Immersed supercomputing device and immersed computing power system

Abstract

The invention provides an immersed supercomputing device and an immersed computing power system. The immersed supercomputing device comprises a machine shell and a computing power heat dissipation module, the machine shell comprises a box body, a first side plate and a third side plate of the box body are opposite, and the first side plate and the third side plate are correspondingly provided with installation guide grooves with opposite openings; the computing power heat dissipation module comprises a first heat dissipation device and a computing power plate, a first heat conduction plate of the first heat dissipation device is attached to the computing power plate, first fins extend out of the two ends of the first heat conduction plate in the opposite direction of the first side plate and the third side plate, and the extending parts of the first fins are inserted into the mounting guide grooves; first gaps are formed between the first side plate and the nearest first fin as well as between the third side plate and the nearest first fin, and the size of the first gaps in the opposite direction is smaller than or equal to two times of the tooth spacing. According to the invention, the manufacturing difficulty of the radiator can be reduced, and the installation reliability of the power radiating module is ensured; and meanwhile, the heat dissipation efficiency of the liquid cooling medium is improved, and the temperature balance of the computing power chips is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of supercomputing equipment heat dissipation technology, and particularly relates to an immersion type supercomputing equipment and an immersion type computing power system. BACKGROUND

[0002] With the continuous improvement of the operation speed and operation amount of supercomputing servers, the heat generated by the computing power chips on the computing power boards of the supercomputing servers also greatly increases. In order to improve the heat dissipation effect of the computing power chips on the computing power boards, an immersion type heat dissipation method is proposed, that is, the supercomputing server is immersed in a liquid cooling medium. The liquid cooling medium can be selected from electrically insulating cooling liquids, such as insulating oil substances, fluorinated liquids, etc.

[0003] The existing immersion type supercomputing equipment has a specific guide groove structure arranged on the heat sink of the computing power heat dissipation module, and a guide rail arranged on the side plate of the casing, so as to install the computing power heat dissipation module in the casing through the cooperation of the guide rail and the guide groove, such as the prior art CN 215956915U. However, this method undoubtedly increases the manufacturing difficulty of the heat sink, because the heat sink itself is composed of a heat conduction plate and fins, and the guide groove structure is added. More importantly, when the liquid cooling medium flows through the heat sink, a large amount of liquid cooling medium is dispersed between the heat sink and the casing, so that part of the liquid cooling medium cannot exert the maximum heat dissipation effect, and the heat dissipation efficiency of the liquid cooling medium is limited. SUMMARY

[0004] The main purpose of the present application is to provide an immersion type supercomputing equipment and an immersion type computing power system, so as to reduce the manufacturing difficulty of the heat sink, ensure the reliability of the installation of the computing power heat dissipation module, improve the heat dissipation efficiency of the liquid cooling medium, and further improve the temperature uniformity of each computing power chip.

[0005] According to a first aspect of the present application, an immersion type supercomputing equipment is provided, comprising a casing and a computing power heat dissipation module, The casing comprises a box body, and the box body comprises first, second, third and fourth side plates connected in sequence, the first and third side plates are opposite to each other and correspondingly provided with open installation guide grooves opposite to each other; The computing power heat dissipation module comprises a first heat sink and a computing power board, the first heat sink comprises a first heat conduction plate and a plurality of first fins arranged at intervals on the same surface of the first heat conduction plate, the first heat conduction plate is arranged in close contact with the computing power board, and the two ends of the first heat conduction plate in the opposite direction of the first and third side plates are both provided with first fins, and the protruding parts are inserted into the installation guide grooves; The first side plate and the third side plate each form a first gap with the nearest first fin, and the size of the first gap in the opposite direction is less than or equal to 2 times the tooth spacing.

[0006] Optionally, the shell has a medium inlet, and the shell is provided with a first shielding structure corresponding to the area of the first gap on the side of the medium inlet.

[0007] Optionally, the box body is a through structure, and the shell further comprises a first panel covering the opening of the box body close to the medium inlet, and the medium inlet and the first shielding structure are formed in the first panel.

[0008] Optionally, at least the first fin in the computing power heat dissipation module is attached to the first panel.

[0009] Optionally, the first shielding structure is formed by bending the first side plate and the third side plate.

[0010] Optionally, the computing power heat dissipation module further comprises a second heat sink, the second heat sink comprises a second heat conduction plate and a plurality of second fins arranged on the same surface of the second heat conduction plate; the first heat conduction plate and the second heat conduction plate are respectively attached to two surfaces of the computing power board. The first side plate and the third side plate each form a second gap with the nearest second fin, and the size of the second gap in the opposite direction is less than or equal to 2 times the tooth spacing.

[0011] Optionally, the computing power board comprises a substrate and a computing power chip arranged on one surface of the substrate; the first heat conduction plate is attached to the other surface of the substrate. The second heat conduction plate is attached to the computing power chip; the height of the second fin is greater than the height of the first fin; in the opposite direction, the size of the second gap is smaller than the size of the first gap.

[0012] Optionally, on the side where the first side plate is located, the outermost second fin is closer to the first side plate than the opening of the mounting guide groove. On the side where the third side plate is located, the outermost second fin is closer to the third side plate than the opening of the mounting guide groove. Optionally, the shell has a medium inlet, and the shell is provided with a second shielding structure corresponding to the area of the second gap on the side of the medium inlet.

[0013] Optionally, a first shielding structure is provided in the area of ​​the casing corresponding to the first gap on the medium inlet side; when projected in a direction perpendicular to the first heat-conducting plate, the projections of the first shielding structure and the second shielding structure are both located outside the projection of the area where the computing chip is located on the computing power board, and in the relative direction, the distance between the first shielding structure and the second shielding structure and the area where the computing chip is located is greater than or equal to 2 times the tooth pitch.

[0014] Optionally, the first side plate and the third side plate are provided with two sets of mounting guide grooves; two computing power heat dissipation modules are provided, which are respectively mounted on the two sets of mounting guide grooves through the first heat conduction plate, wherein the first fin of the computing power heat dissipation module closer to the second side plate faces the second side plate, and the distance between the first fin located in at least the middle area and the second side plate is less than or equal to one tooth pitch; the second fin of the other computing power heat dissipation module faces the fourth side plate, and the distance between the second fin located in at least the middle area and the fourth side plate is less than or equal to one tooth pitch.

[0015] Optionally, the ratio of the thickness of the first fin and the second fin to the tooth pitch is less than or equal to 1 / 4.

