Cooling package for heterogeneous chips

Abstract

This article describes cooling hardware and methods for cooling heterogeneous computing architectures. A system for cooling a heterogeneous computing architecture includes a base stiffener; a top stiffener including mounting channels; a printed circuit board (PCB) including multiple electronic devices and chips, the PCB attached to the base stiffener; and a cooling device mounted atop the top stiffener. One or more heat transfer plates are inserted into the top stiffener via the mounting channels to transfer heat generated by the hardware modules to the cooling device. A resistance channel within the top stiffener is designed to ensure appropriate loading pressure across the entire assembly.

Description

Technical Field

[0001] Embodiments of the present disclosure generally relate to cooling hardware and systems. More specifically, embodiments of the present disclosure relate to cooling hardware and systems for heterogeneous computing architectures. Background Art

[0002] Heterogeneous computing refers to systems that use more than one type of processor or core to maximize performance or energy efficiency. Heterogeneous computing architectures can achieve greater energy efficiency by combining processors with unconventional cores such as custom logic, field-programmable gate arrays (FPGAs), or general-purpose graphics processing units (GPUs).

[0003] Heterogeneous computing architectures pose challenges to the design of accompanying cooling systems because different types of hardware modules in a heterogeneous system may have different thermal design power (TDP) and thermal specifications and requirements.

[0004] Some existing systems for cooling heterogeneous computing architectures use natural convection cooling systems that require pre-attached electronics to the die and the electronics thereon, such as high-speed connectors and voltage regulators (VRs). Other existing cooling systems use cooling devices that deliver a cooling medium (e.g., cooling air or cooling liquid) uniformly to the various types of hardware modules in the heterogeneous computing architecture. Existing thermal systems are either expensive or not flexible enough to accommodate the cooling needs of different types of hardware modules in the heterogeneous computing architecture. Designing cooling systems is becoming increasingly challenging, especially as heterogeneous hardware and chips continue to change. Summary of the Invention

[0005] The present disclosure aims to solve one of the above-mentioned technical problems in the related art.

[0006] An embodiment of the present disclosure provides a system for cooling a heterogeneous computing architecture, comprising: a base reinforcement; a top reinforcement, the top reinforcement including a mounting channel; a printed circuit board, the printed circuit board including a plurality of hardware modules, the printed circuit board being attached to the base reinforcement; a cooling device, the cooling device being mounted on top of the top reinforcement; and one or more heat transfer plates, the one or more heat transfer plates being inserted into the top reinforcement by sliding the one or more heat transfer plates through the mounting channel, wherein the one or more heat transfer plates are in contact with outer surfaces of the plurality of hardware modules to transfer heat generated by the plurality of hardware modules to the cooling device.

[0007] In an embodiment of the present disclosure, the roof reinforcement includes a first side surface and a second side surface, the first side surface is a closed end, and the second side surface is an open end.

[0008] In an embodiment of the present disclosure, the one or more heat transfer plates are inserted into the top reinforcement from the open end through the mounting channel.

[0009] In an embodiment of the present disclosure, the top reinforcement also includes a third side and a fourth side, each of the third side and the fourth side including two resistance channels, wherein the first resistance channel on any one of the third side and the fourth side is on top of the mounting channel, and the second resistance channel on any one of the third side and the fourth side is below the mounting channel.

[0010] In an embodiment of the present disclosure, each of the resistance channels includes one or more elastic structures to protect the one or more heat transfer plates inserted into the top reinforcement via the mounting channel.

[0011] In an embodiment of the present disclosure, the system further includes: a reinforcement mounting structure, which assembles the top reinforcement and the base reinforcement; and a system mounting structure, which assembles the cooling device, the top reinforcement and the base reinforcement.

[0012] In an embodiment of the present disclosure, the cooling device is one of an air cooling device or a liquid cooling device.

[0013] In an embodiment of the present disclosure, the hardware modules on the printed circuit board include one or more central processing units, one or more graphics processing units, one or more voltage regulators, and one or more high bandwidth memories.

[0014] In an embodiment of the present disclosure, each of the heat transfer plates has a mounting arm on each end, wherein the mounting arm is used to mount the heat transfer plate on the top reinforcement through the mounting channel.

[0015] In an embodiment of the present disclosure, the heat transfer plate includes a plurality of different types of heat transfer plates, and is selected based on cooling requirements of the hardware modules on the printed circuit board.

