Chip module cooling system for a rack server

Abstract

A chip module cooling system for rack servers includes a server rack, a plurality of server cabinets, a plurality of chip modules, and a plurality of vapor compression cooling systems. Each vapor compression cooling system is composed of at least one heat-absorbing heat exchanger, at least one compressor, at least one heat-dissipating heat exchanger, and at least one expansion device, and forms a refrigerant circulation inside to absorb heat from the chip modules and then lead the heat out of the server cabinet. The compressor and the heat-absorbing heat exchanger are arranged inside the server cabinet. The refrigerant used in the system is dielectric fluid, which can reduce the risk of liquid leakage and conduction and improve safety and reliability. The design can effectively improve the heat dissipation efficiency and system stability of high-power and high-density server environments, and has modularity and scalability, and is suitable for AI, high-performance computing, and cloud data center applications.

Description

Technical Field

[0001] This utility model relates to the field of chip heat dissipation technology, and in particular to a chip module heat dissipation system for a rack server. Background Technology

[0002] With the rapid development of Artificial Intelligence (AI) technology, applications such as deep learning, big data analytics, and real-time computing are constantly increasing their demands for computing performance. To meet these applications, the computing density and power density of server chips are increasing year by year, resulting in higher heat flux during chip operation. Since the failure rate of electronic components increases exponentially with operating temperature, poor heat dissipation will lead to a significant decrease in reliability, thereby affecting system stability and lifespan. Therefore, efficient heat dissipation design for chips has become an indispensable technical requirement for AI servers.

[0003] In existing technologies, air cooling is widely used in general servers due to its simple structure, low cost, and easy maintenance. Air cooling uses fans and heat sinks for heat exchange, and has advantages such as simple structure, low cost, and convenient maintenance. However, due to the low thermal conductivity of air, its heat dissipation performance is often insufficient when the power density of the chip increases, making it difficult to meet the requirements of high heat flux density.

[0004] In direct-to-chip single-phase liquid cooling, single-phase liquid cooling includes two forms: "water-to-air" and "water-to-water". This method uses a pump to drive a water-containing coolant to circulate in the cooling circuit. This coolant absorbs heat through a cold plate placed on the chip and transfers the heat to the heat dissipation end for release.

[0005] Immersion liquid cooling involves directly immersing server components in an electrically insulating coolant, allowing the liquid to dissipate heat through large-area contact with the electronic components. This method can be divided into single-phase immersion and two-phase immersion: In single-phase immersion, the coolant only experiences sensible heat changes, which are carried away by an external heat exchanger; two-phase immersion utilizes the phase change process of coolant evaporation and condensation to transfer latent heat, resulting in higher heat dissipation efficiency. This technology boasts extremely high heat dissipation performance, reduces fan requirements and operating noise, and ensures uniform chip cooling, making it particularly suitable for high-power-density servers.

[0006] Direct-to-chip two-phase liquid cooling utilizes the refrigerant's phase change process (evaporation and condensation) for heat exchange. Existing technologies often employ a pump-driven circulation structure, where a pump propels the refrigerant between the evaporator and condenser. Through the refrigerant's phase change, it absorbs and releases latent heat, achieving high-efficiency heat dissipation. While this method significantly improves heat dissipation performance and reduces system size, it relies solely on a pump-driven circulation. Since the pump lacks compression capabilities, a significant pressure difference cannot be created between the evaporator and condenser, resulting in similar temperatures. Consequently, a sufficient temperature difference cannot be established within the heat exchanger, hindering effective heat transfer and preventing the removal of substantial amounts of heat energy.

[0007] As mentioned above, existing air cooling systems are ill-suited for high-power chips due to air's poor thermal conductivity, requiring high-speed fans that are noisy and inefficient. Direct-to-chip single-phase liquid cooling improves heat dissipation, but the system requires pumps, piping, and quick connectors, resulting in complex structures, inconvenient maintenance, and potential leakage and conductivity issues. Immersion liquid cooling offers highly efficient heat dissipation, but it has strict requirements for material compatibility, necessitating the avoidance of corrosion caused by prolonged contact between the coolant and the circuit board (PCB), solder joints, and seals, while also considering volatility, thermal conductivity, and dielectric properties. System component maintenance and replacement are cumbersome and involve long downtime; insufficient long-term coolant stability can also affect reliability. Therefore, limitations remain in terms of maintenance convenience and liquid management. Direct-to-chip two-phase liquid cooling offers high efficiency, but the lack of pump compression prevents a significant pressure difference between the evaporator and condenser, resulting in similar temperatures and insufficient temperature differential in the heat exchanger. This hinders effective heat removal, and reliability and control complexity remain challenges. Therefore, providing a suitable cooling technology for AI servers that balances high-efficiency heat dissipation with a compact structure and ease of deployment remains an unresolved issue in current technologies.

