Heat dissipation system and heat dissipation method of graphics processing unit (GPU)

By introducing a cooling system consisting of a main heatsink module, a through-type heat pipe, and a micro vapor chamber on the back of the graphics card, the heat transfer path is optimized, solving the problem of uneven heat dissipation in the dual-sided memory layout. This achieves efficient and uniform heat dissipation for the graphics card, improving the thermal stability of the system and the lifespan of the device.

CN121772091APending Publication Date: 2026-03-31BEIJING BOCO COMM TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-26
Publication Date
2026-03-31

AI Technical Summary

Technical Problem

High-end graphics cards with dual-sided memory layouts suffer from uneven heat dissipation on both sides, resulting in extremely low heat dissipation efficiency for the memory on the back of the PCB. This leads to excessive temperature differences, affecting the consistency of resistors and capacitors and causing signal distortion, thus posing a risk to system thermal stability.

Method used

The heat dissipation system employs a main heat dissipation module, a through-type heat pipe, a rear micro heat spreader, and a backplate. The heat from the memory on the back of the PCB is collected by the backplate and transferred to the rear micro heat spreader. Then, it is transferred to the main heat dissipation module through the through-type heat pipe, and finally, the main heat dissipation module performs the heat dissipation operation. This optimizes the heat transfer path to reduce temperature differences.

Benefits of technology

It effectively reduces the temperature difference between the front and back of the memory on the PCB, ensuring that the temperature is within a safe range, improving the uniformity and stability of heat dissipation, and avoiding frequency reduction or hardware damage caused by uneven heat dissipation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121772091A_ABST
    Figure CN121772091A_ABST
Patent Text Reader

Abstract

The invention discloses a heat dissipation system and a heat dissipation method of a graphics processing unit (GPU). In the heat dissipation system, the heat of the video memory on the back of the PCB is collected through the back plate, and after the heat is transferred to the back micro vapor chamber, the heat is transferred to the penetrating type heat pipe through the back micro vapor chamber, then the heat is transferred to the main heat dissipation module through the penetrating type heat pipe, and finally the heat is dissipated through the main heat dissipation module. In the process, the set heat transfer path is utilized to transfer the heat of the back video memory to the front for heat dissipation operation, so that the temperature difference between the front video memory and the back video memory of the PCB can be reduced to ensure that the temperatures of the front video memory and the back video memory of the PCB are within a safe range.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the technical field of heat dissipation for electronic devices, and in particular to a heat dissipation system and method for a graphics processing unit (GPU). Background Technology

[0002] In the fields of high-performance computing, AI training, and graphics rendering, the memory capacity of graphics processing units (GPUs) has become a key bottleneck restricting the processing of large-scale data and complex models. To achieve greater memory capacity on a single graphics card, the most direct solution in the industry is to increase the number of memory chips. Given a fixed PCB area, changing the traditional single-sided layout of the memory chips to a double-sided layout is currently the most effective technical approach.

[0003] Currently, high-end graphics cards with dual-sided memory layouts rely on large cast copper bases or heat spreaders to cover the GPU core and front memory for heat dissipation. This results in extremely low heat dissipation efficiency for the memory on the back of the PCB, leading to uneven heat dissipation on both sides in practical applications. Summary of the Invention

[0004] This application provides a heat dissipation system and method for a graphics processing unit (GPU), which can provide balanced and efficient heat dissipation for the memory on both sides of the PCB, ensuring that the temperature of the memory on both sides of the PCB is within a safe range.

[0005] In a first aspect, this application provides a heat dissipation system for a graphics processing unit (GPU), characterized in that the heat dissipation system comprises: a main heat dissipation module, a through-type heat pipe, a rear micro heat dissipation plate, and a backplate; the heat dissipation system is applied to a printed circuit board (PCB).

[0006] The main heat dissipation module is located on the front of the PCB in the area of ​​the video memory, and is used for heat dissipation.

[0007] A through-type heat pipe is arranged along the edge of the PCB, with one end thermally coupled to the miniature heat sink on the back and the other end thermally coupled to the main heat dissipation module.

[0008] A miniature heat spreader is placed between the backplate and the PCB to evenly distribute the heat from the back memory.