[0016] Optionally, the thickness of the first fin and the second fin 232 is 0.3mm to 1.2mm, and the tooth pitch is 1.4mm to 2.8mm.

[0017] Optionally, the computing power heat dissipation module further includes a heat spreader and a heat pipe; the first side of the heat spreader is provided with a strip groove, and the heat pipe is disposed in the strip groove; At least one heat spreader is provided between the second heat-conducting plate and the computing chip, and the second side of the heat spreader is in contact with the computing chip, and heat pipes are arranged at the computing chip.

[0018] A second aspect of the present invention provides an immersion computing system, including a liquid cooling tank and an immersion supercomputing device as described in any of the preceding claims, wherein the immersion supercomputing device is disposed in the liquid cooling tank and the medium inlet of the casing faces the bottom of the liquid cooling tank.

[0019] This invention utilizes mounting guide grooves on the casing, directly employing the first heat-conducting plate of the first heat sink to mate with the mounting guide grooves. Furthermore, since only the edge of the first heat-conducting plate needs to extend beyond the first fins, the manufacturing difficulty of the first fins is reduced. The mating of the first heat-conducting plate with the mounting guide grooves on the casing ensures the stability and reliability of the computing power cooling module installation. Moreover, the extended portion of the first heat-conducting plate in this invention only requires a small size to stably mate with the mounting guide groove. Therefore, this invention can simultaneously minimize the first gap between the first fins and the casing, making it less than or equal to twice the tooth pitch. When the immersion supercomputing device is immersed in liquid cooling medium, this significantly reduces the amount of liquid cooling medium flowing between the outermost first fins and the casing, ensuring that most of the liquid cooling medium flows through the channels between the first fins, thereby improving the heat dissipation efficiency of the liquid cooling medium.

[0020] Other beneficial effects of the present invention will be explained in detail through the introduction of specific technical features and technical solutions in specific embodiments. Those skilled in the art should be able to understand the beneficial technical effects brought about by these technical features and technical solutions through the introduction of these technical features and technical solutions. Attached Figure Description

[0021] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings.

[0022] Figure 1 This is a schematic diagram of a preferred embodiment of the immersion supercomputing device of the present invention; Figure 2 This is a cross-sectional view of a preferred embodiment of the immersion supercomputing device of the present invention; Figure 3 for Figure 2 A magnified view of a portion of the text; Figure 4 This is a schematic diagram of a preferred embodiment of the enclosure in the immersion supercomputing device of the present invention; Figure 5 This is a schematic diagram of a preferred embodiment of the computing power heat dissipation module in the immersion supercomputing device of the present invention; Figure 6 This is a schematic diagram of the connection structure of a preferred embodiment of the computing power heat dissipation module's computing power board in the immersion supercomputing device of the present invention. Figure 7 This is a heat distribution diagram of the computing board in the immersion supercomputing device of the present invention.

[0023] In the picture: 10. Housing; 11. Box body; 111. First side panel; 112. Second side panel; 113. Third side panel; 114. Fourth side panel; 115. Mounting guide groove; 1151. First groove wall; 1152. Second groove wall; 12. First panel; 13. Second panel; 14. First shielding structure; 15. Second shielding structure.

[0024] 20. Computing power heat dissipation module; 21. First heat sink; 211. First heat conduction plate; 212. First fin; 22. Computing power board; 221. Substrate; 2211. Chip area; 222. Computing power chip; 23. Second heat sink; 231. Second heat conduction plate; 232. Second fin; 30. First gap; 40. Second gap; 50. Power supply module. Detailed Implementation

[0025] The present invention is described below based on embodiments, but the present invention is not limited to these embodiments. In the following detailed description of the present invention, some specific details are described in detail, but well-known methods, processes, procedures, and elements are not described in detail in order to avoid obscuring the essence of the present invention.

[0026] Furthermore, those skilled in the art should understand that the accompanying drawings provided herein are for illustrative purposes only and are not necessarily drawn to scale.

[0027] Unless the context explicitly requires it, the words "comprising," "including," and similar terms throughout the specification and claims should be interpreted as encompassing rather than being exclusive or exhaustive; that is, meaning "including but not limited to."

[0028] In the description of this invention, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0029] This invention provides an immersion supercomputing device that can be fully submerged in a liquid cooling medium, or only partially submerged, such as only submerging the section containing the computing power heat dissipation module. The liquid cooling medium can be an electrically insulating coolant, such as insulating oils or fluorinated liquids.

[0030] like Figures 1-5As shown, the immersion supercomputing device includes a casing 10 and a computing power heat dissipation module 20. The casing 10 includes a box 11, which includes a first side plate 111, a second side plate 112, a third side plate 113 and a fourth side plate 114 connected end to end in sequence. The first side plate 111 and the third side plate 113 are opposite to each other and are respectively provided with mounting guide grooves 115 with opposite openings.

[0031] The computing power heat dissipation module 20 includes a first heat sink 21 and a computing power plate 22. The first heat sink 21 includes a first heat-conducting plate 211 and a plurality of spaced first fins 212 disposed on the same side of the first heat-conducting plate 211. The first heat-conducting plate 211 is attached to the computing power plate 22 (including direct attachment and indirect attachment, detailed below). In the opposite direction (i.e., the second direction Y) between the first side plate 111 and the third side plate 113, both ends of the first heat-conducting plate 211 are inserted into the mounting guide grooves 115. That is, in the second direction, both ends of the first heat-conducting plate 211 extend beyond the outermost first fins 212, and the extended portion is inserted into the mounting guide grooves 115. In other words, when projected in the first direction X, the projections of the first fins 212 are all located between the two opposite mounting guide grooves 115. Furthermore, when projected in the first direction X, the edge of the computing power plate 22 does not extend beyond the two outermost first fins 212 in the second direction Y.

[0032] The first side plate 111 and the third side plate 113 each form a first gap 30 with the nearest first fin 212. The size of the first gap 30 in the relative direction is less than or equal to twice the tooth pitch; wherein, the tooth pitch refers to the distance between two adjacent first fins 212.