[0008] In view of this, the inventor of this utility model has devoted himself to researching and applying theoretical principles to address the aforementioned problems in the prior art, which has become the target of the inventor's improvement. Utility Model Content

[0009] Therefore, the main objective of this invention is to provide a method for effectively absorbing the heat generated during chip operation by using a low-pressure, low-temperature refrigerant produced by a vapor compression refrigeration system through heat exchanger structures such as cold plates, microchannel caps, flow channels, or cavities mounted on the chip. This low-temperature refrigerant significantly increases heat dissipation efficiency. The refrigerant used in this system is a dielectric material, which effectively reduces the risk of electrical conductivity caused by leakage, greatly improving the overall system safety and reliability.

[0010] To achieve the aforementioned objectives, this utility model proposes a heat dissipation system for a rack server's encapsulated module, comprising: A server rack contains multiple slots. Multiple server chassis are respectively housed in the chassis slots, and each server chassis has an internal accommodating space; Multiple chip modules, each chip module including at least one chip, wherein the at least one chip module is disposed in the accommodating space of each server chassis; and Multiple vapor compression refrigeration systems, each including at least one heat exchanger, at least one compressor, at least one heat exchanger, and at least one expansion device, wherein the vapor compression refrigeration system contains a circulating refrigerant that absorbs heat from the chip module and then dissipates the heat outside the server chassis; The compressor and the heat exchanger are located in the housing space of each server chassis.

[0011] Preferably, each server chassis has a partition in its accommodating space, which is a non-planar plate, thereby dividing the accommodating space into a first accommodating space and a second accommodating space.

[0012] Preferably, each server chassis has a partition in its accommodating space, the partition having at least one open area, which divides the accommodating space into a first accommodating space and a second accommodating space.

[0013] Preferably, the refrigerant in the vapor compression refrigeration system has a vapor volumetric latent heat (VVLH) value of 25°C. The refrigerant mentioned above is, for example, but not limited to, R1233zd(E), R1336mzz(Z), R454C, R134a, R450A, R513A, R1234yf, R1270, R290, R32, R454B, R410A, R515B, or .

[0014] Preferably, the compressor of the vapor compression refrigeration system is a rotary compressor.

[0015] Preferably, the rotary compressor is a horizontal compressor.

[0016] Preferably, the rotary compressor is a variable frequency compressor.

[0017] Preferably, the rotary compressor is either a two-cylinder compressor or a multi-cylinder compressor.

[0018] Preferably, the height of the rotary compressor is below 2 OU (Open Rack Unit).

[0019] Preferably, when there are multiple heat exchangers, these heat exchangers are arranged in series, in parallel, or in parallel-series configuration.

[0020] Preferably, the expansion device is any one of an electronic expansion valve, a thermal expansion valve, a capillary tube, or an orifice device.

[0021] Preferably, the heat exchanger is either air-cooled or liquid-cooled (single-phase or two-phase liquid cooling).

[0022] Preferably, the heat exchanger is a cold plate. The wafer includes a die, a lid, and a substrate. The die is disposed between the substrate and the lid to form an encapsulation structure. A thermal interface material (TIM) is provided between the die and the lid, and between the lid and the heat exchanger, to form a continuous heat conduction path from the die to the heat exchanger.

[0023] Preferably, the heat exchanger is a micro-channel lid. The chip module includes a die, a lid, and a substrate. The die is disposed between the substrate and the lid to form an encapsulation structure. The heat exchanger is disposed above the die and a thermal interface material (TIM) is provided between the two. Multiple microchannels are formed inside the micro-channel lid.

[0024] Preferably, the heat exchanger is a channel or cavity, and the chip module includes a die, a lid, and a substrate. The channel or cavity is formed in the lid and communicates with the surface of the die.

[0025] As can be seen from the above structure, the beneficial effects of this utility model are as follows: By integrating a vapor compression cooling system into the server architecture, the heat dissipation efficiency and system stability in high-power computing environments are effectively improved. This system employs direct refrigerant cooling technology, which significantly reduces the risk of leakage that may occur with traditional liquid cooling systems, thereby improving the overall operational safety and reliability. Through the refrigerant circulation mechanism, the heat generated by the chip modules can be quickly conducted and dissipated to the outside of the server chassis, achieving a highly efficient, stable, and scalable heat dissipation effect. This heat dissipation architecture features a modular design, suitable for multi-chip module configurations, and can flexibly adapt to different server rack requirements, making it particularly suitable for high-density, high-performance computing environments. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the modular arrangement of multiple server racks in the server rack of this utility model.