[0009] The backplate, located on the back of the PCB in the area of ​​the memory chip, is thermally coupled to the memory chip on the back of the PCB via a thermal pad. It is used to collect the heat from the memory chip on the back of the PCB and transfer it to the miniature heat spreader on the back.

[0010] The heat transfer path for the memory on the back of the PCB is, in order, the backplate, the surrounding micro heat spreader, the through-type heat pipe, and the main heat dissipation module.

[0011] Optionally, the main heat dissipation module includes:

[0012] A heat spreader is used to cover the GPU core and the memory on the front of the PCB.

[0013] The internal heat pipe is thermally coupled at one end to the front heat spreader and at the other end to the thick aluminum fin array.

[0014] A thick aluminum fin array is used to receive heat from the front heat spreader and then dissipate it.

[0015] Optionally, the front heat spreader is a 4mm thick copper heat spreader, and the interior of the copper heat spreader includes a 0.3mm thick capillary structure of sintered copper powder.

[0016] The thick aluminum fin array consists of an array of thick aluminum fins with a thickness of 0.2 mm and a spacing of 1.5 mm.

[0017] The internal heat pipe consists of three heat pipes with a diameter of 6mm.

[0018] Optionally, the front heat spreader applies 60psi contact pressure to the GPU core and 30psi contact pressure to the memory on the front of the PCB via a mounting structure.

[0019] Optionally, the backplate adopts an aluminum alloy main structure, and a copper contact part is provided in the area that contacts the back memory to reduce the contact thermal resistance.

[0020] Optionally, airflow channels can be provided at the edge of the area of ​​the rear memory to guide airflow through the rear heat dissipation area.

[0021] Optionally, the cooling system also includes a turbine blower;

[0022] A turbine blower is located on the front side of the PCB in the area of ​​the memory chip, and is used to blow air to the thick aluminum fin array for heat dissipation. The airflow is parallel to the direction of the PCB.

[0023] Optionally, the turbine blower is a double ball bearing turbine blower with a speed range of 1500-6200 RPM and an air volume of 45 CFM.

[0024] Secondly, this application provides a heat dissipation method for a graphics processing unit (GPU), applied to the heat dissipation system described in the first aspect, the method comprising:

[0025] Heat from the memory on the back of the PCB is collected through the backplate and transferred to the miniature heat spreader on the back.

[0026] Heat is transferred to the through-type heat pipe via a miniature heat spreader on the back.

[0027] Heat is transferred to the main heat dissipation module via a through-type heat pipe;

[0028] The main heat dissipation module performs heat dissipation operations based on the amount of heat generated.

[0029] Optionally, the main heat dissipation module performs heat dissipation operations based on the amount of heat, including:

[0030] Heat is dissipated by blowing air onto the thick aluminum fin array using a turbine blower.

[0031] Therefore, this application has the following beneficial effects:

[0032] This application provides a heat dissipation method for a graphics processing unit (GPU). First, heat from the memory chips on the back of the PCB is collected via a backplate and transferred to a micro heat spreader on the back. Then, the heat is transferred through the micro heat spreader to a through-type heat pipe, and finally to the main cooling module. The main cooling module then dissipates the heat. In this process, by utilizing a pre-defined heat transfer path to transfer heat from the back of the PCB to the front for cooling, the temperature difference between the front and back of the memory chips on the PCB can be reduced, ensuring that the temperatures of the memory chips on both sides of the PCB remain within a safe range. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0034] Figure 1 This is a schematic diagram of the structure of a GPU heat dissipation system according to an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a main heat dissipation module in an embodiment of this application; Figure 3 This is a flowchart illustrating a GPU heat dissipation method according to an embodiment of this application. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0036] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant regions.

[0037] First, in order to better understand the technical solutions of the embodiments of this application, the technical terms that appear will be explained and defined.

[0038] (1) GPU (Graphics Processing Unit): Graphics processing unit used for high-speed parallel computing.

[0039] (2) PCB (Printed Circuit Board): Printed circuit board.

[0040] (3) Memory array: A high-density storage unit on the GPU used to store graphics data.

[0041] (4) Heat dissipation system: including heat sinks, heat pipes, fans and other devices used to reduce the operating temperature of electronic components.

[0042] (5) Heat spreader: A vacuum cavity device that utilizes the phase change (evaporation and condensation) of the internal working fluid for efficient two-dimensional heat conduction.

