Cooling systems for high performance computers

Abstract

A cooling system for a blade of a high-performance computer thermally couples blade electronics to a computer liquid cooling system via one or more heat lines without requiring a liquid conduit on or liquid-coupled to the blade. In addition, illustrative embodiments allow the blade to be installed in a high-performance computer and engaged with the cooling system without making a liquid connection, and allow the blade to be disengaged from the cooling system and removed from the high-performance computer without breaking the liquid connection.

Description

Technical Field

[0001] The present invention relates to high performance computer systems, and more particularly to cooling systems for high performance computer systems. Background Art

[0002] The electronic components of high-performance computer systems generate a large amount of heat during normal operation, which is well known to adversely affect the performance and reliability of the computer. Therefore, many computers include cooling systems such as radiators, fans in typical home computers, and water-based cooling systems in more powerful home computers or mainframe computers. Summary of the invention

[0003] The illustrative embodiments transfer heat generated by electronics on a blade of a high performance computer system to a thermal reservoir of the computer without requiring liquid exchange between a cooling circuit at the blade electronics and a cooling element.

[0004] In a first embodiment, a cooling system for a blade of a high performance computer is in communication with a heat accumulator and includes a heat pipe coupled to a DIMM on the blade to thermally conduct heat energy from the DIMM, and a dry heat connection conduit having a high thermal conductivity (preferably at least 100 W / (mK)) that thermally couples the heat pipe to a liquid conduit in liquid communication with the heat accumulator.

[0005] In some embodiments, the dry thermal communication conduit includes a transfer bar, and the heat line and transfer bar are configured to enable installation of the blade into a high performance computer such that the transfer bar operably engages a liquid cooling device without making a liquid connection.

[0006] Some embodiments further include a transfer block thermally disposed between the heat line and the transfer rod such that the transfer block is in thermal communication with the heat line and in thermal communication with the transfer rod to transfer thermal energy from the heat line to the transfer rod. The transfer block may be secured to the transfer rod by at least one fastener and / or may include an intermediate heat line.

[0007] In an illustrative embodiment, the transfer rod includes a cooling system interface for physically and thermally coupling the transfer rod to the fluid conduit.

[0008] An embodiment of a computer system having a liquid cooling circuit includes: a first blade having a first electronic component and a first heat pipe thermally connected to the first electronic component, and a first transfer rod thermally connected to the first heat pipe; and a second blade having a second electronic component and a second heat pipe thermally connected to the second electronic component, and a second transfer rod thermally connected to the second heat pipe, in which a cold storage tank is dry-thermally connected to the first transfer rod and the second transfer rod, so that the first heat pipe and the first transfer rod are configured to transfer thermal energy from the first electronic component to the cold storage tank via the liquid cooling circuit, and the second heat pipe and the second transfer rod are configured to transfer thermal energy from the second electronic component to the cold storage tank via the liquid cooling circuit.

[0009] To promote thermal conductivity, the first transfer rod and / or the second transfer rod may be made of copper or aluminum. Furthermore, in some embodiments, the first transfer rod and / or the second transfer rod may include an intermediate heat line.

[0010] Embodiments of a DIMM module for use in a high-performance computer with a heat accumulator include: a first memory circuit; a first heat pipe, the first heat pipe is thermally connected to the first memory circuit to conductively receive heat generated by the first memory circuit; and a transfer block, the transfer block is thermally connected to the first heat pipe to transfer heat from the first memory circuit to the heat accumulator. Some embodiments also include: a second memory circuit, the second memory circuit is separated from the first memory circuit by a gap pad; and a second heat pipe, the second heat pipe is thermally connected to the second memory circuit to conductively receive heat generated by the second memory circuit. The second heat pipe is thermally connected to the transfer block to transfer heat from the second memory circuit to the heat accumulator. In some embodiments, the first memory circuit and the second memory circuit are arranged between the first heat pipe and the second heat pipe, and in some embodiments, the transfer block includes a hole configured to allow a fastener to secure the DIMM module to the blade. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The foregoing features of the embodiments will be more readily understood by referring to the following detailed description made with reference to the accompanying drawings, in which:

[0012] Figure 1 schematically illustrates a logical view of an exemplary high performance computing system;

[0013] Figure 2 schematically illustrates a physical view of an embodiment of a high performance computing system;

[0014] Figure 3 An embodiment of a single blade chassis is schematically illustrated;

[0015] Figure 4An embodiment of a high performance computer having a liquid cooling system is schematically illustrated;

[0016] Figure 5A and Figure 5B schematically illustrates a blade cooling component;

[0017] Figure 5C and Figure 5D An embodiment of a blade for a high performance computer is schematically illustrated;

[0018] Figure 5E Schematically illustrates an embodiment in which a blade of a high-performance computer is arranged to face another blade of a high-performance computer system, wherein electronic components are interleaved with each other;

[0019] 6A to 6F Schematically illustrates an embodiment of a DIMM;

[0020] 7A to 7E An embodiment of a transfer rod is schematically illustrated;

[0021] FIG. 8A to FIG. 8B An embodiment of a cooling circuit is schematically illustrated. DETAILED DESCRIPTION

[0022] A cooling system for a blade of a high performance computer ("HPC") thermally couples the blade electronics to the liquid cooling system of the computer via dry, high thermal conductivity conduits without requiring liquid coupling at the blade. In illustrative embodiments, the dry conduits form a path of least thermal resistance for heat energy to flow from the blade electronics to the liquid cooling system of the computer, which desirably mitigates or prevents heat energy from flowing from the blade electronics to other components of the blade. In addition, some embodiments allow the blade to be installed into the HPC and operably engaged with the liquid cooling system of the computer without having to make a liquid connection between the blade and the liquid cooling circuit. Some embodiments also allow the blade to be removed from the HPC without having to disconnect the liquid connection between the blade and the liquid cooling circuit.

[0023] Some embodiments achieve 99% liquid cooling of high performance computer systems. This facilitates greater node density, makes high performance computer systems more efficient, and allows users of high performance computer systems greater flexibility in meeting facility requirements. This also reduces audible noise levels in the data center by reducing the need for fans.

[0024] For example, the illustrative embodiments couple one or more heat pipes to the blade electronics (e.g., DIMMs) and couple each heat pipe to the liquid cooling system of the computer via a transfer rod. The blades, heat pipes, and transfer rods can be removed from the HPC without having to disconnect the liquid from the liquid cooling system of the computer. Some embodiments provide cooling sufficient to allow two blades to be placed in close proximity to each other, even to the extent that the electronics of the two blades are interdigitated.