[0027] Figure 2 This is a schematic diagram of the configuration of the single-chip module and circuit board of this utility model.

[0028] Figure 3 This is a schematic diagram (I) of the series configuration of the vapor compression cooling system inside the server chassis of this utility model.

[0029] Figure 4 This is a schematic diagram (I) of the parallel configuration of the vapor compression cooling system inside the server chassis of this utility model.

[0030] Figure 5 This is a schematic diagram (II) of the parallel configuration of the vapor compression cooling system inside the server chassis of this utility model.

[0031] Figure 6This is a schematic diagram showing the configuration of multiple vapor compression cooling systems within the server chassis of this utility model.

[0032] Figure 7 This is a schematic diagram (II) of the series configuration of the vapor compression cooling system inside the server chassis of this utility model.

[0033] Figure 8 This is a schematic diagram (III) of the parallel configuration of the vapor compression cooling system inside the server chassis of this utility model.

[0034] Figure 9 This is a schematic diagram of the first heat dissipation method of the chip in the vapor compression cooling system of this utility model.

[0035] Figure 10 This is a schematic diagram of the second heat dissipation method of the chip in the vapor compression cooling system of this utility model.

[0036] Figure 11 This is a schematic diagram of the third heat dissipation method for the chip in the vapor compression cooling system of this utility model.

[0037] Figure 12 This is a schematic diagram of the basic circuit and component configuration of the vapor compression cooling system of this utility model.

[0038] Figure 13 This is a pressure-enthalpy (ph) diagram of the vapor compression cooling system of this utility model.

[0039] Figure 14 This is a pressure-enthalpy (ph) diagram showing the liquid enthalpy and vapor enthalpy of the vapor compression cooling system of this utility model at a certain pressure (corresponding saturation temperature).

[0040] Figure 15 This is a perspective view of the server chassis of this utility model, showing the configuration of the first and second accommodating spaces separated by a partition, and the various vapor compression cooling systems thereon.

[0041] Figure 16 This is a top view of the server chassis of this utility model, showing the configuration of the first and second accommodating spaces separated by a partition, and the vapor compression cooling systems of each space.

[0042] Figure 17 This is a front view of the server chassis of this utility model, showing the configuration of the first and second accommodating spaces separated by a partition, and the respective vapor compression cooling systems.

[0043] Figure 18 This is a side view of the server chassis of this utility model, showing the configuration of the first and second accommodating spaces separated by a partition, and the vapor compression cooling systems of each space.

[0044] Figure 19This is a cross-sectional view of the single-cylinder horizontal compressor of this utility model.

[0045] Figure 20 This utility model Figure 19 A schematic diagram of a twin-cylinder structure.

[0046] Symbol explanation: Detailed Implementation

[0047] To understand the features, content, advantages, and effects of this utility model, the utility model is described in detail below with reference to the accompanying drawings and embodiments. The drawings used are for illustrative purposes only and to assist in the description. They may not represent the actual proportions and precise configurations of the utility model after implementation. Therefore, the proportions and configurations of the accompanying drawings should not be used to interpret or limit the scope of the utility model in actual implementation.

[0048] The advantages, features, and technical methods of this utility model will be more readily understood by referring to the exemplary embodiments and accompanying drawings. This utility model may be implemented in different forms and should not be construed as being limited to the embodiments set forth herein. Rather, the embodiments provided will make this disclosure more thorough, complete, and fully convey the scope of this utility model to those skilled in the art. This utility model will be defined only by the appended claims.

[0049] Please see Figures 1 to 20 The present invention relates to a chip module heat dissipation system for a rack server, comprising: a server rack 100, multiple server chassis 200, multiple chip modules 300, and multiple vapor compression refrigeration systems 400.

[0050] The server rack 100 can be any existing structure available on the market, and its structure is not limited. The server rack 100 has multiple slots 110 inside. More specifically, the server rack 100 can adopt standard 19-inch or 21-inch width rack specifications, or it can be other non-standard size rack structures, and can include open rack or closed cabinet forms to meet the needs of different application environments.

[0051] Multiple server chassis 200 are respectively disposed within slots 110, and each server chassis 200 has an internal accommodating space 210. The server chassis 200 can be of various existing structural types available on the market, and its structural type is not limited. Furthermore, the server chassis 200 can have a removable or modular design, allowing users to quickly install, maintain, or replace server modules, thereby improving system maintainability and scalability.