[0033] like Figure 4 As shown, the housing 11 has a square structure, specifically a cuboid housing. The first side plate 111, second side plate 112, third side plate 113, and fourth side plate 114 are connected end-to-end, forming a basically closed circumferential structure. The housing forms an inlet and outlet side for the liquid cooling medium at both ends perpendicular to the circumferential direction X, with the liquid cooling medium flowing from the inlet side to the outlet side. The inner surface of the first side plate 111 protrudes with a mounting guide groove 115, the opening of which faces the third side plate 113. The mounting guide groove 115 includes a first groove wall 1151 and a second groove wall 1152, with the first groove wall 1151 being closer to the fourth side plate 114 than the second groove wall 1152. Correspondingly, the inner surface of the third side plate 113 also protrudes with a mounting guide groove 115. The opening of the mounting guide groove 115 faces the first side plate 111, and its first groove wall 1151 is closer to the fourth side plate 114 than the second groove wall 1152. Figure 2 It can be seen that the mounting guide groove 115 on the first side plate 111 and the mounting guide groove 115 on the third side plate 113 form a pair and are used together to install the same first heat conduction plate 211.

[0034] refer to Figure 5 In the computing power heat dissipation module 20, the first heat sink 21 has multiple first fins 212 spaced apart in opposite directions. These first fins 212 are evenly distributed in the second direction Y, and each first heat dissipation fin 212 is substantially parallel to the first direction X. Flow channels are formed between adjacent first heat dissipation fins 212. Thus, the liquid cooling medium entering from the inlet side of the housing 11 is diverted by the multiple first heat dissipation fins and flows through each flow channel to remove heat from the first heat dissipation fins 212. Projected along the third direction, the projection of the computing board 22 does not exceed the projection of the two outermost first heat dissipation fins 212. The first groove wall 1151 is substantially opposite to the computing board 22, and the thickness of the first groove wall 1151 does not exceed the thickness of the computing board 22.

[0035] like Figure 3 As shown, except for the extension portion that mates with the mounting guide groove 115, the outermost first fin 212 of the first heat sink 21 is positioned as close as possible to the first side plate 111, and the outermost first fin 212 of the second side plate 112 is positioned as close as possible to the third side plate 113. Specifically, the first gap between the outermost first fin 212 and the housing 10 in the second direction Y is less than twice the tooth pitch d.

[0036] In the above embodiments, by providing a mounting guide groove 115 on the housing 10, the first heat-conducting plate 211 of the first heat sink 21 directly engages with the mounting guide groove 115. Since only the edge of the first heat-conducting plate 211 needs to extend beyond the first fin 212, the manufacturing difficulty of the first fin 212 is reduced. Furthermore, the engagement of the first heat-conducting plate 211 with the mounting guide groove 115 on the housing 10 ensures the stability and reliability of the computing power cooling module 20 installation. Further, in this invention, the extended portion of the first heat-conducting plate 211 only needs a small size to stably engage with the mounting guide groove 115. Therefore, this invention can also minimize the first gap 30 between the first fin 212 and the housing 10, making it less than or equal to twice the tooth pitch. When the immersion supercomputing device is immersed in the liquid cooling medium, this significantly reduces the amount of liquid cooling medium flowing between the outermost first fin 212 and the housing 10, ensuring that most of the liquid cooling medium flows through the channels between the first fins 212, thereby improving the heat dissipation efficiency of the liquid cooling medium. Figure 7The figure shows the heat distribution on the computing board when the liquid cooling medium, such as liquid cooling oil, has an inlet temperature of 55°C, a flow rate of about 24L / min, and a total power consumption of 6KW. As can be seen from the figure, the setting method of the present invention can achieve better heat dissipation for each computing chip, and the temperature difference between the two ends of the computing chips in the first direction X is basically less than 8°C, so that the entire computing board 22 can better and continuously maintain the optimal working state.

[0037] In one embodiment, such as Figure 4 As shown, the housing 11 extends through in the first direction X, with openings at both ends of the through structure, located on the medium inlet side and the medium outlet side of the housing 10, respectively. A mounting guide groove 115 extends from one opening of the housing 11 to the other. Alternatively, the housing 11 can be a structure where both ends are not fully through, as described below, where the first side plate 111 can be bent to form a first shielding structure 14 and a second shielding structure 15. The housing 11 can also have a bottom plate located on the inlet side. In the case of a partially through structure, the mounting guide groove 115 also extends from the inlet side of the housing 10 to the outlet side.

[0038] In one embodiment, such as Figure 1 and Figure 2 The housing 10 also includes a first panel 12 and a second panel 13. The two panels can be basically plate-shaped structures. The first panel 12 is located on the entrance side of the housing 11, and the second panel 13 is located on the exit side of the housing. When the housing 11 is a through structure, the first panel 12 covers the opening on the entrance side of the housing 11; the second panel 13 covers the opening on the exit side of the housing. When the entrance side of the housing 11 is a non-through structure, the first panel 12 covers at least the other areas on the entrance side of the housing 11, and can of course cover the entire entrance side of the housing 11.

[0039] Specifically, the housing 10 has a medium inlet, which can be directly disposed on the first panel 12 in embodiments including the first panel 12; in embodiments where the housing 11 includes a bottom plate, the medium inlet can be disposed on the bottom plate. The medium inlet can be formed by providing one, two, or more through holes on the first panel 12 or the bottom plate, or by providing a perforated structure, or by other means, to allow liquid cooling medium to flow into the housing 10.

[0040] In a preferred embodiment, a first shielding structure 14 is provided in the region of the housing 10 on the media inlet side corresponding to the first gap 30. The first shielding structure 14 completely shields the first gap 30 on the media inlet side, such as... Figure 2 and Figure 3As shown. By adding the first shielding structure 14, the liquid cooling medium can be blocked from entering the first gap 30 from the medium inlet, thereby further reducing the flow of the liquid cooling medium between the first fin 212 and the first side plate 111 and the third side plate 113, and better improving the heat dissipation efficiency of the liquid cooling medium.

[0041] In an embodiment where the housing 11 has a through-type structure and the casing also includes a first panel, the first panel 12 covers the opening of the housing near the medium inlet. The medium inlet of the casing 10 and the first shielding structure 14 are formed on the first panel 12. That is, the medium inlet is provided on the first panel 12, and it is a closed structure at the corresponding first gap 30, forming the first shielding structure 14. In this way, both the housing 11 and the first panel 12 are easy to form.

[0042] Continue to refer to Figure 3 The computing board 22 includes a substrate 221 and multiple computing chips 222. The substrate 221 has two opposing sides, namely a first side and a second side. A chip area 2211 may be provided on the first side, and multiple computing chips 222 are arranged in multiple rows in the second direction Y. Each row of computing chips 222 is spaced apart in the chip area 2211 along the first direction X. In some embodiments, chip areas 2211 may be provided on the first side and the second side respectively, and multiple computing chips 222 are distributed in each chip area 2211. A distance is left between the chip area 2211 and both ends of the substrate 221 in the second direction Y. After assembly with the first heat sink 21, a distance is left between the chip area 2211 and the outermost first fin 212. In this way, each computing chip 222 basically corresponds to the middle area of ​​the first heat-conducting plate 211, and the heat on it will be more concentrated on the first fin 212 in the middle area of ​​the first heat-conducting plate 211.