[0025] System Architecture

[0026] Figure 1 Schematically illustrates a logical view of an exemplary high performance computing system 100 that can be used with illustrative embodiments of the present invention. Specifically, as known to those skilled in the art, a "high performance computing system" or "HPC system" is a computing system having multiple modular computing resources that are tightly coupled using hardware interconnects so that processors can directly access remote data using a common memory address space.

[0027] HPC system 100 includes a plurality of logical computing partitions 120, 130, 140, 150, 160, 170 for providing computing resources, and a system console 110 for managing the plurality of partitions 120 to 170. A "computing partition" (or "partition") in an HPC system is a management allocation of computing resources that runs a single operating system instance and has a common memory address space. Partitions 120 to 170 can communicate with system console 110 using a logical communication network 180. System users such as scientists or engineers who wish to perform calculations can request computing resources from a system operator who uses system console 110 to allocate and manage these resources. The allocation of computing resources to partitions is described below. HPC system 100 can have any number of computing partitions that are administratively allocated as described in more detail below, and typically has only one partition that contains all available computing resources. Therefore, this figure should not be considered to limit the scope of the present invention.

[0028] Each computing partition, such as partition 160, can be logically viewed as a single computing device similar to a desktop computer. Thus, partition 160 can execute software including a single operating system ("OS") instance 191 (which are used together in the art) using a basic input / output system ("BIOS") 192, and application software 193 for one or more system users.

[0029] Therefore, if Figure 1As also shown in FIG. 1 , a computing partition has various hardware assigned to it by a system operator, including one or more processors 194, volatile memory 195, non-volatile storage 196, and input and output (“I / O”) devices 197 (e.g., network ports, video display devices, keyboards, etc.). However, in the example Figure 1 In an HPC system of an embodiment of the present invention, each computing partition has more processing power and memory than a typical desktop computer. The OS software may include, for example, Microsoft Corporation of Redmond, Washington. operating system, or Linux operating system. In addition, although the BIOS can be provided as firmware by hardware manufacturers such as Intel Corporation of Santa Clara, California, the BIOS is usually customized according to the needs of HPC system designers to support high performance computing, as described in more detail below.

[0030] As part of its system management role, the system console 110 acts as an interface between the computing capabilities of the compute partitions 120 to 170 and the system operator or other computing systems. To this end, the system console 110 issues commands to the HPC system hardware and software on behalf of the system operator that allow, among other things, the following operations: 1) booting up the hardware, 2) partitioning the system computing resources into compute partitions, 3) initializing the partitions, 4) monitoring the health of each partition and any hardware or software errors generated therein, 5) distributing operating systems and application software to the partitions, 6) causing the operating systems and software to execute, 7) backing up the state of the partitions or the software therein, 8) shutting down application software, and 9) shutting down a compute partition or the entire HPC system 100. These specific functions are described in more detail below in the section entitled "System Operations."

[0031] Figure 2 Schematically shows the Figure 1 A physical view of a high performance computing system 100 of an embodiment of the present invention includes Figure 1 The hardware of the HPC system 100 is surrounded by dashed lines. The HPC system 100 is connected to a customer data network 210 to facilitate customer access.

[0032] HPC system 100 includes a system management node ("SMN") 220 that performs the functions of system console 110. Management node 220 may be implemented as a desktop computer, server computer, or other similar computing device provided by a customer or HPC system designer, and includes software required to control HPC system 100 (i.e., system console software).

[0033] HPC system 100 may be accessed using data network 210, which may include any data network known in the art, such as a customer local area network ("LAN"), a virtual private network ("VPN"), the Internet, etc., or a combination of these networks. Any of these networks may allow multiple users to remotely and / or simultaneously access HPC system resources. For example, client computer 230 may use a network known in the art, such as a Tools such as remote desktop services or UNIX secure shells access management node 220 by remote login. If the customer has such a preference, access to HPC system 100 can be provided to remote computer 240. Remote computer 240 can access the HPC system by logging into management node 220 as just described or using a gateway or proxy system known to those skilled in the art.

[0034] The hardware computing resources of the HPC system 100 (e.g., Figure 1 The processors, memory, non-volatile storage devices, and I / O devices shown in FIG. 1 are managed and allocated to one or more "blade chassis" (such as Figure 2 The blade chassis 252, 254, 256, 258 shown in the figure are provided together. The blade chassis is an electronic device chassis configured to accommodate, power, and provide high-speed data communication between multiple stackable modular electronic circuit boards called "blades". Each blade includes sufficient computing hardware to act as a stand-alone computing server. The modular design of the blade chassis allows the blades to be connected to power and data lines with minimal wiring and vertical space.

[0035] Thus, each blade chassis (e.g., blade chassis 252) has a chassis management controller 260 (also referred to as a "chassis controller" or "CMC") for managing system functions in the blade chassis 252, and a plurality of blades 262, 264, 266 for providing computing resources. Each blade (e.g., blade 262) contributes its hardware computing resources to the collective total resources of the HPC system 100. The system management node 220 uses a chassis controller such as chassis controller 260 to manage the hardware computing resources of the entire HPC system 100, while each chassis controller manages resources only for blades in its blade chassis. The chassis controller 260 is physically and electrically coupled to the blades 262 to 266 inside the blade chassis 252 by means of a local management bus 268, which will be described in more detail below. The hardware in the other blade chassis 254 to 258 is similarly configured.

[0036] The chassis controllers communicate with each other using management connection 270. Management connection 270 may be, for example, a high-speed LAN or other data bus running an Ethernet communication protocol. In contrast, the blades communicate with each other using compute connection 280. To this end, compute connection 280 illustratively has a high-bandwidth, low-latency system interconnect such as NumaLink.

[0037] Chassis controller 260 provides system hardware management functions to the rest of the HPC system. For example, chassis controller 260 may receive a system boot command from SMN 220 and respond by issuing a boot command to each of blades 262 to 266 using local management bus 268. Similarly, chassis controller 260 may receive hardware error data from one or more of blades 262 to 266 and store this information for later analysis in conjunction with error data stored by other chassis controllers. Figure 2 In some of the embodiments shown in , the SMN 220 or client computer 230 is provided with access to a single master chassis controller 260 that processes system management commands to control the HPC system 100 and forwards these commands to other chassis controllers, however, in other embodiments, the SMN 220 is directly coupled to the management connection 270 and issues commands to each chassis controller individually. Those of ordinary skill in the art may recognize variations of these designs that allow the same type of functionality, but for clarity, only these designs are presented.