[0043] Currently, high-end graphics cards with dual-sided memory layouts typically employ the following cooling solutions:

[0044] The front side uses a large cast copper base or heat spreader to cover the GPU core and memory, and heat pipes and fans to achieve efficient heat dissipation. However, in order to make the product more universal and meet industry constraints, the heat dissipation manufacturers rely on a thicker thermal pad and PCB backplate for passive heat dissipation of the memory on the back. This makes the heat dissipation efficiency of the memory on the back much lower than that of the memory on the front, resulting in a temperature difference between the memory on the front and the back. When the temperature difference reaches more than 10°C, the resistance and capacitance of the memory chips on the two sides will be different, which may cause clock phase inconsistency and increased signal distortion, thus increasing the bit error rate and posing a risk to the thermal stability of the system.

[0045] Among these, the mandatory industry constraints include the rigid constraints of industry size specifications, economic considerations of materials and manufacturing, and the risks of mechanical stress and long-term reliability. In response, the current heat dissipation solutions do not cover the back of the memory with a large heat dissipation structure. This is essentially the result of heat dissipation manufacturers cooperating with GPU industry specifications under the triple constraints of space, cost, and weight.

[0046] This application provides a heat dissipation system for a graphics processing unit (GPU). The system includes a main heat dissipation module, a through-type heat pipe, a rear micro-vapor chamber, and a backplate. First, the backplate collects heat from the memory chips on the back of the PCB and transfers it to the rear micro-vapor chamber. Then, the rear micro-vapor chamber transfers the heat to the through-type heat pipe, which in turn transfers it to the main heat dissipation module. Finally, the main heat dissipation module dissipates the heat. Therefore, the system provided in this application utilizes a pre-defined heat transfer path to transfer heat from the back of the memory chips to the front for cooling, reducing the temperature difference between the front and back of the PCB and ensuring that the memory chip temperatures on both sides of the PCB remain within a safe range.

[0047] To facilitate understanding of the specific implementation of a heat dissipation system for a graphics processing unit (GPU) provided in this application embodiment, the following description will be provided in conjunction with the accompanying drawings.

[0048] Figure 1 This paper shows a schematic diagram of a GPU heat dissipation system according to an embodiment of this application. See also... Figure 1 The system 100 specifically includes: a main heat dissipation module 101, a through-type heat pipe 102, a rear micro heat dissipation plate 103, and a backplate 104. The heat dissipation system 101 is applied to the PCB.

[0049] The main heat dissipation module 101 is located in the area of ​​the video memory on the front of the PCB and is used for heat dissipation.

[0050] A through-type heat pipe 102 is arranged along the edge of the PCB, with one end thermally coupled to the back micro heat dissipation plate 103 and the other end thermally coupled to the main heat dissipation module 101.

[0051] A miniature heat spreader 103 is disposed between the backplate 104 and the PCB to evenly distribute the heat of the back memory.

[0052] The backplate 104 is located on the back of the PCB in the area of ​​the memory chip. It is thermally coupled to the memory chip on the back of the PCB through a thermal pad and is used to collect the heat from the memory chip on the back of the PCB and transfer it to the back micro heat spreader 102.

[0053] The heat transfer paths for the memory on the back of the PCB are, in order: backplate 104, back micro heat spreader 103, through heat pipe 102, and main heat dissipation module 101.

[0054] The back micro heat spreader 103 is designed to evenly distribute the heat from multiple memory chips, eliminate local hot spots, and thus evenly disperse the heat from the back memory chips.

[0055] It should be noted that the miniature heat spreader 103 on the back of the card in this embodiment can be replaced by a thick pure copper base plate combined with multiple heat pipes. Although the heat dissipation effect is slightly inferior, the cost may be lower. Furthermore, the miniature heat spreader 103 on the back can also be integrated with the metal backplate of the graphics card, making the entire backplate 104 part of the heatsink, connected to the PBC back memory and main heatsink module 101 via thermal pads.

[0056] The backplate 104 can be made of aluminum alloy with copper contact parts in the area where it contacts the back memory chips to reduce thermal resistance. This aluminum alloy structure achieves a balance between lightweight and strength, while the copper contact parts in the memory chip contact area utilize copper's high thermal conductivity to significantly reduce interface thermal resistance, allowing heat to be efficiently conducted to the backplate. This effectively controls the temperature difference between the front and back memory chips to within 10°C, preventing localized overheating that could lead to throttling or hardware damage, and improving overall heat dissipation stability.