[0016] In an embodiment of the present disclosure, the heat transfer plate includes a vapor chamber, a thermoelectric cooler, and a copper heat transfer plate.

[0017] In an embodiment of the present disclosure, the system further includes a gap formed between each pair of the heat transfer plates to prevent heat from spreading between the paired heat transfer plates.

[0018] An embodiment of the present disclosure also provides an electronic rack for a data center, comprising a plurality of server chassis arranged in a stack, each server chassis comprising: one or more servers, and each server comprising one or more hardware modules; and a system according to any one of the above embodiments, wherein the one or more hardware modules of each server are mounted on a printed circuit board of the system.

[0019] The presently disclosed embodiments also provide a method for cooling a heterogeneous computing architecture using a system according to any one of the above embodiments, the method comprising: providing a base reinforcement, a top reinforcement comprising a mounting channel, and a printed circuit board comprising a plurality of hardware modules, the printed circuit board being attached to the base reinforcement; mounting a cooling device on top of the top reinforcement; and inserting the one or more heat transfer plates into the top reinforcement by sliding the one or more heat transfer plates through the mounting channel, wherein the one or more heat transfer plates contact outer surfaces of the plurality of hardware modules to transfer heat generated by the plurality of hardware modules to the cooling device. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In the accompanying drawings, there are illustrated by way of example and not limitation embodiments of the present disclosure.In the drawings, like reference numerals indicate similar elements.

[0021] Figure 1 A side view of a cooling system according to one embodiment is illustrated.

[0022] Figure 2 A cooling system according to one embodiment is further illustrated.

[0023] Figure 3 A heat transfer plate according to one embodiment is illustrated.

[0024] Figures 4A to 4B Side views of two different implementations of a roof reinforcement according to an embodiment are illustrated.

[0025] Figures 5A to 5B A top view of a cooling system according to one embodiment is illustrated.

[0026] Figure 6 It is illustrated that a vapor chamber-based heat transfer plate according to one embodiment is used in the entire cooling system.

[0027] 7A to 7B A cooling system according to one embodiment is further illustrated.

[0028] Figure 8 A cooling system according to one embodiment is further illustrated.

[0029] Figure 9 A method of cooling a heterogeneous computing architecture according to one embodiment is illustrated.

[0030] Figure 10 is a 3D diagram of a roof reinforcement according to one embodiment.

[0031] Figure 11 is a block diagram illustrating an example of an electronics rack according to one embodiment. DETAILED DESCRIPTION

[0032] Various embodiments and aspects of the present disclosure will be described with reference to the details discussed below, and the accompanying drawings will illustrate various embodiments. The following description and drawings are illustrative of the present disclosure and should not be construed as limiting the present disclosure. Many specific details are described to provide a thorough understanding of the various embodiments of the present disclosure. However, in some cases, in order to provide a concise discussion of the embodiments of the present disclosure, well-known or conventional details are not described.

[0033] References in this specification to "one embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present disclosure. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment.

[0034] As described above, a heterogeneous computing architecture can include different types of hardware modules. As used herein, a hardware module in a heterogeneous computing architecture can be any hardware block on a printed circuit board in the heterogeneous computing architecture. Examples of hardware modules can include: processors, such as central processing units (CPUs), graphics processing units (GPUs); cores; high bandwidth memory (HBM); voltage regulators (VRs); dies; security chips; chiplets; system-in-package (SIPs); and system-on-chip (SoCs).

[0035] Each hardware module in a heterogeneous computing architecture may generate heat during operation, and this heat needs to be transferred to an external cooling device for dissipation. However, the hardware modules may have different cooling requirements that cannot be met by using a single heat transfer plate. Furthermore, the heterogeneous computing architecture can be continuously expanded with new hardware modules. The new hardware modules may require a different type of heat transfer plate (HTP) than the existing HTP installed for the heterogeneous computing architecture.

[0036] To address these issues, this document describes a cooling system and method that can meet the cooling needs of different types of hardware modules in a heterogeneous computing architecture. In one embodiment, a system for cooling a heterogeneous computing architecture includes: a base reinforcement; a top reinforcement including mounting channels; a printed circuit board (PCB) including multiple hardware modules, the PCB being mounted on a substrate attached to the base reinforcement; and a cooling device mounted on top of the top reinforcement. One or more heat transfer plates (HTPs) are inserted into the top reinforcement via the mounting channels to transfer heat generated by the hardware modules to the cooling device.