[0052] Multiple chip modules 300, each chip module 300 including at least one chip 310, wherein at least one chip module 300 is disposed within the accommodating space 210 of each server chassis 200. Further, the chip module 300 may be disposed on and electrically connected to a circuit board 320. The chip 310 may be, but is not limited to, a central processing unit (CPU), a graphics processing unit (GPU), an AI accelerator, a dedicated accelerator (e.g., a TPU, FPGA, or ASIC), a digital signal processor (DSP), a memory module, a network interface controller (NIC) module, or other functional circuit modules. They can be configured individually or arranged in a parallel manner within the accommodating space 210 according to application requirements. More specifically, in one embodiment, the chip module 300 may be equipped with a high-performance GPU module, such as NVIDIA GB200, AMD MI350, or other high-performance GPU modules, to support large-scale artificial intelligence training or high-performance computing (HPC) applications.

[0053] Multiple vapor compression refrigeration systems 400 are provided, each including at least one heat exchanger 410, at least one compressor 420, at least one heat exchanger 430, and at least one expansion device 440. Each vapor compression refrigeration system 400 contains a circulating refrigerant. After absorbing heat from each chip 310 of the chip module 300, the refrigerant is effectively transferred and dissipated to the external environment of the server chassis 200, thereby maintaining the stable performance of the chip module 300 under high-power operation.

[0054] The characteristic feature is that the compressor 420 and the heat exchanger 410 are directly disposed within the accommodating space 210 of each serverchassis 200 to form a local cooling loop, making the cooling system closer to the chip module 300, thereby reducing thermal resistance and improving heat dissipation efficiency. The chip module 300 of this invention can adopt the following three heat dissipation methods, further described below: Type 1: such as Figure 9 As shown, this embodiment provides a refrigerant-based chip cooling structure. A heat exchanger 410 is disposed on the surface of the chip 310 of the chip module 300. The heat exchanger 410 can be a cold plate, with internal channels 411 for refrigerant flow to achieve a circulating cooling effect. The chip 310 also includes a die 311, a lid 312, and a substrate 313. The die 311 is disposed between the substrate 313 and the lid 312 to form an encapsulation structure. The substrate 313 provides the necessary electrical connections for each die 311 to achieve signal transmission. A first thermal interface material (TIM) T1 is provided between the die 311 and the lid 312, and a second thermal interface material (TIM) T2 is provided between the lid 312 and the heat exchanger 410, to form a continuous heat conduction path from the die 311 to the heat exchanger. When the refrigerant flows in through the inlet of the flow channel 411 of the heat exchanger 410, it circulates in the internal flow channel, absorbs heat from the die 311, and is discharged from the outlet of the flow channel 411. Due to the high thermal conductivity and high latent heat characteristics of the refrigerant, the heat energy generated by the die 311 can be effectively removed in a short time.

[0055] Type 2: such as Figure 10As shown, this embodiment provides a refrigerant-based microchannel capped chip cooling structure. The heat exchanger 410 is disposed on the chip 310. The heat exchanger 410 can be a microchannel cap. The chip 310 includes a die 311, a cap 312, and a substrate 313. The die 311 is disposed between the substrate 313 and the cap 312 to form an encapsulation structure. The heat exchanger 410 is placed above the die 311 and has multiple microchannels 412 formed inside it. The refrigerant can flow in from the inlet of these microchannels 412 and flow directly through the area above the die 311 along the microchannels 412 before being discharged from the outlet of the microchannels 412. With this configuration, the refrigerant can directly exchange heat within the high heat flux region of the die 311, significantly shortening the heat conduction distance and allowing the heat generated by the die 311 to be quickly carried away by the refrigerant, thereby improving overall heat dissipation efficiency and cooling performance. A thermal interface material (TIM) T3 is provided between the die 311 and the heat exchanger 410 to ensure thermal conductivity.

[0056] Type 3: such as Figure 11 As shown, this embodiment provides a refrigerant-based direct cooling structure, in which the heat exchanger 410 is disposed on the chip 310. The chip 310 includes a die 311, a lid 312, and a substrate 313. The heat exchanger 410 can be a channel or cavity, formed within the lid 312 and communicating with the surface of the die 311, allowing the refrigerant to flow directly to the surface of the die 311 for cooling, thereby absorbing the heat generated by the die 311 during operation. After heat exchange is completed, the refrigerant is discharged from the outlet, thereby improving the overall heat removal efficiency and effectively preventing local overheating. Notably, in this embodiment, no thermal interface material (TIM) is provided between the heat exchanger 410 and the die 311.