[0057] When applying the heat dissipation system 100 in the embodiments of this application, the PCB needs to be optimized in the following aspects to accommodate heat pipe heat transfer, back memory heat dissipation, and airflow guidance:

[0058] (1) Design of heat pipe path

[0059] In order to ensure that the through-type heat pipe 102 can efficiently conduct heat and be tightly connected to the back micro heat sink 103 of the back memory and the main heat dissipation module 101 of the front memory, an appropriate space must be left in the edge area of ​​the PCB for installing the through-type heat pipe 102.

[0060] Specifically, the through-type heat pipe 102 can be composed of two flat copper heat pipes. This flat design not only reduces the space occupied by the heat pipes at the edge of the PCB, but also helps to improve the fit between the heat pipes and the edge of the PCB, thereby reducing contact thermal resistance and improving the lateral heat conduction efficiency. In actual layout, these two flat heat pipes can be arranged symmetrically or asymmetrically along the edge area of ​​the PCB. The specific location needs to be comprehensively considered based on the component distribution on the PCB, heat dissipation requirements, and overall structural strength.

[0061] Through this design, the through-type heat pipe 102 can effectively transfer the heat generated in the back memory area to the main heat dissipation module 101 of the front memory, achieving efficient collaborative heat dissipation of the dual memory sides, thereby improving the overall heat dissipation performance and stability of the GPU.

[0062] (2) Design of the thermal pad and the contact surface between the memory chip and the back of the device

[0063] To ensure effective contact between the back-side memory and the back-side micro heat spreader 103 and achieve optimal heat conduction performance, this embodiment of the application sets a graphene thermal pad between the memory and the backplate 104 to effectively fill the microscopic gap between them and eliminate the air layer. At the same time, adaptive bonding is achieved through surface microstructure design, which significantly reduces the interface thermal resistance and allows heat to be efficiently conducted to the back-side micro heat spreader 103. This ensures that the back-side memory and the micro heat spreader form a tight and reliable heat conduction interface, greatly improves heat dissipation efficiency, controls the temperature difference at the contact surface, and enhances the thermal stability of the dual-sided memory.

[0064] (3) Design of the rear air duct

[0065] To improve the heat dissipation efficiency of the back memory area, this embodiment of the application designs a precision airflow channel structure around and at the edge of the corresponding memory chip on the PCB backplate, so that airflow can effectively pass through the back heat dissipation area, thereby achieving a high-efficiency and stable heat dissipation effect.

[0066] As can be seen, the embodiments of this application utilize the GPU cooling system through the coordinated design of the main cooling module, through-type heat pipe, back micro heat sink and backplate, and the set heat transfer path to transfer the heat of the back memory to the front for heat dissipation. This can reduce the temperature difference between the front and back memory of the PCB and ensure that the memory temperature on both sides of the PCB is within a safe range.

[0067] The following is combined Figure 2 This application presents a schematic diagram of a main heat dissipation module, further explaining and optimizing the GPU's heat dissipation system. Specifically:

[0068] Figure 2 This paper shows a schematic diagram of the structure of a main heat dissipation module according to an embodiment of this application. See also: Figure 2 The main heat dissipation module 200 may specifically include: a front heat spreader 201, an internal heat pipe 202, and a thick aluminum fin array 203, wherein,

[0069] The front heat spreader 201 is used to cover the GPU core and the front memory of the PCB.

[0070] The internal heat pipe 202 is thermally coupled at one end to the front heat spreader and at the other end to the thick aluminum fin array 203.

[0071] The thick aluminum fin array 203 is used to receive heat from the front heat spreader and then dissipate it.

[0072] The front heat spreader can be a 4mm thick copper heat spreader, and the copper heat spreader includes a 0.3mm thick capillary structure of sintered copper powder inside.

[0073] The front heat spreader 201 also contacts the GPU core and the front memory of the PCB through a graphene thermal pad. It should be noted that the graphene thermal pad in this embodiment can be replaced by a high-performance silicone grease or a phase change thermal pad, and the GPU core in this embodiment uses liquid metal thermal paste, which is almost seamless after being tightly attached.