[0043] In a preferred embodiment, on the media inlet side, in the computing power heat dissipation module 20, at least the first fin 212 is attached to the first panel 12. This effectively blocks the liquid cooling medium from entering the first gap 30, allowing the liquid cooling medium to flow more concentratedly through the first fin 212 in the middle region, thereby improving the heat exchange efficiency between the liquid cooling medium and the first heat sink 21. Specifically, only the first fin 212 may be attached to the first panel 12. In embodiments including the second fin 232, it is preferable that both the first fin 212 and the second fin 232 are attached to the first panel 12 to better block the liquid cooling medium from flowing through the first gap 30 and the second gap 40 (detailed below). More preferably, the entire computing power heat dissipation module is attached to the first panel 12. Especially in embodiments where the first panel 12 is provided with a third shielding structure (detailed below), the flow of liquid cooling medium between the first heat-conducting plate and the second heat-conducting plate can be minimized, preventing erosion of the thermal interface material on the computing chip 222.

[0044] In embodiments where the housing 11 is not fully through, the first shielding structure 14 is formed by bending the first side plate 111 and the third side plate 113. Specifically, it can be formed by bending sheet metal, or by forming the first shielding structure 14 by means of mold or mechanical cutting.

[0045] In embodiments where the computing chips 222 are distributed on the same side of the substrate 221, the first heat-conducting plate 211 can be attached to the computing chips 222, especially in embodiments where the computing heat dissipation module 20 only includes the first heat sink 21; it can also be attached to the side of the substrate 221 where the computing chips 222 are not disposed.

[0046] Continue to refer to Figures 2-5 To further improve the heat dissipation efficiency and ensure the reliability of the computing board, the computing power heat dissipation module 20 also includes a second heat sink 23. The second heat sink 23 includes a second heat-conducting plate 231 and multiple second fins 232 disposed on the same side of the second heat-conducting plate 231. That is, both the first heat sink 21 and the second heat sink 23 include a heat-conducting plate and multiple fins, and the multiple fins on the same heat sink are arranged at intervals along the second direction Y on the same side of the corresponding heat-conducting plate. The fins are parallel to each other, that is, the first fins 212 and the second fins 232 are parallel to each other. The tooth pitch between two adjacent second fins 232 on the second heat sink 23 can be equal to or unequal to the tooth pitch between two adjacent first fins 212 on the first heat sink 21. Preferably, the tooth pitches are equal.

[0047] In the embodiment including the second heat sink 23, the first heat-conducting plate 211 and the second heat-conducting plate 231 are respectively attached to both sides of the computing board 22. Furthermore, the first side plate 111 and the third side plate 113 each form a second gap 40 with the nearest second fin 232. The dimension D2 of the second gap 40 in the opposite direction is less than or equal to twice the tooth pitch. That is, the dimensions of both the first gap 30 and the second gap 40 in the second direction Y are less than or equal to twice the tooth pitch, thereby reducing the amount of liquid cooling medium flowing through the second gap 40 and allowing more liquid cooling medium to flow through the second fin 232 in the middle region, thus improving the heat exchange efficiency between the liquid cooling medium and the second heat sink 23.

[0048] In an embodiment where a computing chip 222 is disposed on one side of the substrate 221, the first heat-conducting plate 211 is bonded to the other side of the substrate 221, i.e., bonded to the side of the substrate where the computing chip 22 is disposed. The second heat-conducting plate 231 is bonded to the computing chip 222, i.e., bonded to the side of the computing plate 22 where the computing chip 222 is located. In this embodiment, preferably, the height of the second fin 232 is greater than the height of the first fin 212, i.e., the height of the second fin 232 extending out of the second heat-conducting plate 231 is greater than the height of the first fin 212 extending out of the first heat-conducting plate 211. More preferably, the ratio of the height of the second fin 232 to the height of the first fin 212 is greater than 2:1, such as 2:1, 2.2:1, 2.5:1, 2.8:1, or 3:1, etc.; more preferably, the ratio is 2:1.

[0049] Furthermore, in the relative direction (i.e., the second direction Y) between the first side plate 111 and the third side plate 113, the size of the second gap 40 is smaller than the size of the first gap 30, such as... Figure 2 and Figure 3 As shown, the second fin 232 located on the first side plate 111 is closer to the first side plate 111 than the first fin 212, and the second fin 232 located on the third side plate 113 is closer to the third side plate 113 than the first fin 212.

[0050] Using the above method, most of the heat from the computing chip 222 is conducted to the second fins 232 through the second heat-conducting plate 231, and a small amount of heat is conducted to the first fins 212 through the first heat-conducting plate 211. This arrangement maximizes the distribution of the second fins 232 on the side of the computing chip 222, and increases their height, thereby maximizing the heat dissipation area of ​​the second heat sink 23. This allows for more efficient heat dissipation of the computing chip 222 during operation, while also reducing the flow of liquid cooling medium through the second gap 40. Furthermore, by simultaneously setting two heat sinks, even if the thermal interface material between the computing chip 222 and the second heat-conducting plate 231 is washed away, reducing the heat transfer efficiency between the second heat sink 23 and the computing chip 222, the heat can still be quickly transferred away through the first heat sink 21, thus improving the reliability and stability of heat transfer in the computing chip 222.

[0051] In the embodiment with the second heat sink 23, the first groove wall 151 of the mounting guide groove 115 is located between the first heat-conducting plate 211 and the second heat-conducting plate 231, so as to better reduce the size of the first heat sink 21, the computing board 22 and the second heat sink 23 in the stacking direction, so as to reduce the space occupation when the entire server is immersed in the liquid cooling medium, and enable the entire system to simultaneously achieve heat exchange for more servers, thereby improving the heat dissipation efficiency of the immersion system.