[0037] In one embodiment, one or more HTPs are inserted into the top reinforcement from the open end through the mounting channel. Each HTP has a mounting arm on each end. The mounting arm is used to mount the HTP on the top reinforcement. A key feature is that the HTP can be designed with different heat transfer technologies. The HTP is selected based on the cooling requirements of the hardware modules on the PCB. The HTP includes a heat sink, a thermoelectric cooler, and a copper heat transfer plate. A gap is provided between each pair of HTPs to prevent heat from spreading between the paired HTPs.

[0038] In one embodiment, each resistance channel includes one or more resilient structures to protect one or more HTPs inserted into the top reinforcement via the mounting channel.

[0039] In one embodiment, the top reinforcement has four sides, one of which is a solid or closed end, one of which is an open end, and each of the other two sides has two resistance channels. The first resistance channel on each of the other two sides is on top of the mounting channel, and the second resistance channel on each of the other two sides is below the mounting channel.

[0040] In one embodiment, the system further comprises: a reinforcement mounting structure that assembles the top reinforcement and the base reinforcement; and a system mounting structure that assembles the cooling device, the top reinforcement, and the base reinforcement.

[0041] In one embodiment, the hardware modules on the PCB may include one or more central processing units (CPUs), one or more graphics processing units (GPUs), one or more voltage regulators (VRs), and one or more high bandwidth memories (HBMs).

[0042] Various embodiments enable HTPs of different types and / or technologies to be integrated into the cooling system based on the cooling needs of heterogeneous computing architectures without modifying the general cooling system framework. When new hardware modules are added to the PCB of a heterogeneous computing architecture, the new HTPs can be inserted into the cooling system via mounting channels without modifying other components of the cooling system, thereby increasing the scalability of the heterogeneous computing architecture.

[0043] Furthermore, the various embodiments described herein can improve the reliability of heterogeneous computing architectures by eliminating potential damage to hardware modules on PCBs during the assembly process due to the use of elastic structures and separation of different layers. The various embodiments can also improve the cooling performance of heterogeneous computing architectures without increasing the cost of the cooling system.

[0044] Figure 1 A side view of a cooling system 100 according to one embodiment is illustrated. Figure 1 As shown, cooling system 100 includes a top reinforcement 103 and a base reinforcement 107. Top reinforcement 103 includes a mounting channel 109, a top resistance channel 110, and a bottom resistance channel 111. The top resistance channel is located on top of mounting channel 109, and the bottom resistance channel 111 is located below mounting channel 109. Each of top reinforcement 103 and base reinforcement 107 can be a frame reinforcement.

[0045] As further shown, one side (e.g., the left side) of the top reinforcement 103 can be a solid edge, which makes the top reinforcement 103 closed on that side. The other side (e.g., the right side) of the top reinforcement 103 is open, which allows one or more heat transfer plates (HTPs) to be inserted into the top reinforcement 103 through the mounting channel 109.

[0046] In one embodiment, one or more elastic structures 113 may be positioned at the edge of the top resistance channel 110. Similarly, one or more elastic structures 114 may be positioned at the edge of the bottom resistance channel 111. Each of the elastic structures 113 and 114 may provide appropriate cushioning and appropriate pressure to the HTP inserted into the top reinforcement 103.

[0047] Cooling system 100 may further include cooling device 115, which may be an air-cooled heat sink or a liquid-cooled cold plate. The HTP inserted into top stiffener 103 can transfer heat generated by hardware modules on PCB 105 to cooling device 115. It should be noted that the electronic devices packaged on PCB 105 are not shown in the figure.

[0048] Figure 2 A cooling system 100 according to one embodiment is further illustrated. Figure 2 As shown, the cooling system 100 may further include a reinforcement member mounting structure 203 for assembling the top reinforcement member 103 and the base reinforcement member 107 . The PCB 105 packaged with a plurality of different hardware modules is sandwiched between the top reinforcement member 103 and the base reinforcement member 107 .

[0049] The resistance channels 110 and 111 can ensure appropriate loading pressure on the hardware modules on the PCB 105 using elastic structures in the resistance channels 110 and 111 so that the loading pressure does not damage the PCB 105 and the hardware modules. The hardware modules can be fragile and can be damaged without proper protection against the pressure applied by components such as the cooling device 115 on top of the top stiffener 103. Both resistance channels 110 and 111 provide additional protection for the electronics and the HTP during both the hardware integration process and normal operation.