[0057] Furthermore, the compressor 420 drives the refrigerant circulation and compresses the low-pressure, low-temperature refrigerant gas into a high-pressure, high-temperature state to facilitate subsequent heat dissipation. The heat exchanger 430 receives the high-pressure, high-temperature refrigerant gas generated by the compressor 420 and releases its heat energy outside the server chassis 200. The expansion device 440 throttles and reduces the pressure of the high-pressure refrigerant cooled by the heat exchanger 430, converting it into a low-pressure, low-temperature liquid, and then introduces it back into the heat exchanger 410 to complete the circulation process. The heat exchanger 430 and the expansion device 440 can be adjusted according to actual needs and can be configured inside or outside the server chassis 200, or at the boundary between the inside and outside of the chassis 200, such as the edge of the server chassis 200, to further improve the flexibility of space utilization and achieve a highly integrated modular design. This configuration can be selected based on system heat dissipation requirements, structural layout limitations, and ease of maintenance, offering flexibility and compatibility.

[0058] In terms of application effectiveness, the vapor compression cooling system 400 significantly improves the heat dissipation performance of the high-power chip modules 300 within the server chassis 200, reducing the risk of frequency throttling, performance degradation, or hardware failure caused by overheating, thereby enhancing the overall computing performance and reliability of the server. This design is particularly suitable for applications in artificial intelligence (AI), high-performance computing (HPC), cloud data centers, and edge computing, providing a highly reliable and scalable heat dissipation solution.

[0059] Furthermore, when the server chassis 200 contains one or more chip modules 300, the number of vapor compression refrigeration systems 400 can be the same as the number of chip modules 300, forming a one-to-one configuration. In this case, each chip module 300 is cooled by an independent vapor compression refrigeration system 400, where the number of heat exchangers 410 corresponds to the number of chips 310 in that chip module 300. This configuration ensures that each chip module 300 has its own dedicated cooling circuit, can operate independently, and effectively avoids mutual interference of thermal loads, thereby ensuring stable and efficient heat dissipation even under high-power operation. In addition, in a single-module scenario, the number of the vapor compression refrigeration system 400 can be greater than the number of the chip module 300, forming a many-to-one configuration, which further improves heat dissipation performance and system reliability, and is suitable for occasions that require long-term continuous operation or have high reliability requirements.

[0060] Furthermore, when the server chassis 200 contains one or more chip modules 300, the number of vapor compression refrigeration systems 400 can be less than the number of chip modules 300, forming a one-to-many configuration. In this case, at least one vapor compression refrigeration system 400 is provided, and components such as the heat exchanger 410, compressor 420, heat exchanger 430, and expansion device 440 can be added as needed, providing centralized heat dissipation for multiple chip modules 300 through the same refrigerant circuit. This configuration effectively simplifies the system architecture and achieves resource sharing, making it particularly suitable for applications with limited space or where cost-effectiveness is emphasized.

[0061] Based on the above structure, further explanation is provided below: Each server chassis 200 may have a partition 220 installed in its accommodating space 210. The partition 220 is fixed to the server chassis 200, thereby dividing the accommodating space 210 into a first accommodating space 211 and a second accommodating space 212 to achieve a partitioned layout. The first accommodating space 211 and the second accommodating space 212 are used to house the chip module 300 and the compressor 420 and heat exchanger 410 of the vapor compression refrigeration system 400. The heat exchanger 430 and the expansion device 440 can be adjusted according to actual needs and can be configured inside or outside the server chassis 200, or at the boundary between the inside and outside of the chassis 200, to complete the refrigerant circulation loop.

[0062] More specifically, the partition 220 can be a non-planar structural plate, and its surface morphology can include curved surfaces, angles, corrugations, or multi-step structures. Its shape can be optimized according to the installation requirements of the compressor 420 inside the server chassis 200. The partition 220 can also be manufactured using stamping technology to form a concave-convex structure or hollow areas to provide installation space for the compressor 420, further improving the overall structural integration and space configuration efficiency. In other embodiments, the partition 220 can be designed as a detachable or modular structure, allowing users to replace different types of partitions 220 according to the size, cooling capacity, or maintenance needs of the compressor 420, achieving highly flexible design and convenient maintenance.