[0052] To further increase the number of second fins 232 and improve heat dissipation, on the side where the first side plate 111 is located, the outermost second fin 232 (i.e., the second fin 232 closest to the first side plate 111) is closer to the first side plate 111 than the opening of the mounting guide groove 115; similarly, on the side where the third side plate 113 is located, the outermost second fin 232 (i.e., the second fin 232 closest to the third side plate 113) is closer to the third side plate 113 than the opening of the mounting guide groove 115 on that side. That is, in the projection along the third direction Z perpendicular to the first direction X and the second direction Y, the second fin 232 has an overlapping area with the mounting guide groove 115 on the first side plate 111 side and the mounting guide groove 115 on the third side plate 113 side. Furthermore, a second shielding structure 15 is provided in the area of ​​the casing 10 on the medium inlet side corresponding to the second gap 40, such as... Figure 2 and Figure 3 As shown, the second shielding structure 15 completely blocks the second gap 40 on the medium inlet side. By adding the second shielding structure 15, the liquid cooling medium can be blocked from entering the second gap 40 from the medium inlet, thereby further reducing the flow of the liquid cooling medium between the second fin 232 and the first side plate 111 and the third side plate 113, and better improving the heat dissipation efficiency of the liquid cooling medium. The second shielding structure 15 can be set in the same way as the first shielding structure 14, such as being formed by bending the first side plate 111 and the third side plate 113 directly, or being formed directly on the first panel 12, as detailed in the description of the first shielding structure 14.

[0053] In one embodiment, when projected in a direction perpendicular to the first heat-conducting plate 211 (i.e., a third direction Z perpendicular to the first direction X and the second direction Y), the projections of the first shielding structure 14 and the second shielding structure 15 are both located outside the projection of the area where the computing chip 222 of the computing board 22 is located, that is, outside the projection of the chip area 2211. In other words, neither the first shielding structure 14 nor the second shielding structure 15 extends onto the area corresponding to the chip area 2211 on the first panel 12. Furthermore, in the relative direction (i.e., the second direction Y) between the first side plate 111 and the third side plate 113, the distance between the first shielding structure 14 and the second shielding structure 15 and the area where the computing chip is located (i.e., the chip area 2211) is greater than or equal to twice the tooth pitch. By adopting this configuration, sufficient liquid cooling medium can be ensured to flow between the first fins 212 and between the second fins 232, while also better preventing excessive flow of liquid cooling medium through the gaps between the first radiator 21, the second radiator 23 and the casing (the aforementioned first gap 30 and second gap 40). At the same time, the extension of the first shielding structure 14 and the second shielding structure 15 can also increase the structural strength of the entire first panel 12 to avoid its impact deformation on the flow of liquid cooling medium.

[0054] In one embodiment, the first side plate 111 and the third side plate are provided with two sets of mounting guide grooves 115. That is, two mounting guide grooves 115 with opposite openings form a set for inserting the same first heat-conducting plate 211. The housing 11 is provided with two sets of mounting guide grooves 115. The first side plate 111 and the third side plate 113 are each provided with two mounting guide grooves 115. Each mounting guide groove 115 of the first side plate 111 and one mounting guide groove 115 on the third side plate 113 with opposite openings form a set. Specifically, the two sets of mounting guide grooves 115 can be located in areas close to the second side plate 112 and the fourth side plate 114, respectively.

[0055] Accordingly, two computing power heat dissipation modules 20 are provided, that is, the immersion supercomputing device includes two computing power heat dissipation modules 20. The two computing power heat dissipation modules 20 are respectively installed in two sets of mounting guide slots 115 through their respective first heat conduction plates 211, that is, each is installed in a corresponding mounting guide slot 115. The first fins 212 of the computing power heat dissipation module 20 closer to the second side plate 112 face the second side plate 112, and the second fins 232 of the other computing power heat dissipation module 20 face the fourth side plate 114, as shown. Figure 2 and Figure 3 As shown.

[0056] Furthermore, in the computing power heat dissipation module 20 near the second side plate 112, the distance between the first fin 212 located in the middle region and the second side plate 112 is less than or equal to one tooth pitch; in the computing power heat dissipation module 20 near the fourth side plate 114, the distance between the second fin 232 located in the middle region and the fourth side plate 114 is less than or equal to one tooth pitch. That is to say, the distance between each computing power heat dissipation module 20 and the casing 10 in the third direction Z is basically less than or equal to one tooth pitch.

[0057] By adopting the above settings, the flow of liquid cooling medium in areas other than the fins can be further reduced, thereby concentrating the flow between each fin, improving the overall heat dissipation efficiency of the server, making the temperature difference between the computing chips 222 on the medium inlet side and the medium outlet side smaller, and ensuring the optimal computing performance of each computing chip 222.

[0058] In some embodiments, the ratio of the thickness of the first fin 212 and the second fin 232 to the tooth pitch is less than or equal to 1 / 4, meaning the thickness of each fin is one-quarter of the tooth pitch, or even less. This method maximizes the number of fins within a limited space, increasing the heat dissipation area of ​​each radiator and maximizing the flow area, thereby further improving heat dissipation efficiency. Specifically, in one embodiment, the thickness of the first fin 212 and the second fin 232 is 0.3mm to 1.5mm, such as 0.3mm, 0.5mm, 0.6mm, 0.7mm, 0.9mm, 1.0mm, 1.2mm, or 1.5mm. More preferably, in the same radiator, the thickness of the outermost fin in the arrangement direction is between 0.6mm and 1.5mm, and the thickness of the remaining fins is between 0.3mm and 0.8mm. The tooth pitch is 1.4mm~2.8mm. In each heat sink, the tooth pitch is 1.4mm~2.8mm, such as 1.4mm, 1.7mm, 2.0mm, 2.5mm, or 2.8mm. Preferably, except for the outermost fin, the thickness of the first fin 212 and the second fin 232 is 0.6mm, and the tooth pitch is 2.6mm or 2.8mm. More preferably, the tooth pitch of the first heat sink 21 is 2.6mm, and the tooth pitch of the second heat sink 23 is 2.8mm. Using this range ensures that each fin has sufficient strength to avoid deformation caused by the impact of the liquid cooling medium flow, while maximizing the flow area of ​​each heat sink, enabling each computing chip 222 in each computing power cooling module 20 to achieve good temperature uniformity.

[0059] To better verify that reducing the gap 14 can improve the heat dissipation effect, in a specific embodiment, the size of the gap (i.e., the first gap 30 and the second gap 40) 14 between the first side plate 111 and the third side plate 113 and the nearest fin is set to be basically 1 times the tooth pitch, such as 2.8mm (that is, both the first gap 30 and the second gap 40 are set to 2.8mm), and the embodiment in which the size of these two gaps is set to 12.5mm (greater than twice the width of the medium flow channel) is tested respectively. The test results show that by reducing the size of the gap 14 from 12.5mm to 2.8mm, the temperature gain of the computing chip 222 can reach 5.2℃, that is, the temperature on the computing chip 222 can be reduced by an average of about 5℃, and the heat dissipation efficiency of the entire computing board 222 can be improved by more than 13%.