[0050] In one embodiment, the cooling system 100 further includes a system mounting structure 201 for assembling the external cooling device 115, the top stiffener 103, the base stiffener 107, and the PCB 105. The resistance channels 110 and 111 can ensure proper attachment and mounting pressure applied by the cooling unit 115 on the PCB 105. In an embodiment, the mounting structures 201 and 203 and the resistance channels 110 and 111 work together to ensure proper pressure loading between the surface of the electronic device and the HTP, and between the HTP and the cooling device 115.

[0051] Figure 3 1 illustrates a heat transfer plate 300 according to one embodiment. As shown, the heat transfer plate (HTP) 300 is a 3D vapor chamber, which is a two-phase device used to spread heat from a heat source to a cooling device. In this example, the heat source is one or more hardware modules on the PCB 105, and the cooling device is a PCB. Figure 1 The cooling device 115.

[0052] In one embodiment, the 3D vapor chamber may include multiple sealed metal bags filled with liquid (e.g., metal bag 305). When a heat source heats the liquid in the metal bags, the liquid ensures that the heat is evenly dissipated to the copper heat pipes (not shown) in the 3D vapor chamber 300, thereby eliminating hot spots.

[0053] The main section of the 3D vapor chamber 300 is the main heat transfer body 301, the bottom side 306 of the body 301 is attached to the heat source, and the top side 304 of the body 301 is attached to the cooling device 115. Figure 1 As described above, the mounting arms 302 and 303 at both ends of the vapor chamber body 301 can be used to attach the vapor chamber body 301 to the top reinforcement 103 .

[0054] Heat transfer plate 300 can also be made of a variety of other types, including thermoelectric coolers (TECs) and pure copper. It's important to note that the vapor chamber structure is fragile, meaning it can easily be crushed when the applied pressure exceeds its design specifications. If deformation occurs, the vapor chamber will not function properly.

[0055] Figures 4A to 4B Two different embodiments of the roof reinforcement 103 are illustrated in side views according to an embodiment. For each embodiment, the side view is a view of one of the two mounting sides of the roof reinforcement 103.

[0056] Figure 4A It is shown that the top reinforcement 103 is implemented as a single-section reinforcement. Figure 4A The top reinforcement in FIG has four sides. One side is a solid end or closed end 403, and one side is an open end 402, which allows the HTP to be inserted into the top reinforcement. The other two sides are mounting sides for mounting the HTP inserted into the top reinforcement. Figure 4A One of two mounting sides is shown.

[0057] As an illustrative example, if the vapor chamber HTP 300 is to be inserted into Figure 4A In the top reinforcement 103 in FIG. 1 , each mounting arm 302 and 303 can be mounted on one of the two mounting sides through the mounting channel.

[0058] The top reinforcement includes two resistance channels 401 and 402 with elastic structures 407 and 409 to provide protection for the HTP inserted into the top reinforcement 103 when the top reinforcement is assembled into the cooling system, and also for the PCB (e.g. Figure 1 The PCB 105 in the device and the hardware modules packaged on the PCB 105 provide protection.

[0059] In one embodiment, the two resistance channels 401 and 405 may be on a physical entity that can be inserted into the top reinforcement 103. The HTP may then be inserted into the top reinforcement 103 to be protected by the resistance channels 401 and 405.

[0060] Figure 4B It is shown that the reinforcement 103 is implemented as a multi-section reinforcement. Figure 4B The top reinforcement in FIG. 4 also has four sides. One side is a solid or closed end 411, and one side is an open end 414, which allows one or more HTPs to be inserted into the top reinforcement. The other two sides are mounting sides for mounting the HTPs inserted into the top reinforcement.

[0061] The top reinforcement includes two resistance channels 401 and 402 , each of which is divided into multiple sections, and each section includes one or more elastic structures. For example, a section in resistance channel 413 includes elastic structure 417 , and a section in resistance channel 415 includes elastic structure 419 .

[0062] In one embodiment, the multi-segment embodiment of the top reinforcement 103 provides flexibility for inserting and assembling multiple HTPs. The elastic structure in both embodiments can provide vertical displacement redundancy, limit the maximum vertical displacement, and protect the HTP inserted into the top reinforcement. In an embodiment, as Figure 4B The various sections of the resistance channels 413 or 415 shown may be different to accommodate different HTP implementations and functions.