[0074] In this embodiment, the front heat spreader 201 uses a copper heat spreader due to its excellent thermal conductivity and stable chemical properties. Furthermore, the 4mm thickness combined with a 0.3mm sintered copper powder capillary structure increases the contact area and optimizes capillary suction, resulting in faster thermal response and more uniform temperature distribution (the temperature difference between any two points can be controlled within 10℃), making it particularly suitable for cooling dense heat sources in high-power GPUs. While other materials (such as aluminum) are lower in cost and lighter, their thermal conductivity (approximately 200 W / m·K) and corrosion resistance are poor, potentially affecting long-term reliability.

[0075] The front heat spreader 201 applies a 60psi contact pressure to the GPU core and a 30psi contact pressure to the front memory of the PCB through a fastening structure, so that the heat spreader and the chip / memory form a low thermal resistance contact, thereby improving the overall heat dissipation efficiency.

[0076] The thick aluminum fin array 203 in this embodiment may include a thick aluminum fin array with a thickness of 0.2 mm and a spacing of 1.5 mm. The 0.2 mm thin fins maximize the heat dissipation surface area in a limited space, and the 1.5 mm spacing ensures smooth airflow and avoids hot air stagnation. In addition, aluminum is lighter than copper and has a lower processing cost, making it suitable for mass production.

[0077] In this embodiment, the internal heat pipe 202 consists of three 6mm diameter heat pipes. The 6mm diameter heat pipes have a larger internal working fluid circulation space, and the heat transfer capacity of a single heat pipe is better than that of a thinner pipe. They can quickly transfer the heat from the GPU core to the fin array. Moreover, compared to a single thick heat pipe, the three 6mm heat pipes ensure heat dissipation efficiency while also taking into account manufacturing costs and weight control.

[0078] It should be noted that the heat dissipation system in the embodiments of this application may further include:

[0079] The turbine blower 204 is located in the memory area on the front of the PCB and is used to blow air to the thick aluminum fin array 203 for heat dissipation, wherein the airflow is parallel to the direction of the PCB.

[0080] The turbine blower 204 in this embodiment uses a dual-ball bearing turbine blower with a speed range of 1500-6200 RPM, an airflow of 45 CFM, and supports 4-wire PWM control. The 45 CFM airflow can quickly penetrate the 1.5mm pitch thick aluminum fin array, expelling hot air parallel to the PCB and preventing airflow short-circuiting. Furthermore, the PWM speed adjustment capability of 1500-6200 RPM can be intelligently adjusted according to the load, with noise levels below 30dB under low load and ensuring the memory temperature remains stable below 85℃ under high load. The dual-ball bearing design supports 24 / 7 continuous operation with a lifespan of 50,000 hours, making it suitable for harsh environments such as servers. This design improves overall heat dissipation efficiency by optimizing the airflow and heatsink coordination, while also balancing noise and energy consumption.

[0081] The main heat dissipation module 200 in this embodiment adopts an integrated design of heat sink and turbine blower, eliminating the need for external auxiliary fans. This makes it perfectly compatible with OEM complete machines, workstations and data center server scenarios that have strict requirements for size and airflow.

[0082] In OEM system applications, the integrated structure of the radiator and turbine blower avoids conflicts between the external fan and the chassis airflow, and reduces dust accumulation through positive pressure ventilation (the intake air volume is slightly greater than the exhaust air volume), thereby improving long-term reliability.

[0083] In workstation and data center server scenarios, the PWM speed control of the turbine blower supports dynamic load adjustment, maintaining stable airflow in high-density server deployments and preventing local overheating that could lead to frequency reduction.

[0084] As can be seen, this application provides a heat dissipation system for dual-sided GPU memory, achieving efficient thermal management through multi-module collaborative design, making it particularly suitable for harsh environments such as high-power graphics cards and data center servers. The system consists of four core components: a front main heat dissipation module, a rear micro vapor chamber, a through-type heat pipe heat transfer structure, and a backplate. It employs a collaborative airflow control scheme, utilizing a turbine blower to blow air parallel to the PCB direction. The airflow is guided by rear airflow channels, forming a laminar flow that penetrates the fin array, and heat is exhausted through the chassis airflow duct. Through three-dimensional collaborative design, the temperature difference between the front and back memory sides is significantly reduced (controlled within 10°C), avoiding frequency throttling or hardware damage caused by uneven heat dissipation.