[0038] The blade chassis 252, its blades 262-266, and the local management bus 268 may be provided as known in the art. However, the chassis controller 260 may be implemented using hardware, firmware, or software provided by the HPC system designer. Each blade provides the HPC system 100 with a certain number of processors, volatile memory, non-volatile storage devices, and I / O devices known in the art of stand-alone computer servers. However, each blade also has hardware, firmware, and / or software to allow these computing resources to be grouped together and viewed together as a computing partition, as described in more detail below in the section entitled "System Operation."

[0039] although Figure 2 An HPC system 100 having four chassis and three blades in each chassis is shown, but it should be understood that these figures do not limit the scope of the invention. An HPC system may have dozens of chassis and hundreds of blades; in fact, HPC systems are often required because they provide a large number of tightly coupled computing resources.

[0040] Figure 3A single blade chassis 252 is schematically shown in greater detail. In this figure, portions not relevant to the immediately following description have been omitted. A chassis controller 260 and its connections to the system management node 220 and to the management connection 270 are shown. The chassis controller 260 may be provided with a chassis data storage device 302 for storing chassis management data. In some embodiments, the chassis data storage device 302 is a volatile random access memory ("RAM"), in which case the SMN 220 can access the data in the chassis data storage device 302 as long as power is applied to the blade chassis 252, even if one or more of the compute partitions fails (e.g., due to an operating system crash) or a blade fails. In other embodiments, the chassis data storage device 302 is a non-volatile storage device such as a hard disk drive ("HDD") or a solid state drive ("SSD"). In these embodiments, the data in the chassis data storage device 302 is accessible after the HPC system has been powered off and restarted.

[0041] For discussion purposes, Figure 3 Relevant portions of a specific implementation of blades 262 and 264 are shown. Blade 262 includes a blade management controller 310 (also referred to as a "blade controller" or "BMC") that performs system management functions at the blade level in a manner similar to the functions performed by the chassis controller at the chassis level. See the section entitled "HPC System Operation" below for more details on the operation of the chassis controller and blade controllers. Blade controller 310 may be implemented as custom hardware designed by an HPC system designer to allow communication with chassis controller 260. In addition, blade controller 310 may have its own RAM 316 to perform its management functions. As shown in Figure 3 As shown in previous figures, chassis controller 260 communicates with the blade controller of each blade using local management bus 268 .

[0042] Blade 262 also includes one or more microprocessors 320, 322 (alternatively referred to as "processors 320 or 322" or generally referred to as "processor 320") connected to RAM 324, 326. Blade 262 may alternatively be configured so that multiple processors can access a common set of RAM on a single bus, as is known in the art. It should also be understood that the processors 320, 322 may include any number of central processing units ("CPUs") or cores, as is known in the art. The processors 320, 322 in blade 262 are connected to other items such as a data bus that communicates with I / O devices 332, a data bus that communicates with non-volatile storage devices 334, and other buses common in stand-alone computing systems. (For clarity, Figure 3Only connections from processor 320 to these other devices are shown.) Processors 320, 322 may be, for example, Intel Corporation Core TM The processor. The I / O bus may be, for example, a PCI or PCI Express ("PCIe") bus. The storage bus may be, for example, a SATA, SCSI, or Fibre Channel bus. It will be appreciated that other bus standards, processor types, and processor manufacturers may be used according to illustrative embodiments of the present invention.

[0043] Each blade (e.g., blades 262 and 264) includes an application specific integrated circuit 340 (also referred to as an "ASIC," "hub chip," or "hub ASIC") that controls many of its functions. More specifically, in order to logically connect processors 320 and 322, RAM 324 and 326, and other devices 332 and 334 together to form a managed multi-processor, (optionally) coherently-shared, distributed memory HPC system, processors 320, 322 are electrically connected to hub ASIC 340. Hub ASIC 340 thus provides an interface between the HPC system management functions generated by SMN 220, chassis controller 260, and blade controller 310 and the computing resources of blade 262.

[0044] In this regard, the hub ASIC 340 is connected to the blade controller 310 by way of a direct connection or by way of a field programmable gate array ("FPGA") 342 or similar programmable device for passing signals between the integrated circuits. One skilled in the art may select an appropriate connection between the hub ASIC 340 and the blade controller 310. The discussion of direct or indirect connections should not limit the various embodiments of the present invention.

[0045] In particular, signals are generated on output pins of blade controller 310 in response to commands issued by chassis controller 260. In the case of an indirect connection, these signals are converted by FPGA 342 into commands for certain input pins of hub ASIC 340, and vice versa. For example, a "power on" signal received by blade controller 310 from chassis controller 260 requires (among other things) that a "power on" voltage be provided to a certain pin on hub ASIC 340; FPGA 342 facilitates this task.

[0046] The field programmable nature of FPGA 342 allows the interface between blade controller 310 and ASIC 340 to be reprogrammed after manufacture. Thus, for example, blade controller 310 and ASIC 340 may be designed with certain common functions, and FPGA 342 may be advantageously used to program the use of these functions in a dedicated manner. If a hardware design error is discovered in either module, the communication interface between blade controller 310 and ASIC 340 may also be updated, allowing for rapid system repair without the need to manufacture new hardware.

[0047] Also related to its role as an interface between computing resources and system management, hub ASIC 340 is connected to processors 320, 322 via a high-speed processor interconnect 344. In one embodiment, processors 320, 322 are manufactured by Intel Corporation, which provides a processor interconnect 344 for this purpose. Quick Path Interconnect ("QPI"), and hub ASIC 340 includes modules for communicating with processors 320, 322 using QPI. Other embodiments may use other processor interconnect configurations.

[0048] The hub chip 340 in each blade also provides connections to other blades for high-bandwidth, low-latency data communications. Therefore, the hub chip 340 includes a link 350 to the compute connection 280 that connects different blade chassis. For example, this link 350 can be implemented using a network cable. The hub ASIC 340 also includes connections to other blades in the same blade chassis 252. The hub ASIC 340 of blade 262 is connected to the hub ASIC 340 of blade 264 via a chassis compute connection 352. The chassis compute connection 352 can be implemented as a data bus on the backplane of the blade chassis 252, rather than using a network cable, thereby advantageously allowing very high-speed data communications between blades required for high-performance computing tasks. Data communications on both the inter-chassis compute connection 280 and the intra-chassis compute connection 352 can be implemented using the NumaLink protocol or a similar protocol.