[0063] Figures 5A to 5B 1. Top views of a cooling system 100 according to one embodiment are illustrated. Each top view is a view of the cooling system 100 after the cooling device 115 is removed.

[0064] Figure 5A A top view of the cooling system is shown after the HTP 501 is inserted into the top stiffener 103. In this example, all of the hardware modules on the PCB 105 can use the HTP 501 to transfer heat to the cooling device 115.

[0065] Figure 5B The HTP 501 is shown inserted into the top reinforcement 103. As shown, the mounting channel 109 on one mounting side (eg, Figure 1 ) and a mounting channel 509 on the other mounting side can be used to insert the HTP 501.

[0066] Figure 6 A vapor chamber-based HTP 601 is illustrated as being used in the overall cooling system 100 according to one embodiment.

[0067] As shown in the figure, multiple hardware modules are packaged on PCB 105. The hardware modules can be different types of chips or integrated circuits, as well as other supporting electronic devices. As an example in the figure, the hardware modules may include two voltage regulators (VRs) 604 and 612, two high-bandwidth memories 606 and 610, and a small chip 608.

[0068] All hardware modules packaged on the PCB 105 may use the vapor chamber-based HTP 601 to transfer heat to the cooling device 115 . Figure 6The embodiments described in

[0044] can use the cooling device 115 for normal cooling workloads as well as for enhanced cooling workloads. It can be seen that the current design also enables easy implementation of the vapor chamber device to high power heterogeneous electronic devices.

[0069] 7A to 7B A cooling system 100 according to one embodiment is further illustrated. More specifically, 7A to 7B Multiple HTPs of different types are shown being used to match different types of hardware modules on PCB 105 .

[0070] Figure 7A A side view of the cooling system 100 is shown with a copper heat transfer plate 701 inserted for VR 604 and another copper heat transfer plate 703 inserted for VR 612 . Figure 7A Also shown is the insertion of a phase change heat transfer plate 704 for both HBMs 606 and 610 and the chiplet 608. Different HTPs are inserted to match the different cooling and packaging requirements of the hardware modules on the PCB 105.

[0071] like Figure 7A As further shown, gaps 707 and 709 are retained between different HTPs for thermal insulation and for preventing heat from diffusing between a pair of HTPs separated by the gap. Figure 7B yes Figure 7A A top view of cooling system 100 is shown.

[0072] The systems and methods described in this disclosure are used to cool heterogeneous architectures. Therefore, the cooling devices on the top layer and HTP can be different to accommodate different heterogeneous chip packages.

[0073] Figure 8 Further illustrated is a cooling system 100 according to one embodiment. In this embodiment, various HTPs are integrated to transfer heat generated by hardware modules / packages on PCB 105 to cooling device 115. Cooling device 115 can be a general cooling plate or an air heat sink with uniform fins.

[0074] like Figure 8 As shown, the HTP may include a copper heat transfer plate 805 for VR 612 and another copper heat transfer plate 807 for VR 604. The HTP also includes a vapor chamber 801 for chiplet 608 and two TECs 803 and 804, respectively, for HBMs 610 and 606. Gaps 807, 809, 811, and 813 between the HTPs may provide thermal insulation between the HTPs.

[0075] Figure 9 A method 900 for cooling a heterogeneous computing architecture according to one embodiment is illustrated. Figure 9As shown, in block 901, a base reinforcement, a top reinforcement including mounting channels, and a printed circuit board (PCB) including multiple hardware modules are provided, and the PCB is attached to the base reinforcement. In block 903, a cooling device is mounted on top of the top reinforcement. The cooling device is a cold plate or heat sink. In block 905, one or more heat transfer plates (HTPs) are inserted into the top reinforcement via the mounting channels to transfer heat generated by the multiple hardware modules to the cooling device.

[0076] Figure 10 FIG. 1 is a 3D diagram of a roof reinforcement 1000 according to one embodiment. The roof reinforcement 1000 is symmetrical in top and bottom directions and in left and right directions. Figure 10 Only a portion of the roof reinforcement is shown, primarily one side of the roof reinforcement 1000 .

[0077] The top reinforcement 1000 includes two resistance channels on each side. For example, the top reinforcement 1000 includes a bottom resistance channel 1023 and a top resistance channel 1024 on the right side. Each resistance channel can include multiple elastic structures (e.g., elastic structure 1021). The mounting channel 1025 can be used to insert the HTP D 1013 into the top reinforcement 1000. The mounting channel 1025 includes two separate channels, one between the two resistance channels on each side of the top reinforcement 1000.