[0063] Please refer to Table 1, which lists the vapor enthalpy, liquid enthalpy, enthalpy difference (Δh), vapor density, and latent heat per unit volume (VVLH) of the refrigerant vapor for different refrigerants at an evaporation temperature of 25°C. Figure 14 As shown, the enthalpy difference (Δh) corresponds to h1-h4' in the figure. The above data was calculated and compiled based on the REFPROP V10 thermophysical property database published by the National Institute of Standards and Technology (NIST) for reference in refrigerant performance analysis and system design.

[0064] Among them, the latent heat per unit volume (VVLH) of refrigerant vapor is an important indicator for evaluating the heat dissipation capacity of refrigerant, which can simultaneously reflect the amount of latent heat and vapor density of refrigerant. When the VVLH value is higher, it means that the refrigerant vapor can cool more heat energy under the same volume conditions, which can significantly improve the heat dissipation efficiency for the space-constrained server chassis 200.

[0065] In one embodiment, the refrigerant of the vapor compression refrigeration system 400, at an evaporation temperature of 25°C, has a vapor volumetric latent heat (VVLH) value that can reach [value missing]. The above. The refrigerant may be, for example, but not limited to, R1233zd(E), R1336mzz(Z), R454C, R134a, R450A, R513A, R1234yf, R1270, R290, R32, R454B, R410A, R515B, or... wait.

[0066] In terms of application differences, refrigerants with higher VVLH values ​​are suitable for cooling high heat flux density modules (such as NVIDIA GB200, AMD MI350, and other high-performance GPU modules); while refrigerants with medium VVLH values ​​are suitable for general servers or data center computing nodes, balancing energy efficiency and system cost. This tiered refrigerant usage model allows for flexible adjustments to the cooling system design based on different server application scenarios.

[0067] However, although water is also listed in Table 1, its VVLH value at 25°C is only about The refrigerant concentration is far lower than that of the aforementioned refrigerants, and in practical applications, it has problems such as high conductivity and potential damage to electronic components after leakage. Therefore, water is explicitly excluded as a refrigerant in this invention. All other refrigerant types listed in Table 1 are applicable to the vapor compression cooling system 400 described in this invention.

[0068]

[0069] In one embodiment, the compressor 420 used in the vapor compression refrigeration system 400 can be a rotary compressor, which has the characteristics of simple structure, stable operation and high efficiency, and is particularly suitable for applications within the limited space of the server rack 100. Furthermore, the rotary compressor can be a horizontal compressor to effectively reduce the height of the unit and facilitate installation in the housing space 210 of the server chassis 200. In another embodiment, the compressor 420 can be a variable frequency compressor, which can adapt the compression ratio and speed in real time according to the changes in the computing load of each chip 310 of the chip module 300 in the server, thereby improving energy efficiency and reducing unnecessary energy consumption. Moreover, the compressor 420 can also be a twin-cylinder compressor (such as...). Figure 20 (As shown) or multi-cylinder compressors are used to meet the requirements of low vibration and high refrigerant flow rate, maintain reliability during long-term operation, and improve refrigerant delivery efficiency and stability. The horizontal compressor can be of various existing structural types available on the market, and its structural type is not limited.

[0070] The horizontal compressor may include a housing 421, an electric motor 422, a compression pump 423, and an oil suction pipe 424. The electric motor 422 includes a stator 4221 fixed to the inner wall of the housing 421, and a rotor 4222 rotatably disposed inside the stator 4221. The compression pump 423 is disposed within the housing 421 and is linked to the electric motor 422. The compression pump 423 includes at least a cylinder 4231, an upper support 4232, a lower support 4233, and a crankshaft 4234. A compression chamber 42311 is centrally located within the cylinder 42311, and a piston 42312 is disposed within the compression chamber 42311. The upper support 4232 and the lower support 4233 are respectively disposed within the cylinder 4231. The crankshaft 4234 has at least one eccentric portion 42341 to define an upper shaft section 42342 and a lower shaft section 42343. The upper shaft section 42342 is fitted with the rotor 4222 of the electric motor 422 and the upper support 4232, while the lower shaft section 42343 is fitted with the lower support 4233. The eccentric portion 42341 is fitted with the piston 42312 of the cylinder block 4231 to drive compression. One end of the oil suction pipe 424 is connected to an oil trough 425 at the bottom of the compressor pump 423, and an oil suction cover 4235 is added at the lower support 4233, so that the other end of the oil suction pipe 424 can extend into the oil suction cover 4235 and be located at the bottom of the crankshaft 4234. By means of the centrifugal force and pressure difference generated by the rotation of the crankshaft 4234, the lubricating oil is delivered to the contact part of the compressor pump 423 to ensure the lubrication and reliability of the compressor 420 during long-term operation, maintain stable oil supply, and further improve the system life and operational stability.