[0049] Blade cooling

[0050] Figure 4A high performance computer system 100 including a housing 401 and a liquid cooling system is schematically illustrated. The cooling system receives a cooling fluid 411 (e.g., water, ethylene glycol, a mixture of water and ethylene glycol, etc.) from a source 410, which may be a thermal storage tank (also referred to as a cold storage tank) or a water supply from a facility where the computer 100 is located. In the illustrative embodiment, the cooling fluid 411 flows from its source 410 via a first fluid line 402 to a cooling distribution unit 420. In some embodiments, the cooling distribution unit 420 may include a heat exchanger 444 in which the cooling fluid 411 from the source 410 receives heat from the cooling fluid 411 that has been circulated to one or more blades (e.g., such as blade 262 in blade chassis 252 and blade 266 in blade chassis 256 in blade housing 416). However, in some embodiments, the cooling distribution unit 420 omits the heat exchanger 444 and circulates the cooling fluid 411 from the source 410 to the cooling blades 262 without passing through the heat exchanger.

[0051] The cooling fluid 411 is circulated around and / or within the high performance computer 100 via one or more pumps 446. Figure 4 In some embodiments, the cooling fluid 411 is circulated through the liquid cooling system 800 within the computer housing 401 between the blades 262 disposed within the blade enclosure 416. In some embodiments, the cooling fluid 411 is returned from the high performance computer 100 to the source 410 via the return line 404.

[0052] Figure 5A A cooling arrangement for blade electronics 502 of blade 262 of high performance computer 100 is schematically illustrated in accordance with various embodiments. The illustrative embodiments describe using the cooling arrangement with DIMMs, but the blade electronics are not limited to DIMMs and may be, for example, another integrated circuit.

[0053] The thermally conductive conduit 501 on the blade 262 extends between the heat line 601 (coupled to the blade electronics 502) and the liquid cooling circuit 810, and thereby enables thermal energy to be transferred from the blade electronics 502 to the liquid cooling circuit 810 laterally across the face 512 of the blade 262, even though the blade electronics 502 and the liquid cooling circuit 810 are not in direct physical contact with each other. Such an embodiment is beneficial for blades 262 where it is not possible or desirable to bring the liquid cooling circuit 810 into direct thermal contact with the blade electronics 502. In a preferred embodiment, the thermally conductive conduit 501 is a dry thermal communication path on the blade 262 that interfaces with the cooling circuit 810 that is not on the blade 262. As used herein, two objects are in "dry thermal communication" when thermal energy can be transferred from one object to another without the flow of liquid between them.

[0054] Some embodiments of the heat conducting pipe 501 change the direction of heat flow from the heat pipe 601 to the cooling circuit 810 by rotating the angle 505, for example, as in Figure 5B Schematically illustrated in . Such a conduit may be described as being curved or bent and following a circuitous path. In such an embodiment, a component of the heat conductive conduit 501 (such as a transfer rod 700) may include a bend at an angle 505, or two or more components (such as two transfer rods 700) may be in thermal contact with each other and arranged at an angle 505 relative to each other. Without restricting the path of the heat conductive conduit 501 to a straight line, such an embodiment allows a blade designer to design the heat conductive conduit 501 to transfer heat from the heat line 601 to the cooling circuit 810, and gives the board designer the freedom to arrange other components at the location of the blade 262 (e.g., the processor 320) and / or the cooling circuit 810 as desired, because the heat conductive conduit 501 can be configured to accommodate such an arrangement and move heat around the blade 262 from its source to the cooling circuit.

[0055] To this end, in some embodiments, each component of the heat-conducting conduit 501 between the heat line 601 and the cooling loop 810 has a thermal conductivity of at least 50 W / (mK) [or Wm -1 K -1 ] or a thermal conductivity (k) of 100 W / (mK). For example, the component can be made of one or more of aluminum having a thermal conductivity of about at least 200 W / (mK) or copper having a thermal conductivity (k) of about 300 W / (mK). In contrast, the thermal conductivity (k) of FR4 circuit board material is less than 1.

[0056] Furthermore, in a preferred embodiment, the thermal conductivity of all components of the heat conductive conduit 501 is greater than the thermal conductivity of the surrounding environment and greater than the thermal conductivity of the substrate (e.g., circuit board) 511 of the blade 262. In this manner, the components of the heat conductive conduit 501 form a path of least thermal resistance for thermal energy to flow from the blade electronics 502 to the cooling circuit 810. This desirably mitigates or prevents the flow of thermal energy from the blade electronics 502 to other components on the blade 262 and delivers such flow to the reservoir 410.

[0057] Illustrative embodiments of the heat pipe 501 described further below include a transfer rod 700 thermally coupled between the heat line 601 and the cooling pipe 811. In some embodiments, the heat pipe 501 also includes a transfer block 610 for physically coupling the heat line 601 to the transfer rod 700 and / or thermally coupling the heat line 601 to the transfer rod 700, such as by physical contact, to facilitate heat conduction. In some embodiments, the transfer rod 700 includes an interface to the liquid cooling system 800, which is configured to receive thermal energy from the transfer rod 700 and conduct the thermal energy to the thermal storage 410.

[0058] Figure 5C and Figure 5D An embodiment of a blade 262 for a high performance computer 100 is schematically illustrated in accordance with an illustrative embodiment.

[0059] Blade 262 includes four DIMMs 2111, 2112, 2113, and 2114 longitudinally adjacent to another four DIMMs 2121, 2122, 2123, and 2124, all of which are mounted to blade substrate 511. In this illustrative embodiment, each "DIMM" is a dual in-line memory module that includes at least one memory circuit. Embodiments of DIMMs herein may also be referred to as DIMM modules. As an example, substrate 511 may be a printed circuit board made of FR4 material known in the electronics industry.

[0060] In this embodiment, the DIMMs define a blade gap 550 below the center of the substrate 511, wherein the blade gap 550 is disposed with one or more microprocessors 320 mounted on the substrate 511. The DIMMs 2111, 2112, 2113, and 2114, and the DIMMs 2121, 2122, 2123, and 2124 are in digital communication with the one or more processors 320.

[0061] The configuration of the blade 262 leaves little room for a liquid conduit on the base plate 511 to couple the blade 262 to the thermal storage device 410. Furthermore, fluidly coupling the liquid conduit to the blade 262 would have the undesirable consequence of requiring such a liquid conduit to be connected each time the blade 262 is installed in the high performance computer 100 and to be disconnected each time the blade 262 is removed from the high performance computer 100.

[0062] In some embodiments, the two blades may be mounted face to face in what may be referred to as a "twinning" arrangement, and may even be mounted very close to each other so that some of their components intersect. Figure 5E In FIG. 1 , two blades 262 and 264 are arranged adjacent to each other, and DIMMs 2111, 2112, 2113, and 2114 of blade 262 intersect with DIMMs 2211, 2212, 2213, and 2214 of blade 264 because a line drawn parallel to the X-axis would pass through DIMM 2111, DIMM 2213, and DIMM 2112. Pairing increases the component density of a high-performance computer, putting more computing power into a smaller space, thereby reducing the space required for the high-performance computer, and in some cases it is desirable to shorten the electronic communication paths between blades.