[0078] Removable layer 1022 can be the bottom of top resistance channel 1024 and the top of the right side of mounting channel 1025. Similarly, another removable layer 1026 can be the bottom of the right side of mounting channel 1025 and the top of bottom resistance channel 1023. Both removable portions 1022 and 1026 can move up and down to provide protection for hardware modules on the HTP and / or PCB.

[0079] As further shown, one end 1019 of the roof reinforcement 1000 is solid, while the other end is open so that an HTP (e.g., HTP 1013) can be inserted into the roof reinforcement 1000. HTPs 1007, 1009, and 1011 represent different types of HTPs that have been inserted into the roof reinforcement 1000.

[0080] The top stiffener 1000 has a partially open bottom 1017 and a partially open top 1015 to allow direct contact between the electronic device and the HTP, and between the cooling device and the HTP.

[0081] Figure 11 1 is a block diagram illustrating an electronic rack according to one embodiment. Electronic rack 1100 may represent any electronic rack in a data center. Figure 11According to one embodiment, an electronics rack 1100 includes, but is not limited to, a CDU 1101, a rack management unit (RMU) 1102, and one or more server chassis 1103A-1103E (collectively, server chassis 1103). The server chassis 1103 can be inserted into a server slot array (e.g., a standard shelf) from either the front 1104 or rear 1105 of the electronics rack 1100. Note that while five server chassis 1103A-1103E are shown herein, more or fewer server chassis can be maintained within the electronics rack 1100. It should also be noted that the specific locations of the CDU 1101, RMU 1102, and / or server chassis 1103 are shown for illustrative purposes only; other arrangements or configurations of the CDU 1101, RMU 1102, and / or server chassis 1103 can also be implemented. In one embodiment, electronics rack 1100 can be open to the environment or partially contained by a rack container, as long as cooling fans can generate airflow from front to back.

[0082] In addition, for at least some of the server chassis 1103, an optional fan module (not shown) is associated with the server chassis. Each fan module includes one or more cooling fans. The fan modules can be mounted on the rear end of the server chassis 1103 or on the electronics rack to generate airflow that flows from the front end 1104, travels through the air space of the server chassis 1103, and is present at the rear end 1105 of the electronics rack 1100.

[0083] In one embodiment, CDU 1101 primarily includes a heat exchanger 1111, a liquid pump 1112, and a pump controller (not shown), as well as other components such as a liquid reservoir, a power supply, monitoring sensors, and the like. Heat exchanger 1111 may be a liquid-to-liquid heat exchanger. Heat exchanger 1111 includes a first circuit having an inlet and an outlet, each having a first pair of liquid connectors that connect to external liquid supply / return lines 131-132 to form a primary circuit. Connectors connected to external liquid supply / return lines 131-132 may be arranged or mounted on rear end 1105 of electronics rack 1100. Liquid supply / return lines 131-132, also referred to as room liquid supply / return lines, may be connected to an external cooling system (e.g., a data center room cooling system).

[0084] In addition, heat exchanger 1111 also includes a second loop having two ports with a second pair of liquid connectors. The second pair of liquid connectors are coupled to liquid manifold 1125 (also referred to as a rack manifold) to form a secondary loop. The secondary loop may include a supply manifold (also referred to as a rack liquid supply line or rack supply manifold) that supplies cooled liquid to server chassis 1103 and a return manifold (also referred to as a rack liquid return line or rack return manifold) that returns warmer liquid to CDU 1101. Note that CDU 1101 may be any type of commercially available or custom CDU. Therefore, the details of CDU 1101 will not be described herein.

[0085] Each server chassis 1103 may include one or more IT components (e.g., a central processing unit or CPU, a general / graphics processing unit (GPU), memory, and / or storage devices). Each IT component may perform data processing tasks, wherein the IT component may include software installed in the storage device, loaded into the memory, and executed by one or more processors to perform the data processing tasks. The server chassis 1103 may include a host server (referred to as a master node) connected to one or more computing servers (also referred to as computing nodes, such as CPU servers and GPU servers). The host server (having one or more CPUs) typically interfaces with clients over a network (e.g., the Internet) to receive requests for specific services, such as storage services (e.g., cloud-based storage services such as backup and / or recovery), executing applications to perform certain operations (e.g., image processing, deep data learning algorithms or modeling as part of a software-as-a-service or SaaS platform, etc.). In response to the request, the host server assigns the task to one or more of the computing nodes or computing servers (having one or more GPUs) managed by the host server. The computing servers perform the actual tasks, which may generate heat during operation.