[0060] At the four corners of the housing 11, mounting holes are provided protruding inwards to facilitate the installation of the first panel 12 and the second panel 13. Correspondingly, the second fin 232 and the first fin 212 at these locations are smaller in height than the second fin 232 and the first fin 212 in other areas.

[0061] In both the first heat-conducting plate 211 and the second heat-conducting plate 231, in the embodiment where they are attached to the computing power chip 222, the first heat-conducting plate 211 and the second heat-conducting plate 231 can be provided with multiple heat-conducting strips on the side facing the computing power plate 22. Each heat-conducting strip extends along the first direction X, and the multiple heat-conducting strips are correspondingly arranged with multiple rows of computing power chips 222. In the embodiment where the first heat-conducting plate 211 or the second heat-conducting plate 231 is attached to the computing power chip 222, each heat-conducting strip is attached to one row of computing power chips 222 respectively.

[0062] To further improve the heat dissipation effect of the computing chip 222, the computing power heat dissipation module 20 also includes a vapor chamber and heat pipes; the first side of the vapor chamber is provided with a strip groove, and the heat pipes are disposed in the strip groove. At least a vapor chamber and heat pipes are disposed between the second heat-conducting plate 231 and the computing chip 222, and the second side of the vapor chamber is in contact with the computing chip 222, and heat pipes are arranged at each of the computing chips 222, that is, multiple strip grooves are respectively disposed corresponding to multiple rows of computing chips 222.

[0063] In embodiments where the first heat-conducting plate 211 or the second heat-conducting plate 231 is bonded to the computing chip 222, both the first heat-conducting plate 211 and the second heat-conducting plate 231 are bonded to the computing chip 222 via a thermal interface material (TIM). The thermal interface material can be silicone grease, gel, a thermal pad, liquid metal, etc., preferably a curable liquid thermally conductive material. In embodiments where a thermally conductive strip is provided, the thermal interface material is disposed between the thermally conductive strip and the computing chip 222.

[0064] In some embodiments, the immersion supercomputing device also includes a power module 50, which is installed inside the housing 10. In embodiments with two computing power heat dissipation modules 20, the power module 50 is located between the two computing power heat dissipation modules 20.

[0065] Whether it is the first heat sink 21 or the second heat sink 23, the air-facing surface of each fin forms a flow-guiding slope (such as the first flow-guiding slope 213 and the second flow-guiding slope 233). The end of the flow-guiding slope away from the first panel 12 is inclined in the direction away from the heat-conducting plate, and the inclination angle is 10°~42°. That is, the end face of each fin facing the first panel 12 is an inclined plane. The angle α between the inclined plane and the heat-conducting plate is 10°~42°. The end of the flow-guiding slope close to the first panel 12 is the starting end, and the other end is the ending end. Furthermore, the ending end extends into the side where the computing chip 222 is located in the first direction X. The projection of the flow-guiding slope in the thickness direction of the heat-conducting plate (such as the first heat-conducting plate and the second heat-conducting plate) overlaps with the chip area 2211. By employing the aforementioned flow-guiding slope, sufficient mixing space can be formed between the flow-guiding slope and the first panel 12, allowing the liquid cooling medium to be distributed more evenly before entering each medium channel. Consequently, under the guidance of the flow-guiding slope, it enters each medium channel more evenly. Furthermore, by setting the specific tilt angle of the flow-guiding slope, the temperature of each computing chip 222 in the through direction can be balanced, further improving the temperature consistency of each computing chip 222. In fact, the temperature difference between the computing chips at both ends in the first direction can be less than 8°C. The temperature uniformity performance can be improved by more than 32% compared to the solution without the flow-guiding slope. At the same time, the liquid pressure drop of the liquid cooling medium is reduced by 18%, or in other words, the energy consumption requirement of the liquid cooling pump of the entire device is reduced by 18%. This makes it easier to achieve temperature control of each computing board 22, thereby enabling the entire computing board 22 to maintain optimal working condition continuously.

[0066] Multiple fins are arranged at intervals along a direction perpendicular to the through-flow direction, forming a medium flow channel between adjacent fins. In the fin arrangement direction (i.e., the second direction Y), the ratio of the sum of the widths of all medium flow channels to the width of the radiator is between (0.65~0.85):1. That is, if the sum of the dimensions of all medium flow channels in the second direction Y is M, and the width of the radiator is L, then the ratio of M to L is 0.65~0.85, such as 0.65, 0.68, 0.7, 0.73, 0.75, 0.76, 0.77, 0.78, 0.8, 0.83, or 0.85. Radiators using this method, in conjunction with the aforementioned flow hole ratio and guide slope settings, maximize the fin area while ensuring a more balanced flow rate into each medium flow channel. Furthermore, each medium flow channel can be continuously filled with liquid cooling medium, thereby continuously, efficiently, and stably improving heat dissipation efficiency and effect.

[0067] Furthermore, in the second heat sink 23, the tilt angle α is between 20° and 28°, such as 20°, 22°, 23°, 24°, 25°, 26°, or 28°, preferably 23°. Considering that the heat of the computing chip 222 is mainly dissipated from the side facing the second heat sink 23, by setting the angle of the second flow guide slope 233 of the second heat sink 23 to be slightly larger, the space between the first panel and the flow guide slope is reduced. This allows the liquid cooling medium to enter the medium flow channel faster at the lowest possible temperature to dissipate heat from the computing chip 222, and also allows the liquid cooling medium to flow through more computing chips 222 at the lowest possible temperature, thereby further improving the heat dissipation effect.

[0068] In some embodiments, the ratio of the height to the thickness of the fins of the second heat sink 23 is between (30~56):1, that is, the height of the second fin 232 is 30 to 56 times the thickness, such as 30, 33, 35, 38, 40, 43, 45, 50, 53, or 56 times the thickness, etc., preferably 38 to 42 times the thickness. Using this ratio range, the heat dissipation area of ​​the second fin can be increased by increasing the fin height, while avoiding the impact of excessively thick fins on the flow of the liquid cooling medium, and the strength of the second fin 232 can be increased as much as possible. This improves the overall heat dissipation effect while ensuring the stability of the liquid cooling medium flow in each medium flow channel.