[0085] To make the methods provided in the embodiments of this application clearer and easier to understand, the following is combined with... Figure 3 This method is applied to Figure 1 A specific example of the GPU cooling system 100 shown is illustrated.

[0086] To better understand the assembly and use of the heat dissipation system, this application provides a dual-sided graphics card product with specific modules as shown in Table 1.

[0087]

[0088] First, a 2.4mm thick FR408HR PCB is selected and subjected to immersion gold surface treatment. The PCB is pre-installed with a through-type heat pipe as required. Next, the main heatsink is installed: liquid metal thermal paste is precisely applied to the GPU core die (bare die, chip), and 1.5mm thick graphene thermal pads are attached to both sides of the memory chips. The aluminum alloy backplate is then attached to the thermal pads. A 2.5mm high vapor chamber for the back memory chips is attached to the backplate. One end of the through-type heat pipe is thermally coupled to the back micro vapor chamber, and the other end is thermally coupled to the front main heatsink. The main heatsink is aligned and installed on the front of the PCB, thermally coupled to the through-type heat pipe. Pressure is applied to the main heatsink using the mounting brackets, ensuring the GPU core reaches 60psi and the memory reaches 30psi, while simultaneously pressing the back vapor chamber and the main heatsink baseplate firmly. Finally, assembly and software flashing are performed: a dual ball bearing turbine blower (4-wire PWM interface) is installed.

[0089] Subsequent thermal and power consumption tests will verify whether the memory hotspot temperature is consistently below 95°C under full load, whether the output voltage ripple is within ±10mV, whether the fan speed is smoothly adjusted according to the composite temperature, and whether the overall design works normally within the target chassis airflow.

[0090] like Figure 3 As shown, this embodiment may include the following steps:

[0091] S301: Collects heat from the memory on the back of the PCB through the backplate and transfers the heat to the miniature heat spreader on the back.

[0092] To efficiently dissipate heat from the memory on the back of the PCB, the first step is to use a backplate to collect the heat from the memory and transfer it to a miniature heat spreader on the back. The heat is then transferred to a through-type heat pipe via the miniature heat spreader, and finally to the main heat dissipation module. The main heat dissipation module then performs heat dissipation operations based on the amount of heat. Therefore, this embodiment requires the backplate to collect the heat from the memory on the back of the PCB, providing a prerequisite for efficient heat dissipation of the memory on the back.

[0093] In this embodiment, the backplate (such as an aluminum alloy-copper composite structure) can quickly absorb the heat from the memory and distribute the heat evenly through the phase change cycle of the micro heat spreader on the back, avoiding local hot spots. At the same time, the micro heat spreader can effectively balance the temperature of the memory on the front and back, keeping the temperature difference within 10°C, thus improving the long-term stability of the graphics card.

[0094] S302: Heat is transferred to the through-type heat pipe via a miniature heat spreader on the back.

[0095] In this embodiment, a flat design is used to adapt to a compact layout, and heat is directed to the main heat dissipation module through heat pipes to achieve three-dimensional collaborative heat dissipation.

[0096] S303: Heat is transferred to the main heat dissipation module through a through-type heat pipe.

[0097] In this embodiment, the through-type heat pipe utilizes the working fluid phase change (evaporation-condensation) principle, and its heat transfer efficiency far exceeds that of solid metals, enabling it to quickly conduct heat from the heat source (such as the GPU core) over long distances to the main heat dissipation module.

[0098] S304: Heat dissipation is performed by the main heat dissipation module based on the amount of heat generated.

[0099] In this embodiment, the main heat dissipation module and the fan module can perform heat dissipation operations based on the heat. For example, the turbine blower can blow air onto the thick aluminum fin array to dissipate heat, which can efficiently remove the heat from the fins, ensuring that the heat dissipation system operates stably under high load and avoiding frequency reduction due to overheating.