[0086] The electronics rack 1100 also includes an optional RMU 1102 that is configured to provide and manage power to the servers and CDU 1101. The RMU 1102 can be coupled to a power supply unit (not shown) to manage the power consumption of the power supply unit. The power supply unit can include the necessary circuitry (e.g., an alternating current (AC) to direct current (DC) or DC to DC power converter, a battery, a transformer, or a regulator, etc.) to power the remaining components of the electronics rack 1100.

[0087] In one embodiment, the RMU 1102 includes an optimization module 1121 and a rack management controller (RMC) 1122. The RMC 1122 may include a monitor to monitor the operating status of various components within the electronic rack 1100, such as the computing nodes 1103, the CDU 1101, and the fan modules. Specifically, the monitor receives operational data from various sensors representing the operating environment of the electronic rack 1100. For example, the monitor can receive operational data representing the temperature of the processor, cooling liquid, and airflow, and the operational data can be captured and collected via various temperature sensors. The monitor can also receive data representing the fan power and pump power generated by the fan module and the liquid pump 1112, and the fan power and pump power can be proportional to the speed of each of the fan module and the liquid pump 1112. These operational data are referred to as real-time operational data. Note that the monitor can be implemented as a separate module within the RMU 1102.

[0088] Based on the operational data, the optimization module 1121 performs optimization using a predetermined optimization function or optimization model to derive a set of optimal fan speeds for the fan module and an optimal pump speed for the liquid pump 1112, such that the total power consumption of the liquid pump 1112 and the fan module is minimized, while the operational data associated with the liquid pump 1112 and the cooling fan of the fan module are within their respective design specifications. Once the optimal pump speed and the optimal fan speed have been determined, the RMC 1122 configures the liquid pump 1112 and the cooling fan of the fan module based on the optimal pump speed and the fan speed.

[0089] As an example, based on the optimal pump speed, the RMC 1122 communicates with the pump controller of the CDU 1101 to control the speed of the liquid pump 1112, which in turn controls the liquid flow rate of the cooling liquid supplied to the liquid manifold 1125 for distribution to at least some of the server chassis 1103. Similarly, based on the optimal fan speed, the RMC 1122 communicates with each fan module to control the speed of each cooling fan of the fan module, which in turn controls the airflow rate of the fan module. Note that each fan module can be individually controlled with its own specific optimal fan speed, and different fan modules and / or different cooling fans within the same fan module can have different optimal fan speeds.

[0090] Note that Figure 11The rack configuration shown is shown and described for illustrative purposes only; other configurations or arrangements are also applicable. For example, the CDU 1101 can be an optional unit. The cold plate of the server chassis 1103 can be coupled to the rack manifold, and the rack manifold can be coupled directly to the room manifolds 131-132 without using a CDU. Although not shown, a power supply unit can be provided within the electronics rack 1100. The power supply unit can be implemented as a standard chassis that is the same as or similar to the server chassis, wherein the power supply chassis can be inserted into any standard shelf and replace any server chassis 1103. In addition, the power supply chassis can also include a battery backup unit (BBU) to provide battery power to the server chassis 1103 when the main power source is unavailable. The BBU can include one or more battery packs, and each battery pack includes one or more battery cells and the necessary charging and discharging circuits for charging and discharging the battery cells.

[0091] In one embodiment, the cooling devices arranged in each server chassis as shown may represent any cooling devices described throughout this application.

[0092] In the foregoing description, embodiments of the present disclosure have been described with reference to specific exemplary embodiments thereof. However, it will be apparent that various modifications may be made thereto without departing from the broader spirit and scope of the present disclosure as set forth in the following claims. Accordingly, the description and drawings are to be regarded as illustrative rather than restrictive.

[0093] As previously described, embodiments of the present disclosure may be (or include) a non-transitory machine-readable medium (e.g., microelectronic memory) having stored thereon instructions that program one or more data processing components (generally referred to herein as "processors") to perform airflow management operations, such as controlling the fan speed of one or more fans of a battery module (and / or BBU shelf). In other embodiments, some of these operations may be performed by specific hardware components containing hardwired logic. These operations may optionally be performed by any combination of programmed data processing components and fixed hardwired circuit components of any battery module described herein.