[0063] Figure 5E Schematically illustrates an end view of DIMMs 2111, 2112, 2113, and 2114 when these DIMMs are configured in pairs with corresponding blades 262 and 264 of DIMMs 2211, 2212, 2213, and 2214. FIG. 5C to FIG. 5E As shown in , the pairs leave little room for cooling air to flow around and between the blades and DIMMs.

[0064] Thus, some embodiments extend the cooling loop 810 across and adjacent to the face 512 of the blade 262. This configuration allows the cooling liquid 411 to be closer to the heat generating electronic devices (e.g., processor 320, DIMM 262) than in systems where the liquid is confined to the rack 101, thereby providing greater heat removal capabilities.

[0065] DIMM module

[0066] exist Fig. 6A and Figure 6B In the embodiment of the present invention, the DIMM 2111 has a heat pipe 601 arranged on one side 2131, and optionally two or more heat pipes 601 arranged on opposite sides 2131 and 2132 of the DIMM 2111. As an example, Figure 6Cschematically illustrates an end view of a DIMM 2111 having two heat pipes 601 (one on each side of the DIMM 2111), and Fig.6D The same DIMM 2111 is schematically illustrated.

[0067] Each such heat pipe 601 is thermally connected to the memory circuit 515 in the DIMM 2111. As is known in the art, a "heat pipe" 601 is a sealed tube with two opposite ends, in which an internal core and a evaporative fluid are sealed. Heat from a heat source (e.g., a circuit; a DIMM) near a first end of the heat pipe causes the evaporative fluid within the heat pipe to evaporate from a liquid state to a gaseous state. The gaseous fluid travels through the heat pipe to a second end that is arranged near a cooling source. As heat from the gas is transferred to the cooling source, the gaseous fluid condenses back to a liquid state at the second end. The liquid evaporative fluid then travels back to the first end via the core. Such a heat pipe can be described as a closed-loop system in which the evaporative fluid flows, but the evaporative fluid does flow into or out of the system.

[0068] The DIMM 2111 in this embodiment has a gap pad 516 between two memory circuits 515. The gap pad 516 is preferably thermally conductive to facilitate heat transfer from the memory circuit 515 to the heat pipe 601. In this embodiment, the gap pad 516 is 0.5 mm thick and has a thermal conductivity of 5.0 W / mK. As an example, in some embodiments, the gap pad 516 can be a Bergquist gap pad 5000S35, which is a glass fiber reinforced filler and polymer with high thermal conductivity characteristics available from Bergquist Company.

[0069] Figure 6C The DIMM 2111 also has a DIMM cover 517 that encapsulates the memory module 515. The DIMM cover 517 may include two halves coupled by fasteners 614 through holes 518. In other embodiments, such as in Fig. 6E and Fig. 6F As schematically illustrated in FIG. 5 , the DIMM cover 517 may be secured by one or more compression DIMM clips 519 .

[0070] In some embodiments, the heat pipes 601 are secured to the DIMMs 2111 to secure the heat pipes 601 in position relative to the DIMMs 2111 and to facilitate heat conduction from the DIMMs 2111 to the heat pipes 601. For example, in some embodiments, the heat pipes 601 are soldered to the DIMMs 2111 and / or the transfer block 610. As shown, the longest dimension of each heat pipe 601 extends along the DIMM 2111. Thus, when the DIMMs 2111 are mounted to the blade substrate 511 (e.g., to the face 512 of the blade substrate 511), the position and orientation of the heat pipes 601 relative to other features of the blade or cooling loop 810 are determined by and matched to the placement and orientation of the DIMMs.

[0071] The transfer block 610 is configured to physically and thermally interface between the heat line 601 and the transfer rod 700. 7A to 7E In some embodiments, the transport block 610 is configured to physically and thermally interface to multiple heat pipes (eg, two heat pipes 601 on the same DIMM 2111).

[0072] In the illustrated embodiment, each heat pipe 601 includes a flange 611 extending at a right angle relative to the sealing tube 602 of the heat pipe 601. The flange 611 nests with the transfer block 610 to fix the heat pipe 601 to the transfer block 610 and conducts heat from the heat pipe 601 to the transfer block 610.

[0073] In some embodiments, the transfer block 610 is a unitary device having a base 618 wide enough to accommodate two heat lines 601 , and two wings 619 extending upward from the base 618 to nest with flanges 611 on the heat lines 601 .

[0074] In other embodiments, for example, Fig. 6E , the transfer block 610 has a base 618 configured to accommodate a single heat pipe 601, and wings 619 extending upward from the base 618 to nest with flanges 611 on the heat pipe 601. Some embodiments include two transfer blocks 610, wherein the bases 618 of the transfer blocks 610 are configured to interlock with each other, such as in Fig. 6E 2131 is schematically illustrated in which the transfer block 610 on the side 2131 is straight and the base 618 of the transfer block 610 on the side 2132 is shaped so that the two bases 618 are nested in each other. In some such embodiments, the holes 613 of each base 618 can overlap so that a single fastener 614 can pass through the two holes 613 and secure the two transfer blocks 610 and the DIMM 2111 to the transfer rod 700.

[0075] In the illustrated embodiment, the heat pipe 601 is installed between the DIMM 2111 and the blade substrate (e.g., circuit board) 511. To this end, in some embodiments, the transfer block 610 includes a hole 613 configured to allow a fastener 614 to pass through the transfer block 610 and be secured to the transfer block, and thus secure the rest of the DIMM 2111 to the transfer rod 700. For example, the threaded fastener 614 can mate with the threaded hole 713 in the transfer rod 700.

[0076] Transmission rod

[0077] exist Fig. 7A , Figure 7B , Figure 7C , Fig.7D and Fig. 7E Schematically illustrated in FIG. 1 , an embodiment of a transfer rod 700 is shown. The transfer rod 700 is configured to conductively receive thermal energy from one or more heat lines 601, for example, via a transfer block 610, and conduct the thermal energy to a liquid cooling loop 810. To this end, the transfer rod 700 is made of a thermally conductive material (such as copper or aluminum, to name a few) to facilitate heat transfer from the DIMM 2111 to the liquid conduit 811.

[0078] Fig. 7A A top view of the transfer rod 700 is presented, Figure 7B A side view of the transfer bar 700 is presented, and Figure 7C An end view of a transfer bar 700 is presented. The transfer bar 700 has a flat bottom 701 configured to mount to the base plate 511 of the blade 262.