[0100] This embodiment provides a method for heat dissipation of a graphics processing unit (GPU). First, a backplate collects heat from the memory chips on the back of the PCB and transfers it to a micro heat spreader on the back. The heat is then transferred through the micro heat spreader to a through-type heat pipe, which in turn transfers the heat to the main cooling module. Finally, the main cooling module dissipates the heat based on its volume. During this process, the backplate collects heat from the memory chips and efficiently conducts it to the main cooling module through the through-type heat pipe, effectively reducing the temperature difference between the front and back of the memory chips and improving heat dissipation uniformity. By optimizing the heat dissipation path and enhancing heat conduction, localized overheating is avoided, allowing the GPU to operate stably under high loads. This reduces the risk of performance degradation and hardware damage caused by high temperatures, thereby extending the device's lifespan. Furthermore, the integrated turbine blower design eliminates the need for external auxiliary fans, making it perfectly compatible with OEM systems, workstations, and data center servers that have strict requirements for size and airflow.

[0101] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the methods of the above embodiments can be implemented by means of software plus a general-purpose hardware platform. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as a read-only memory (ROM) / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, a server, or a network communication device such as a router) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0102] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are merely illustrative. Modules described as separate components may or may not be physically separate. Components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the objectives of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0103] The above description is merely an exemplary implementation of this application and is not intended to limit the scope of protection of this application.

Claims

1. A heat dissipation system of a graphics processing unit (GPU), comprising: The heat dissipation system comprises a main heat dissipation module, a through heat pipe, a back micro heat plate and a back plate; the heat dissipation system is applied to a printed circuit board (PCB); The main heat dissipation module is arranged in the area of the front display memory of the PCB and is used for heat dissipation operation; The through heat pipe is arranged along the edge of the PCB, one end of which is thermally coupled with the back micro heat plate, and the other end is thermally coupled with the main heat dissipation module; The back micro heat plate is arranged between the back plate and the PCB and is used for uniformly dispersing the heat of the back display memory; The back plate is arranged in the area of the back display memory of the PCB, is thermally coupled with the back display memory of the PCB through a heat conduction pad and is used for collecting the heat of the back display memory of the PCB and transferring the heat to the back micro heat plate; The heat transfer path of the back display memory of the PCB is the back plate, the back micro heat plate, the through heat pipe and the main heat dissipation module in sequence.

2. The heat dissipation system of claim 1, wherein, The main heat dissipation module comprises: A front heat plate for covering the core of the GPU and the front display memory of the PCB; An internal heat pipe, one end of which is thermally coupled with the front heat plate and the other end of which is thermally coupled with a thick aluminum fin array; The thick aluminum fin array is used for receiving the heat conduction of the front heat plate and dissipating heat.

3. The heat dissipation system of claim 2, wherein, The front heat plate is a 4mm-thick copper heat plate, and the copper heat plate internally comprises a 0.3mm-thick sintered copper powder capillary structure; The thick aluminum fin array comprises a thick aluminum fin array with a thickness of 0.2mm and a pitch of 1.5mm; The internal heat pipe is a 3mm-diameter heat pipe.

4. The heat dissipation system of claim 2, wherein, The front heat plate applies a 60psi contact pressure to the core of the GPU and a 30psi contact pressure to the front display memory of the PCB through a buckle structure.

5. The heat dissipation system of claim 2, wherein, A turbo blower is further included; The turbo blower is arranged in the area of the front display memory of the PCB and is used for blowing air to the thick aluminum fin array, and the air flow of the blowing is parallel to the direction of the PCB.

6. The heat dissipation system of claim 5, wherein, The turbo blower adopts a double-ball-bearing turbo blower with a rotating speed range of 1500-6200 RPM and an air volume of 45 CFM.

7. The heat dissipation system of claim 1, wherein, The back plate adopts an aluminum alloy main structure, and a copper contact part is arranged in the area in contact with the back display memory to reduce the contact thermal resistance.

8. The heat dissipation system of claim 1, wherein, A flow guide groove is arranged at the edge of the area of the back display memory and is used for guiding the air flow to pass through the back heat dissipation area.

9. A method for dissipating heat from a graphics processing unit (GPU), the method comprising: The method is applied to the heat dissipation system of claims 1-8, and the method comprises: Collecting the heat of the back display memory of the PCB through the back plate and transferring the heat to the back micro heat plate; Transferring the heat to the through heat pipe through the back micro heat plate; Transferring the heat to the main heat dissipation module through the through heat pipe; Performing heat dissipation operation according to the heat through the main heat dissipation module.

10. The method according to claim 9, characterized in that, The heat dissipation operation according to the heat through the main heat dissipation module comprises: Blowing air to the thick aluminum fin array through the turbo blower.