[0094] While certain aspects have been described and shown in the accompanying drawings, it should be understood that these aspects are merely illustrative of the broad disclosure and not restrictive, and that the disclosure is not limited to the specific configurations and arrangements shown and described, as various other modifications may occur to those skilled in the art. Accordingly, this description is to be regarded as illustrative and not restrictive.

[0095] In some aspects, the present disclosure may include language, such as, “at least one of [element A] and [element B]”. This language may refer to one or more elements. For example, “at least one of A and B” may refer to “A”, “B”, or “A and B”. Specifically, “at least one of A and B” may refer to “at least one of A and at least one of B” or “at least one of A or B”. In some aspects, the present disclosure may include language, such as, “[element A], [element B], and / or [element C]”. This language may refer to any one element or any combination thereof. For example, “A, B, and / or C” may refer to “A”, “B”, “C”, “A and B”, “A and C”, “B and C”, or “A, B, and C”.

Claims

1. A system for cooling a heterogeneous computing architecture, comprising: base reinforcement; a roof reinforcement, the roof reinforcement including a mounting channel; a printed circuit board comprising a plurality of hardware modules, the printed circuit board being attached to the base reinforcement; a cooling device mounted on top of the top reinforcement; as well as one or more heat transfer plates, the one or more heat transfer plates being inserted into the top stiffener by sliding the one or more heat transfer plates through the mounting channels, wherein the one or more heat transfer plates contact exterior surfaces of the plurality of hardware modules to transfer heat generated by the plurality of hardware modules to the cooling device; The top reinforcement includes a third side and a fourth side, each of the third side and the fourth side including two resistance channels, wherein the first resistance channel on any one of the third side and the fourth side is on top of the mounting channel, and the second resistance channel on any one of the third side and the fourth side is below the mounting channel.

2. The system according to claim 1, wherein: The roof reinforcement also includes a first side and a second side, the first side being a closed end and the second side being an open end.

3. The system according to claim 2, wherein: The one or more heat transfer plates are inserted into the top reinforcement from the open end through the mounting channel.

4. The system according to claim 3, wherein: Each of the resistance channels includes one or more elastic structures to protect the one or more heat transfer plates inserted into the top reinforcement via the mounting channels.

5. The system according to claim 1, further comprising: a reinforcement mounting structure, the reinforcement mounting structure assembling the top reinforcement and the base reinforcement; and A system mounting structure is provided to assemble the cooling device, the top reinforcement, and the base reinforcement.

6. The system according to claim 1, wherein: The cooling device is one of an air cooling device or a liquid cooling device.

7. The system according to claim 1, wherein: The hardware modules on the printed circuit board include one or more central processing units, one or more graphics processing units, one or more voltage regulators, and one or more high bandwidth memories.

8. The system according to claim 1, wherein: Each of the heat transfer plates has a mounting arm on each end, wherein the mounting arm is used to mount the heat transfer plate on the top reinforcement through the mounting channel.

9. The system according to claim 1, wherein: The heat transfer plate includes a plurality of different types of heat transfer plates and is selected based on cooling requirements of the hardware modules on the printed circuit board.

10. The system according to claim 1, wherein: The heat transfer plate includes a vapor chamber, a thermoelectric cooler and a copper heat transfer plate.

11. The system of claim 1, further comprising a gap formed between each pair of the heat transfer plates to prevent heat from spreading between the paired heat transfer plates.

12. An electronic rack for a data center, comprising a plurality of server chassis arranged in a stacked manner, each server chassis comprising: One or more servers, and each server includes one or more hardware modules; as well as According to any one of claims 1-11, the one or more hardware modules of each server are mounted on a printed circuit board of the system.

13. A method for cooling a heterogeneous computing architecture using the system according to any one of claims 1 to 11, the method comprising: providing a base reinforcement, a top reinforcement including a mounting channel, and a printed circuit board including a plurality of hardware modules, the printed circuit board attached to the base reinforcement; installing a cooling device on top of the top reinforcement; as well as The one or more heat transfer plates are inserted into the top stiffener by sliding the one or more heat transfer plates through the mounting channels, wherein the one or more heat transfer plates contact outer surfaces of the plurality of hardware modules to transfer heat generated by the plurality of hardware modules to the cooling device.

Images