[0079] The transfer rod 700 is not a heat pipe and in some embodiments only includes solid material (e.g., copper, aluminum). In other embodiments, the transfer rod 700 may include one or more intermediate heat pipes 720, or be coupled to one or more intermediate heat pipes 720, such as in Fig.7D For example, the intermediate heat line 720 can be arranged between the transfer rod 700 and the base plate 511 of the blade 262. To this end, the intermediate heat line 720 can be arranged in a channel 730 in the bottom 701 of the transfer rod 700. Fig. 7E The transfer bar 700 is schematically illustrated with an intermediate heat line 720 installed in the channel 730. In any case, the intermediate heat line 720 can be soldered to the transfer bar 700. Such an intermediate heat line 720 enhances the thermal conductivity of the transfer bar 700, thereby helping to carry heat from the DIMM to the liquid conduit 811.

[0080] The transfer bar 700 also includes one or more heat pipe interfaces 710 configured to be thermally conductively connected to one or more heat pipes 601. For example, the heat transfer block 610 of the DIMM 2111 can be nested with the heat pipe interface 710. To this end, in this embodiment, each heat pipe interface 710 has a recess 712 arranged between and defined by two ridges 711. The size and position of the recess 712 are set to receive one or more heat pipes 601 and DIMM 2111. In some embodiments, two transfer blocks 610 from two longitudinally arranged DIMMs 2111, 2121 can be nested in the heat pipe interface 710. Such a configuration promotes heat transfer from the heat pipe 601 to the transfer bar 700, and can also be used to fix the heat pipe 601 relative to the transfer bar 700 and / or the blade 262. As can be understood from the previous description, the heat pipe 601 is in dry thermal communication with the transfer bar 700.

[0081] Some embodiments of the transfer rod 700 also have one or more conduit interfaces 705 shaped to receive the liquid conduit 811. Such an interface may be referred to as a cooling system interface. Typically, the shape of the conduit interface 705 is complementary to the shape of the liquid conduit 811, so that the liquid conduit 811 is nested in the conduit interface, thereby increasing the contact area between the liquid conduit 811 and the conduit interface 705 to promote heat conduction. To this end, in a preferred embodiment, the conduit interface 705 is contoured (uneven) and complementary to the shape of the liquid conduit 811, so that the conduit 811 fits tightly into the conduit interface 705. As an example, if the liquid conduit 811 has a convex contour (e.g., a curved or circular cross-section), the conduit interface 705 has a concave contour (e.g., an arcuate or semicircular cross-section), as in Figure 7B Schematically illustrated in FIG.

[0082] The fluid conduit 811 in some embodiments may extend parallel to the face 512 of the blade substrate 511 and simply rest in the conduit interface 705. Typically, there is no need to bolt or otherwise secure the transfer rod 700 to the fluid conduit 811. Among other things, bolting or otherwise securing the transfer rod 700 to the fluid conduit 811 would undesirably complicate the installation and removal of the blade 262.

[0083] Other embodiments may include thermally conductive material between the fluid conduit 811 and the delivery rod 700. The conduit interface 705 may help secure the fluid conduit 811 in place relative to the delivery rod 700 and / or the blade 262 and facilitate heat transfer from the delivery rod 700 to the fluid conduit 811.

[0084] As from Fig. 8A and Figure 8BAs can be understood in FIG. 1 , the DIMMs (2111, etc.) are parallel to the liquid conduit 811 and the DIMMs and their associated heat pipes 601 are therefore not directly coupled to the liquid conduit 811. Therefore, one benefit of the transfer rod 700 is that it can direct the heat flow direction from the heat pipe (i.e., in Fig. 8A and Figure 8B The Y-axis direction) is changed to a direction intersecting with the liquid conduit 811 (i.e., in Fig. 8A and Figure 8B This presents the blade designer with greater flexibility regarding the location and orientation of the DIMM 2111 etc. relative to the fluid conduit 811.

[0085] Liquid cooling loop

[0086] Fig. 8A and Figure 8B Portions of an embodiment of a cooling system 800 are schematically illustrated, which absorbs heat from the DIMMs and transports the heat to the thermal storage 410. To this end, the cooling system includes a cooling circuit 810 and four processors 320, the cooling circuit having liquid conduits 811 arranged adjacent to components of the blade 262 (including the DIMMs 2111, etc.).

[0087] As described above, the liquid conduit 811 is in liquid communication with the heat accumulator 410. In operation, the cooling fluid 411 from the heat accumulator 410 flows into the liquid conduit 811 via the inlet coupling 821 in the valve 820, circulates around the blade 262, collects heat generated by the electronic devices on the blade (e.g., DIMM 2111, etc.), and flows out of the liquid conduit 811 via the outlet coupling 822 to flow back to the heat accumulator 410. However, if the cooling fluid 411 from the heat accumulator 410 flows into the liquid conduit 811 via the inlet coupling 821 in the valve 820, the cooling fluid 411 flows into the liquid conduit 811 via the inlet coupling 821 in the valve 820, circulates around the blade 262, collects heat generated Fig. 8A and Figure 8B As can be understood, the liquid conduit 811 does not contact the DIMM 2111, etc.

[0088] As described above, on blade 262 , dry heat conduit 501 and liquid conduit 811 between heat lines 601 in thermal communication with the DIMMs conduct heat from DIMM 2111 to liquid conduit 811 .

[0089] Replace DIMMs and / or other components

[0090] According to the foregoing embodiments, the thermally conductive dry conduit 501 can be modular because the heat pipe 601 and one or more of the transfer block 610 and / or transfer rod 700 are separable from each other. This allows the DIMM 2111 to be easily removed from the blade 262. For example, the DIMM 2111 can be removed from the blade 262 by disconnecting the heat pipe 601 of the blade 262 from the transfer block 610 associated therewith, or by disconnecting the transfer block 610 from the transfer rod 700 (by removing the fastener 614). The DIMM 2111 can then be pulled out of the blade 262. The DIMM 2111 can then be replaced by a different DIMM (e.g., such as a different DIMM with greater memory capacity or higher or lower heat generation).

[0091] Similarly, modularity allows for removal and / or replacement of heat lines 601 and / or transport blocks 610, such as replacing a defective heat line 601 or transport block 610, or replacing a heat line 601 and / or transport block 610 with a new version having greater cooling capacity.

[0092] Dry Blade Installation and Removal

[0093] Some embodiments are configured to enable blade 262 to be installed into high performance computer 100 without having to create a liquid connection between the liquid conduit (e.g., 811) and the blade assembly. Similarly, in a preferred embodiment, blade 262 is configured to be removable from the high performance computer without having to disrupt such a liquid connection because the cooling circuit (e.g., liquid conduit 811) is not part of blade 262 or is fixed to blade 262.