[0071] When the refrigerant enters the compression chamber 42311 through the suction pipe 426, it is compressed into a high-pressure, high-temperature refrigerant gas by the rotor 4222 and then delivered to the heat exchanger 430 through the discharge port 427. Its main body is arranged horizontally, so that the motor 422 and the compressor pump 423 are aligned with their axes at a horizontal or inclined angle of ±15 degrees. This design effectively reduces the overall height of the compressor 420, making it suitable for installation within the limited space 210 of the server chassis 200 and minimizing its vertical footprint. In addition to reducing the vertical footprint, this design also allows for a coplanar configuration with the heat exchanger 410 and refrigerant piping, improving the compactness and modular flexibility of the heat dissipation module. Furthermore, by lowering the center of gravity, vibration and noise are reduced, thereby enhancing the stability and reliability of the server cooling system.

[0072] Furthermore, the height of the rotary compressor can be limited to below 2 OU (Open Rack Unit), where 1 OU = 48 mm and 2 OU = 96 mm, to meet the space specifications of the standard server rack 100. Through this miniaturized design, the compressor 420 can be coplanarly configured with components such as the heat exchanger 410 or cooling pipes, which not only improves the compactness of the internal heat dissipation module layout of the server but also increases the flexibility of modular integration, facilitating system expansion and maintenance in different application scenarios. Therefore, the miniaturization and modularity of the compressor 420 significantly enhance the applicability and practical value of this heat dissipation system in high-density server environments.

[0073] In one embodiment, the number of the heat exchangers 410 can be multiple, and under different design requirements, these heat exchangers 410 can be arranged in series, parallel, or parallel-series configurations to correspond to different cooling load conditions and optimize heat dissipation. Furthermore, the expansion device 440 can be an electronic expansion valve or a thermostatic expansion valve, capable of precisely controlling and adjusting the refrigerant flow to meet dynamic heat dissipation needs; it can also be a traditional capillary tube or orifice device, offering advantages of simple structure and high reliability to meet the application requirements of server cooling systems. On the other hand, the heat exchanger 430 can adopt an air-cooled structure, dissipating heat to the ambient air through air cooling; or it can be a liquid-cooled structure (single-phase or two-phase liquid cooling), using cooling water or other refrigerants for efficient heat dissipation. Through the above-mentioned diverse component selection and configuration methods, the chip module heat dissipation system of this utility model possesses high flexibility and scalability, adapting to the needs of different server architectures and computing environments.

[0074] In terms of design advantages, the series-connected heat exchanger 410 provides tiered heat dissipation and has a simpler structure; while the parallel configuration effectively reduces the flow resistance of a single cooling loop, making it suitable for scenarios where multiple chip modules 300 operate at high speed simultaneously. The parallel-series combination offers both efficiency and flexibility in flow distribution, making it suitable for heterogeneous computing server environments. On the other hand, while the air-cooled heat exchanger 430 has a simple structure and is suitable for general edge server applications, the liquid-cooled (single-phase or two-phase liquid cooling) heat exchanger 430 is more advantageous in high-density servers or cloud data centers. It can efficiently dissipate heat through external cooling water systems or data center cooling infrastructure, significantly improving overall system performance and reliability. Therefore, the heat dissipation system of this invention combines configuration flexibility and application adaptability, providing optimal heat dissipation solutions for different computing environments.

[0075] By integrating the Vapor Compression Refrigeration system 400 into the server architecture, heat dissipation efficiency and system stability in high-power computing environments can be effectively improved. This system employs direct refrigerant cooling technology, significantly reducing the risk of leakage that may occur with traditional liquid cooling systems, thereby enhancing overall operational safety and reliability. Through a refrigerant circulation mechanism, the heat generated by the chip modules can be rapidly dissipated to the external environment of the server chassis, achieving efficient, stable, and scalable heat dissipation.

[0076] Furthermore, the heat dissipation architecture of this invention features a modular design, which can be adjusted according to the number and power density of the 300 multi-chip modules inside the server, flexibly adapting to both single and multiple cooling loops. This design not only improves heat dissipation efficiency but also facilitates subsequent maintenance and expansion, reduces system downtime, and ensures long-term stable operation of the data center.

[0077] Furthermore, this system can be paired with a horizontal compressor to further reduce space requirements and forms a coplanar configuration with the heat exchanger 410 and refrigerant piping, thereby improving the overall compactness of the heat dissipation module. Its applications cover various fields such as artificial intelligence (AI), high-performance computing (HPC), cloud data centers, and edge computing, and it is particularly suitable for high-density server racks and GPU cluster environments. Through these features, this invention not only effectively solves the limitations of traditional heat dissipation methods in high-power environments but also demonstrates high practical value and industrial application potential.