[0069] The ratio of the height of the fins of the second heat sink 23 to the width of a single medium flow channel is between (8~16):1, that is, the height of the second fin 232 is 35~56 times the width of the medium flow channel, such as 8, 10, 10.5, 11, 13 or 16 times the width of the medium flow channel, etc., preferably 10~11 times the width of the medium flow channel. Through this ratio range, the flow area in each medium flow channel can be better controlled, and it can be ensured that each medium flow channel can be continuously filled with liquid cooling medium, so that the liquid cooling medium is more rationally distributed, achieving the maximum efficiency of heat dissipation for the computing chip 222.

[0070] Furthermore, by defining the aforementioned ratio range between the second fin 232 and its thickness, and the ratio range between the second radiator 23 and the width of the medium flow channel, it is possible to better balance the flow rate and flow stability in each medium flow channel while ensuring that the fin has the largest possible heat dissipation area and the strength to resist the impact of the medium flow, and to achieve the optimal heat dissipation effect in conjunction with the fin.

[0071] In some embodiments, the plurality of computing chips 222 are arranged in two groups in the through direction, and the arrangement density of computing chips 222 in the group closer to the entrance is greater than the arrangement density of computing chips 222 in the other group.

[0072] The terminating end of the second flow guiding slope 233 extends to the 1st to 5th computing chips. That is, the second flow guiding slope 233 projects onto the slope of the computing board 22 along the third direction Z (i.e., the thickness direction of the heat-conducting plate), covering the 1st to 5th computing chips 222 on the computing board 22 near the first panel 12. In this way, the temperature of each computing chip 222 in the through direction can be better balanced. Preferably, when the tilt angle of the second flow guiding slope 233 is between 20° and 28°, the projection along the thickness direction of the heat-conducting plate, the projection of the flow guiding slope of the second heat sink 23 (i.e., the second flow guiding slope 233) overlaps with at least 3 to 5 computing chips 222. Specifically, it overlaps with the 3rd to 5th computing chips closest to the first panel 12 in the computing chip group with high density (i.e., the computing chip group near the first panel 12). This configuration aims to conduct heat from as many computing chips 222 as possible away when the liquid cooling medium is at a low temperature, thereby further achieving temperature uniformity of the computing chips 222 along the through-path. The tilt angle is negatively correlated with the number of computing chips N covered; that is, the larger the tilt angle, the smaller the number of computing chips N covered. For example, at a tilt angle of 28°, 3 computing chips 222 are covered; at 23°, 4; at 20°, 5; and at 42°, only one computing chip 222 may be covered. It should be noted that covering the computing chips 222 can mean covering the entire computing chip 222 or only a portion of it. Covering the entire computing chip 222 is preferred.

[0073] In some embodiments, the fin height of the second radiator 23 is greater than the fin height of the first radiator 21 to increase the heat dissipation effect of the second radiator 23. In a preferred embodiment, the ratio of the fin height of the second radiator 23 to the fin height of the first radiator 21 is between (1.5~2.5):1, that is, the height of the second fin 232 is 1.5 to 2.5 times the height of the first fin 212, such as 1.5 times, 1.8 times, 2 times, 2.2 times or 2.5 times, preferably 2 times.

[0074] In other embodiments, the width of the medium flow channel of the first heat sink 21 (i.e., the tooth pitch of the first heat sink 21) is less than or equal to the width of the medium flow channel of the second heat sink 23 (i.e., the tooth pitch of the second heat sink 23).

[0075] By setting up the above, the heat dissipation area of ​​the second heat sink 23 and the flow area of ​​the liquid cooling medium are increased within a limited space. At the same time, the heat conduction on both sides of the computing chip 222 can be balanced, so that the computing chip 222 can better maintain the temperature balance of its various parts and take into account the temperature consistency of each computing chip 222.

[0076] The tilt angle α of the first heat sink can be between 10° and 26°, such as 10°, 12°, 12.5°, 15°, 18°, 20°, 22°, 25°, 25.5°, or 26°, preferably between 12.5° and 15.5°, and more preferably 14.6°. By setting the angle of the first flow guiding slope 213 of the first heat sink 21 to be slightly smaller, a larger mixing space and a longer flow guiding path are provided. At the same time, the heat dissipation area of ​​the first fin 212 at the inlet end is appropriately reduced, so that the liquid cooling medium flows through the downstream computing chip 222 at the lowest possible temperature. Thus, in conjunction with the second heat sink 23, the temperature consistency of the upstream and downstream computing chips 222 can be further improved.

[0077] In some embodiments, in order to better enable the computing board 22 to achieve consistent heat dissipation effect and temperature of each computing chip 222 through the heat dissipation cooperation of the upper and lower sides, it is preferable that the termination ends of the guide slopes of the first heat sink 21 and the second heat sink 23 are aligned, that is, in the through direction, the termination ends of the first guide slope 213 and the second guide slope 233 are at the same position.

[0078] In the first heat sink 21, the ratio of fin height to fin thickness is between (15~29):1, meaning the height of the first fin 212 is 15~29 times its thickness, such as 15, 18, 20, 23, 25, 27, or 29 times its thickness, etc., preferably 18~22 times its thickness. Using this ratio range allows for increased heat dissipation area of ​​the first fin 212 by increasing fin height, avoiding the impact of excessively thick fins on the flow of the liquid cooling medium, while maximizing the strength of the first fin 212. This improves the overall heat dissipation effect while ensuring the stability of the liquid cooling medium flow in each medium flow channel.

[0079] The ratio of the height of the fins of the first heat sink 21 to the width of a single medium flow channel is between (4~9):1, that is, the height of the first fin 212 is 4 to 9 times the width of the medium flow channel, such as 4, 5, 6, 7, 8, or 9 times the width of the medium flow channel, etc., preferably 5 to 7 times the width of the medium flow channel. Through this ratio range, the flow area in each medium flow channel can be better controlled, the liquid cooling medium can be more rationally distributed, and it can be ensured that each medium flow channel can be continuously filled with liquid cooling medium, thereby achieving the maximum efficient heat dissipation effect for the computing chip 222.

[0080] Furthermore, by defining the ratio range of the first fin 212 to its thickness and the ratio range of the medium flow channel width of the first heat sink 21, it is possible to better balance the flow rate and flow stability in each medium flow channel while ensuring that the fin has the largest possible heat dissipation area and the strength to resist the impact of the medium flow, and to achieve the optimal heat dissipation effect in conjunction with the fin.

[0081] The housing 10 also has a media outlet, which can be located on the second panel 13 or in the area of ​​the housing 11 near the outlet side, such as on any one or more of the first side panel 111, the second side panel 112, the third side panel 113, and the fourth side panel 114. In a preferred embodiment, a media outlet is provided on both the second panel 13 and the housing 11. Furthermore, to facilitate the movement of the entire immersion computing system, a handle is also provided on the second panel 13.