[0094] This is an advantage over other possible cooling concepts that would not provide the benefits of liquid cooling for the electronics 502 on the blade and / or would require a technician to engage a liquid coupling through which cooling liquid can flow between the cooling system and some liquid conduit on the blade, and to disconnect such a liquid coupling in order to remove the blade from the high performance computer 100. Making and / or breaking such a liquid connection increases the complexity of installing and removing the blade from the computer, requiring appropriate training of the technician and increasing the time required for installation and removal. Furthermore, making and / or breaking such a liquid connection increases the risk of spilling liquid onto the blade or other components of the computer, thereby potentially damaging the blade and / or the computer, and also potentially compromising the computer's cooling system due to the loss of cooling liquid.

[0095] As an example, the conduit interface 705 in some embodiments is configured to enable the blade 262 to be installed in the high-performance computer 100 (e.g., into the blade chassis 252) and to be operably engaged with the cooling system 800 (e.g., the liquid conduit 811) to conduct thermal energy from the blade electronics 502 to the cooling system 800 without making a liquid connection, and to disconnect the blade from the cooling system 800 and remove it from the high-performance computer 100 without disrupting the liquid connection.

[0096] Fig. 7A The catheter interface 705 in is aligned along the Y axis, and Fig. 7A and Figure 7B At least a portion of the liquid conduit 811 in the blade chassis 252 is aligned along the Y axis. The blade chassis 252 in which the blade 262 is installed can be configured so that the blade 264 slides into and out of the blade chassis by moving along the Y axis. In such a configuration, the conduit interface 705 can receive a portion of the liquid conduit 811 and can slide along a portion of the liquid conduit 811 until the blade 262 is fixed in the blade chassis 252, at which position the liquid conduit 811 remains nested in the conduit interface 705 and is in thermal communication with the transfer rod 700. Therefore, the transfer rod 700 can be operably engaged with the liquid conduit 811; that is, the transfer rod 700 is thermally connected to the transfer rod 700 (e.g., dry thermal communication) without having to establish a liquid connection between the liquid conduit 811 and the transfer rod 700. In addition, the transfer rod 700 can be operably engaged with the liquid conduit 811 without having to fasten the transfer rod 700 to the liquid conduit 811, such as by bolts or intermediary devices.

[0097] Similarly, the blade 262 may be removed from the blade chassis 252 of the high performance computer 100 without having to disrupt the fluid connection between the fluid conduit 811 and the transfer rod 700. In keeping with the aforementioned installation example, wherein the conduit interface 705 is aligned along the Y-axis and at least a portion of the fluid conduit 811 is aligned along the Y-axis, the blade 262 may be removed from the blade chassis 252 by sliding the blade 262 in the opposite direction along the Y-axis, and in doing so, a portion of the fluid conduit 811 is disengaged from the conduit interface 705 without disconnecting or disrupting the fluid connection and / or without removing the fastener. Note that in the aforementioned illustrative example, when the blade 262 is removed, the cooling system 800 (e.g., the fluid conduit 811) remains in the high performance computer 100, whereas when the blade 262 is installed, the transfer rod 700 of the cooling system 800 (and thus its one or more heat lines 601 and one or more transfer blocks 610) is in dry thermal communication with the cooling system 800 (e.g., the fluid conduit 811).

[0098] The illustrative embodiments described herein achieve the benefits of liquid cooling of electronic components on blade 262 without requiring liquid to flow into and / or out of blade 262 from liquid conduit 811 and / or thermal accumulator 410. More specifically, the illustrative embodiments described herein transfer heat generated by blade electronics to a cooling loop of a computer through dry thermal communication without liquid exchange between the blade electronics and the cooling element. Although the embodiments are described by describing a system for cooling DIMMs, they are not limited to cooling DIMMs. For example, the blade electronics may be a computer processor circuit, a communication interface circuit, or a power supply circuit, to name a few.

[0099] As described above, the heat pipe 601 coupled to the DIMM 2111 can be in dry thermal communication with the liquid cooling loop 810 via the dry heat pipe 501, so that heat from the DIMM 2111 is transferred from the heat pipe 601 to the transfer rod 700 without liquid flowing from the heat pipe 601 to the transfer rod 700. In addition, the transfer rod 700 can be in dry thermal communication with the liquid cooling loop 810, so that even if cooling liquid flows in the cooling loop 810, such liquid does not flow from the transfer rod 700 to the cooling loop 810, or from the cooling loop 810 to the transfer rod 700.

[0100] The following is a list of reference numbers used herein.

[0101] 100: High performance computer systems;

[0102] 110: System console;

[0103] 120: Calculate partitions;

[0104] 130: Calculate partitions;

[0105] 140: Calculate partitions;

[0106] 150: Calculate partitions;

[0107] 160: Calculate partitions;

[0108] 170: Calculate partitions;

[0109] 180: logical communication network;

[0110] 191: Operating system instance;

[0111] 192: Basic Input / Output System (“BIOS”);

[0112] 193: Application software;

[0113] 194: Processor;

[0114] 195: Volatile memory;

[0115] 196: Non-volatile storage device;

[0116] 197: Input / output devices ("I / O");

[0117] 210: Enterprise data network;

[0118] 220: System Management Node ("SMN");

[0119] 230: Enterprise computers;

[0120] 240: Remote computer;

[0121] 252: Blade chassis;

[0122] 254: Blade chassis;

[0123] 256: Blade chassis;

[0124] 258: Blade chassis;

[0125] 260: controller;

[0126] 262: Blade;

[0127] 264: Blade;

[0128] 266: Blade;

[0129] 268: local management bus;

[0130] 270: Management connection;

[0131] 280: Calculate connection;

[0132] 302: chassis data storage device;

[0133] 310: blade management controller ("blade controller" or "BMC");

[0134] 316: Random Access Memory ("RAM");

[0135] 320: processor;

[0136] 322: processor;

[0137] 324: Random Access Memory ("RAM");

[0138] 326: Random Access Memory ("RAM");

[0139] 332: Input / output devices ("I / O");

[0140] 334: non-volatile storage device;

[0141] 340: Application-specific integrated circuit (“ASIC” or “central chip” or “central ASIC”);

[0142] 342: Programmable devices (e.g., field programmable gate arrays, “FPGAs”);

[0143] 344: High-speed processor interconnect;

[0144] 401: High performance computer housing;

[0145] 402: Pipeline from the heat accumulator;

[0146] 403: Pipeline from cooling distribution unit;

[0147] 404: reflux line;

[0148] 410: heat storage device;

[0149] 411: cooling fluid;

[0150] 416: Blade housing;

[0151] 420: cooling distribution unit;

