Solenoid valve; 26. Temperature sensor; 3. Air outlet; 4. Exhaust fan; 5. Air inlet. Detailed Implementation
[0024] The technical solutions of this utility model will be clearly and completely described below with reference to the embodiments of this utility model. Obviously, the described embodiments are only some embodiments of this utility model, and not all embodiments. Based on the embodiments of this utility model, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this utility model.
[0025] like Figures 1-5 As shown, a GPU server heat dissipation structure includes an air outlet 3 and an air inlet 5 on both sides of the device body 1. An exhaust fan 4 is installed at the air outlet 3. A cooling mechanism 2 is provided at the air outlet 3. The cooling mechanism 2 includes a protective shell 21 fixed to the device body 1. An air inlet 22 is provided on the protective shell 21. A cooling component 25 is fixed inside the protective shell 21. The cooling component 25 includes a plurality of heat dissipation fins 251 and water-cooling pipes 252 passing through the heat dissipation fins 251. The inlet and outlet of the water-cooling pipes 252 are respectively connected to an external circulation pipeline through a connector 253.
[0026] In this embodiment, the air outlets 3 and air inlets 5 on both sides of the device body 1 form a complete heat dissipation channel. Driven by the exhaust fan 4, a directional airflow is formed. External cold air enters the device body 1 through the air inlet 5, and after internal heat exchange, it carries heat and is discharged from the air outlet 3. This forced convection circulation achieves heat dissipation for the internal components of the server. The cooling mechanism 2, configured on one side of the air inlet 5, first performs preliminary heat exchange and cooling of the high-temperature air when outside air enters the protective shell 21. It then performs secondary deep cooling in conjunction with the surrounding water cooling pipes 252. This allows the low-temperature air, which has undergone double cooling, to enter the server body and participate in convection, significantly improving the overall heat dissipation efficiency. The efficient heat dissipation structure ensures that the GPU server operates within a suitable temperature range, guarantees the stable operation of artificial intelligence tasks, and improves the reliability and availability of the system.
[0027] The external circulation pipeline includes a circulation pump, a liquid storage tank, and a radiator. Two connectors 253 serve as the inlet and outlet of the coolant, respectively, and are connected to the external circulation pump, liquid storage tank, and radiator to introduce low-temperature coolant. After circulation, the coolant flows to the return pipeline, cools down, and then circulates again, forming a closed-loop cooling cycle.
[0028] like Figure 4 As shown, the air inlet 22 is covered with a dustproof mesh 23, which is fixed to the protective shell 21. During operation, the overall heat dissipation structure draws in air. Dust, fibers, hair, and other particles carried in the air may accumulate on the internal components. The dustproof mesh 23 physically filters and intercepts these particles, preventing them from entering the server. It is typically made of fine mesh or fiber material.
[0029] like Figure 4 As shown, a drying net 24 is also provided between the cooling mechanism 2 and the equipment body 1. The drying net 24 covers the air inlet 5 and is fixed to the equipment body 1. The drying net 24 is made of a highly breathable and moisture-absorbing material. While ensuring smooth air circulation, it can effectively block humid air from entering the interior of the equipment body 1, avoiding corrosion damage to electronic components caused by water vapor condensation.
[0030] like Figures 4-5 As shown, several heat dissipation fins 251 are evenly distributed inside the protective shell 21, and both ends of each heat dissipation fin 251 penetrate the protective shell 21 and extend to the outside of the protective shell 21 to increase the heat dissipation effect of the heat dissipation fins 251. Multiple water-cooling pipes 252 are arranged side-by-side and are meanderingly inserted into the heat dissipation fins 251, which increases the contact area between the water-cooling pipes 252 and the heat dissipation fins 251, improving the cooling effect. When the coolant circulates within the water-cooling pipes 252, on the one hand, when high-temperature air flows through the protective shell 21, it first undergoes initial heat exchange with the exposed heat dissipation fins 251, and then contacts the surface of the water-cooling pipes 252, achieving secondary cooling of the air through conduction via the pipe wall. On the other hand, the water-cooling pipes 252 simultaneously absorb the accumulated heat generated by the heat dissipation fins 251 themselves due to heat exchange, and the heat is quickly carried away from the cooling mechanism 2 by the forced convection of the coolant.
[0031] like Figures 4-5As shown, a temperature sensor 26 is fixed on the protective shell 21, and both ends of the water-cooling pipe 252 are connected to solenoid valves 254. The temperature sensor 26 is used to detect the ambient temperature of the device body 1. When the temperature sensor 26 detects that the ambient temperature exceeds a preset threshold, such as 35°C, the control system synchronously triggers all solenoid valves 254 to fully open, so that the coolant flows through the water-cooling pipe 252 at maximum flow rate, and starts the full-power heat dissipation mode. When the ambient temperature is lower than the threshold, the algorithm dynamically closes some pipeline solenoid valves 254, adjusting the coolant flow rate to an economical mode that adapts to the current heat dissipation requirements, effectively reducing the power consumption of the circulating pump.
[0032] The above are merely preferred embodiments of the present utility model and are not intended to limit the present utility model. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present utility model shall be included within the protection scope of the present utility model.
Claims
1. A GPU server heat dissipation structure, comprising an air outlet (3) and an air inlet (5) on both sides of a device body (1), wherein an exhaust fan (4) is installed at the air outlet (3), characterized in that: The air outlet (3) has a cooling mechanism (2), which includes a protective shell (21) fixed on the equipment body (1). The protective shell (21) has an air inlet (22). A cooling component (25) is fixed inside the protective shell (21). The cooling component (25) includes several heat dissipation fins (251) and a water cooling pipe (252) passing through the heat dissipation fins (251). The water inlet and outlet of the water cooling pipe (252) are connected to the external circulation pipeline through a connector (253). A drying net (24) is also provided between the cooling mechanism (2) and the equipment body (1). The drying net (24) covers the air inlet (5) and is fixed to the equipment body (1).
2. The GPU server heat dissipation structure according to claim 1, characterized in that, The air inlet (22) is covered with a dustproof net (23), and the dustproof net (23) is fixed to the protective shell (21).
3. The GPU server heat dissipation structure according to claim 1, characterized in that, Several heat dissipation fins (251) are evenly distributed inside the protective shell (21), and both ends of each heat dissipation fin (251) penetrate the protective shell (21) and extend to the outside of the protective shell (21).
4. The GPU server heat dissipation structure according to claim 1, characterized in that, The water-cooling pipes (252) are arranged in a row, and each of them is serpentine and interspersed on several heat dissipation fins (251).
5. The GPU server heat dissipation structure according to claim 1, characterized in that, The external circulation pipeline includes a circulation pump, a liquid storage tank, and a radiator.
6. The GPU server heat dissipation structure according to claim 1, characterized in that, A temperature sensor (26) is fixed on the protective shell (21), and both ends of the water cooling pipe (252) are connected to a solenoid valve (254).