[0078] In summary, this invention provides a method for effectively absorbing the heat generated during chip operation by using a low-pressure, low-temperature refrigerant produced by a vapor compression refrigeration system through a heat exchanger 410 structure, such as a cold plate, microchannel cap, flow channel, or cavity, mounted on the chip 310. This low-temperature refrigerant significantly increases heat dissipation efficiency. The refrigerant used in this system is a dielectric material, which effectively reduces the risk of electrical conductivity due to leakage, greatly improving the overall system safety and reliability.

[0079] The above description is merely an embodiment of this utility model and should not be construed as limiting the scope of implementation of this utility model. Any simple equivalent changes and modifications made in accordance with the claims and patent specification of this utility model shall still fall within the scope of this utility model patent.

Claims

1. A chip module heat dissipation system for a rack server, comprising: include: A server rack with multiple cabinet layers inside; Multiple server boxes are respectively set in the box layer, and each server box has an accommodating space inside; Multiple chip modules, each chip module comprising at least one chip, wherein the at least one chip module is disposed in the accommodating space of each server chassis; and Multiple vapor compression cooling systems, each vapor compression cooling system including at least one heat-absorbing heat exchanger, at least one compressor, at least one heat-dissipating heat exchanger and at least one expansion device, wherein the vapor compression cooling system contains a refrigerant circulating inside, and after absorbing heat from the chip module, dissipates the heat outside the server chassis; The compressor and the heat exchanger are located in the housing space of each server box.

2. The rack server's chip module heat dissipation system of claim 1, wherein, Each server box has a partition in its accommodating space. The partition is a non-planar plate, which divides the accommodating space into a first accommodating space and a second accommodating space.

3. The rack server's chip module cooling system of claim 1, wherein, Each server box has a partition in its accommodating space, and the partition has at least one open area, which divides the accommodating space into a first accommodating space and a second accommodating space.

4. The rack server's chip module cooling system of claim 1, wherein, The latent heat per unit volume of the refrigerant in this vapor compression cooling system, at an evaporation temperature of 25°C, is [value missing]. The refrigerant mentioned above is R1233zd(E), R1336mzz(Z), R454C, R134a, R450A, R513A, R1234yf, R1270, R290, R32, R454B, R410A, R515B or .

5. The rack server's chip module cooling system of claim 1, wherein, The compressor in this vapor compression cooling system is a rotary compressor.

6. The rack server's chip module cooling system of claim 5, wherein, This rotary compressor is a horizontal compressor.

7. The rack server's chip module cooling system of claim 5, wherein, This rotary compressor is a variable frequency compressor.

8. The rack server's chip module cooling system of claim 5, wherein, This rotary compressor is either a two-cylinder compressor or a multi-cylinder compressor.

9. The rack server's chip module cooling system of claim 5, wherein, The height of this rotary compressor is less than 2 OU.

10. The rack server's chip module cooling system of claim 1, wherein, When there are multiple heat exchangers, these heat exchangers are arranged in series, in parallel, or in parallel-series configurations.

11. The rack server's chip module heat dissipation system of claim 1, wherein, The expansion device is any of an electronic expansion valve, a thermal expansion valve, a capillary tube, or an orifice device.

12. The rack server's chip module cooling system of claim 1, wherein, The heat exchanger is either air-cooled or liquid-cooled (single-phase or two-phase liquid cooling).

13. The rack server's chip module cooling system of claim 1, wherein, The heat exchanger is a cold plate. The chip includes a bare die, a cap and a substrate. The bare die is disposed between the substrate and the cap to form an encapsulation structure. A thermally conductive interface material is provided between the bare die and the cap and between the cap and the heat exchanger to form a continuous heat conduction path from the bare die to the heat exchanger.

14. The rack server's chip module cooling system of claim 1, wherein, The heat exchanger is a microchannel cap. The chip module includes a bare die, a cap and a substrate. The bare die is disposed between the substrate and the cap to form an encapsulation structure. The heat exchanger is disposed above the bare die and a thermally conductive interface material is provided between the two. Multiple microchannels are formed inside the microchannel cap.

15. The rack server's chip module cooling system of claim 1, wherein, The heat exchanger is a channel or cavity. The chip module includes a bare die, a cap and a substrate. The channel or cavity is formed in the cap and communicates with the surface of the bare die.

Images