[0082] The present invention also provides an immersion computing system, including a liquid cooling tank and an immersion supercomputing device as described in any of the above embodiments. The immersion supercomputing device is disposed in the liquid cooling tank, and the medium inlet of the housing 10 faces the bottom of the liquid cooling tank. The immersion computing system also includes a liquid cooling pump and a cooling assembly. The liquid cooling pump is located outside the immersion supercomputing device and is used to drive the liquid in the liquid cooling tank to circulate along the direction of the medium inlet, the inner cavity of the housing 10, the medium outlet, the cooling assembly, and the medium inlet, so that the heat on the computing chips 222 is conducted out of the housing 10 during the circulation of the liquid cooling medium, so that the temperature of each computing chip 222 is basically the same and maintained within the optimal operating temperature range.

[0083] It should be noted that the bonding mentioned in the text is not limited to the two parts being tightly bonded to achieve a seal. It can also mean that there is basically no gap between the two parts, or the gap is very small, which can be the gap caused by permissible manufacturing and assembly errors.

[0084] It will be understood by those skilled in the art that the above-described preferred solutions can be freely combined and superimposed without conflict. It should be understood that the above-described embodiments are merely exemplary and not restrictive. Various obvious or equivalent modifications or substitutions made by those skilled in the art regarding the above details without departing from the basic principles of the invention will be included within the scope of the claims of this invention.

Claims

1. An immersion supercomputing device, comprising a casing and a computing power heat dissipation module, characterized in that, The housing includes a box body, which includes a first side plate, a second side plate, a third side plate and a fourth side plate connected end to end in sequence. The first side plate and the third side plate are opposite to each other and are respectively provided with mounting guide grooves with opposite openings. The computing power heat dissipation module includes a first heat sink and a computing power plate. The first heat sink includes a first heat-conducting plate and a plurality of spaced first fins disposed on the same side of the first heat-conducting plate. The first heat-conducting plate is attached to the computing power plate. The two ends of the first heat-conducting plate are inserted into the mounting guide groove in the opposite direction of the first side plate and the third side plate. The first side plate and the third side plate each form a first gap with the nearest first fin, and the size of the first gap in the relative direction is less than or equal to twice the tooth pitch; wherein, the tooth pitch refers to the distance between two adjacent first fins.

2. The immersion supercomputing device according to claim 1, characterized in that, The housing has a medium inlet, and a first shielding structure is provided on the housing in the area corresponding to the first gap on the medium inlet side.

3. The immersion supercomputing device according to claim 2, characterized in that, The housing is a through structure; the casing also includes a first panel, which covers the opening of the housing near the medium inlet, and the medium inlet and the first shielding structure are formed on the first panel.

4. The immersion supercomputing device according to claim 3, characterized in that, In the computing power heat dissipation module, at least the first fin is attached to the first panel.

5. The immersion supercomputing device according to claim 2, characterized in that, The first shielding structure is formed by bending the first side plate and the third side plate.

6. The immersion supercomputing device according to claim 1, characterized in that, The computing power heat dissipation module further includes a second heat sink, which includes a second heat-conducting plate and a plurality of second fins disposed on the same side of the second heat-conducting plate; the first heat-conducting plate and the second heat-conducting plate are respectively attached to two sides of the computing power plate; The first side plate and the third side plate each form a second gap with their nearest second fin, and the size of the second gap in the relative direction is less than or equal to twice the tooth pitch.

7. The immersion supercomputing device according to claim 6, characterized in that, The computing board includes a substrate and a computing chip disposed on one side of the substrate; the first heat-conducting plate is attached to the other side of the substrate; The second heat-conducting plate is attached to the computing chip; the height of the second fin is greater than the height of the first fin; In the relative direction, the size of the second gap is smaller than the size of the first gap.

8. The immersion supercomputing device according to claim 7, characterized in that, On the side where the first side plate is located, the outermost second fin is closer to the first side plate than the opening of the mounting guide groove; On the side where the third side plate is located, the outermost second fin is closer to the third side plate than the opening of the mounting guide groove.

9. The immersion supercomputing device according to claim 6, characterized in that, The housing has a medium inlet, and a second shielding structure is provided on the housing in the area corresponding to the second gap on the medium inlet side.

10. The immersion supercomputing device according to claim 9, characterized in that, A first shielding structure is provided in the area of ​​the casing corresponding to the first gap on the medium inlet side; when projected in a direction perpendicular to the first heat-conducting plate, the projections of the first shielding structure and the second shielding structure are both located outside the projection of the area where the computing chip is located on the computing power board, and in the relative direction, the distance between the first shielding structure and the second shielding structure and the area where the computing chip is located is greater than or equal to 2 times the tooth pitch.

11. The immersion supercomputing device according to claim 6, characterized in that, The first side plate and the third side plate are provided with two sets of mounting guide slots; there are two computing power heat dissipation modules, which are respectively mounted on the two sets of mounting guide slots through the first heat conduction plate. The first fin of the computing power heat dissipation module closer to the second side plate faces the second side plate, and the distance between the first fin located in at least the middle area and the second side plate is less than or equal to one tooth pitch; the second fin of the other computing power heat dissipation module faces the fourth side plate, and the distance between the second fin located in at least the middle area and the fourth side plate is less than or equal to one tooth pitch.

12. The immersion supercomputing device according to claim 6, characterized in that, The ratio of the thickness of the first fin and the second fin to the tooth pitch is less than or equal to 1 / 4.

13. The immersion supercomputing device according to claim 12, characterized in that, The thickness of the first fin and the second fin 232 is 0.3mm to 1.2mm, and the tooth spacing is 1.4mm to 2.8mm.

14. The immersion supercomputing device according to claim 6, characterized in that, The computing power heat dissipation module also includes a heat spreader and a heat pipe; the first side of the heat spreader is provided with a strip groove, and the heat pipe is disposed in the strip groove; At least one heat spreader is provided between the second heat-conducting plate and the computing chip, and the second side of the heat spreader is in contact with the computing chip, and heat pipes are arranged at the computing chip.

15. An immersive computing system, characterized in that, It includes a liquid cooling tank and an immersion supercomputing device as described in any one of claims 1-14, wherein the immersion supercomputing device is disposed in the liquid cooling tank and the medium inlet of the casing faces the bottom of the liquid cooling tank.

Images