[0152] 444: heat exchanger;

[0153] 446: Pump;

[0154] 501: High thermal conductivity conduit;

[0155] 502: Blade electronics;

[0156] 505: angle;

[0157] 511: a blade substrate (e.g., a printed circuit board);

[0158] 512: surface of blade substrate;

[0159] 515: memory circuit;

[0160] 516: gap pad;

[0161] 517: DIMM cover;

[0162] 518: Hole in DIMM cover;

[0163] 519: DIMM cover clip;

[0164] 550: Blade clearance

[0165] 2111 to 2114: DIMM;

[0166] 2121 to 2124: DIMM;

[0167] 2211 to 2214: DIMM;

[0168] 2131: first side of DIMM;

[0169] 2132: the second side of the DIMM;

[0170] 601: heat pipe;

[0171] 602: Sealing pipe of heat pipe;

[0172] 610: Transmit block;

[0173] 611: flange;

[0174] 613: Connector hole;

[0175] 614: Fasteners;

[0176] 618: Base of the transport block;

[0177] 619: Wing;

[0178] 700: transmission rod;

[0179] 701: bottom of the transmission rod;

[0180] 702: top of the transmission rod;

[0181] 705: catheter interface;

[0182] 710: heat pipe interface;

[0183] 711: ridge;

[0184] 712: Notch;

[0185] 713: threaded hole;

[0186] 720: intermediate heat pipeline;

[0187] 730: channel;

[0188] 800: Cooling system;

[0189] 810: cooling circuit;

[0190] 811: catheter;

[0191] 820: valve;

[0192] 821: Entrance;

[0193] 822: Exit.

[0194] The embodiments of the present invention described above are intended to be exemplary only; many variations and modifications will be apparent to those skilled in the art. All such variations and modifications are intended to fall within the scope of the present invention as defined in any of the appended claims.

Claims

1. A cooling system for a blade of a high-performance computer in communication with a heat storage device, the cooling system comprising: a heat pipe coupled to a dual in-line memory module DIMM on the blade to conductively receive thermal energy from the DIMM; as well as a dry thermal communication conduit having a high thermal conductivity and disposed on the face of the blade, the dry thermal communication conduit thermally coupling the heat line to a liquid conduit disposed across the face of the blade but not mounted to the face of the blade, The dry heat connection conduit includes a transfer rod, the transfer rod includes a cooling system interface, and the cooling system interface is formed to accommodate a liquid conduit so that the cooling system interface can slide along a portion of the liquid conduit, so that a portion of the liquid conduit can be engaged with the cooling system interface without making a liquid connection, or a portion of the liquid conduit can be disengaged from the cooling system interface without destroying the liquid connection.

2. The cooling system according to claim 1, wherein: The heat line and the transfer rod are configured such that the blade can be installed to and removed from a blade chassis containing the liquid conduit by engaging or disengaging the transfer rod from the liquid conduit without making or breaking a liquid connection.

3. The cooling system according to claim 2, wherein: The dry thermal communication conduit includes a transfer block in thermal communication with the heat line and in thermal communication with the transfer rod to transfer thermal energy from the heat line to the transfer rod.

4. The cooling system according to claim 3, wherein: The transfer block is secured to the transfer rod by at least one fastener.

5. The cooling system according to claim 2, wherein: The transfer rod includes an intermediate heat line.

6. The cooling system according to claim 2, wherein: The transfer rod includes a bottom side mounted to the face of the blade and a top side opposite the bottom side, wherein the cooling system interface includes one or more grooves recessed into the top side of the transfer rod and arranged to receive a portion of the liquid conduit and slide along a portion of the liquid conduit for thermally coupling the transfer rod to the liquid conduit.

7. The cooling system of claim 3, further comprising a second heat pipe coupled to the DIMM such that the DIMM is interposed between the heat pipe and the second heat pipe, and both the heat pipe and the second heat pipe are coupled to the transport block.

8. A computer system having a liquid cooling circuit, the computer system comprising: a first blade having a first electronic component and a first heat line in thermal communication with the first electronic component, and a first transfer bar in thermal communication with the first heat line; a second blade having a second electronic component and a second heat line in thermal communication with the second electronic component, and a second transfer bar in thermal communication with the second heat line; as well as a liquid cooling circuit connected to the cold accumulator, the liquid cooling circuit comprising a liquid conduit in dry thermal communication with the first transfer rod and the second transfer rod; the first heat line and the first transfer rod being configured to transfer thermal energy from the first electronic component to the cold storage via the liquid cooling circuit, and the second heat line and the second transfer rod being configured to transfer thermal energy from the second electronic component to the cold storage via the liquid cooling circuit, Wherein, the first transmission rod includes a cooling system interface, and the cooling system interface is formed to accommodate a liquid conduit, so that the cooling system interface can slide along a portion of the liquid conduit, so that a portion of the liquid conduit can be engaged with the cooling system interface without making a liquid connection, or a portion of the liquid conduit can be disengaged from the cooling system interface without destroying the liquid connection.

9. The computer system of claim 8, wherein: The first blade is arranged to face the second blade, and the first electronic component and the second electronic component intersect with each other.

10. The computer system of claim 8, wherein: The first transmission rod comprises copper or aluminum.

11. The computer system of claim 8, wherein: The first transfer rod further includes an intermediate heat line.

12. A dual in-line memory module (DIMM) module for use in a high performance computer having a heat storage device, the DIMM comprising: a first memory circuit; a first heat line in thermal communication with the first memory circuit to conductively receive heat generated by the first memory circuit; a transfer block in thermal communication with the first heat line to transfer heat from the first storage circuit to the thermal accumulator; as well as A transfer rod configured to transfer heat from the transfer block to the heat accumulator through a liquid cooling circuit including a liquid conduit, the transfer rod including a cooling system interface shaped to accommodate the liquid conduit so that the cooling system interface can slide along a portion of the liquid conduit to engage the portion of the liquid conduit with the cooling system interface without making a liquid connection, or to disengage the portion of the liquid conduit from the cooling system interface without breaking the liquid connection.

13. The DIMM module of claim 12, further comprising: a second memory circuit, the second memory circuit being separated from the first memory circuit by a gap pad; as well as a second heat pipe in thermal communication with the second storage circuit to conductively receive heat generated by the second storage circuit, The second heat line is in thermal communication with the transfer block to transfer heat from the second storage circuit to the thermal accumulator.

14. The DIMM module of claim 13, wherein: The first memory circuit and the second memory circuit are arranged between the first heat line and the second heat line.

15. The DIMM module of claim 12, wherein: The transport block includes holes configured to allow fasteners to secure the DIMM module to a blade.

Images