WO2024001769A1 - 一种计算机***和服务器 - Google Patents
一种计算机***和服务器 Download PDFInfo
- Publication number
- WO2024001769A1 WO2024001769A1 PCT/CN2023/100071 CN2023100071W WO2024001769A1 WO 2024001769 A1 WO2024001769 A1 WO 2024001769A1 CN 2023100071 W CN2023100071 W CN 2023100071W WO 2024001769 A1 WO2024001769 A1 WO 2024001769A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- layer
- heat dissipation
- computer system
- processor
- cable
- Prior art date
Links
- 230000017525 heat dissipation Effects 0.000 claims abstract description 161
- 238000001816 cooling Methods 0.000 claims description 29
- 239000007788 liquid Substances 0.000 claims description 21
- 238000000034 method Methods 0.000 claims description 18
- 230000008878 coupling Effects 0.000 claims description 4
- 238000010168 coupling process Methods 0.000 claims description 4
- 238000005859 coupling reaction Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 13
- 238000013461 design Methods 0.000 description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 230000008054 signal transmission Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 230000009977 dual effect Effects 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 244000132059 Carica parviflora Species 0.000 description 1
- 235000014653 Carica parviflora Nutrition 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/16—Constructional details or arrangements
- G06F1/18—Packaging or power distribution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/16—Constructional details or arrangements
- G06F1/18—Packaging or power distribution
- G06F1/183—Internal mounting support structures, e.g. for printed circuit boards, internal connecting means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/16—Constructional details or arrangements
- G06F1/20—Cooling means
Definitions
- the present application relates to the field of computer technology, and in particular, to a computer system and a server.
- each processor is usually deployed with a separate heat sink, and the heat exchange area of a single processor is limited.
- signal cables need to be added in the computer system to realize communication connections between processors and between processors and other devices, within the limited overall space of the computer system, there is interference between the signal cables and the radiator, and then As the number of processors continues to increase, the conflict between signal cable routing space and heat dissipation space has become more and more serious.
- Limited by the chassis size and heat dissipation capacity of the computer system existing computer systems cannot meet the increasing heat dissipation requirements of high-power processors. Therefore, how to provide a solution that meets the increasing heat dissipation requirements of processors in computer systems with limited space has become an urgent technical problem to be solved.
- This application provides a computer system and a server, which are used to improve the heat dissipation of the computer system and improve the deployment space utilization of the processor.
- a computer system in a first aspect, can be divided into the following three layers according to logical layers: a heat dissipation layer, a computing layer and a high-speed layer.
- the computing layer has a processor for generating electrical signals.
- the heat dissipation layer has a heat dissipation structure
- the high-speed layer has a cable module for transmitting electrical signals
- the heat dissipation structure is coupled to the processor
- the cable module is coupled to the processor.
- the heat dissipation structure can dissipate heat to the processor, and the cable module can transmit electrical signals.
- the heat dissipation structure is coupled to the processor (also called coupling), and the cable module is coupled to the processor, thereby releasing the cable routing space in the cable module and solving the problem.
- the problem of cable routing space and heat dissipation interference improves the heat dissipation of computer systems and improves the utilization of processor deployment space.
- the cable module includes internal terminals, cables, and external terminals.
- the internal terminals and external terminals are connected through cables.
- the processor has an output terminal, and the electrical signal generated by the processor is transmitted through the output port. Output, the internal terminal is coupled to the output terminal, the electrical signal is transmitted to the cable through the internal terminal, and the cable outputs the electrical signal to the external terminal.
- the external terminal is coupled to the motherboard of the computer system.
- the processor in the computing layer has an output terminal.
- the output terminal is used to output the electrical signal generated by the processor.
- the output terminal is coupled to the internal terminal in the cable module, so that the electrical signal is transmitted to the cable through the internal terminal.
- the electrical signal is transmitted to the external terminal through the cable, and the electrical signal is output to the motherboard of the computer system through the external terminal.
- the external terminal is detachably connected to the main body of the cable module.
- the main body of the cable module refers to the collection of other components in the cable module other than external terminals.
- internal terminals, external terminals, and cables can be regarded as the main body of the cable module.
- the external terminal can be completely removed from the main body of the cable module to replace other external terminals according to the needs of the actual application scenario.
- the external terminal includes multiple sub-modules, and each sub-module is detachably connected to the main body of the cable module.
- the main body of the cable module includes a UBC terminal, and each sub-module is detachably connected to the UBC terminal.
- the external terminal has One sub-module can be detached from the main body of the cable module and replaced with another sub-module, thereby improving the adaptability of the external interface of the computer system.
- the external terminal is a blind plug connector.
- the blind plug connector is a specific form of external terminal implementation.
- the blind plug connector is coupled to the motherboard of the computer system.
- the blind plug connector is fixed on the PCB interface board of the computer system, and the interface of the motherboard is plugged into the PCB.
- the blind plug connector can adjust itself to the correct connection position based on the self-configured guidance system, thereby facilitating the connection between the high-speed layer and the motherboard of the computer system.
- the heat dissipation structure includes: a thermal pad, wherein the thermal pad is coupled to the processor (for example, the thermal pad is connected to the processor through an output terminal of the processor).
- the thermal pad is coupled to the processor, and the processor generates heat when outputting electrical signals.
- the heat generated by the processor can be quickly conducted to the heat dissipation structure, and the heat dissipation structure is used for cooling.
- a thermal pad can cover the upper layer of the chip, and the thermal pad can conduct heat to the cold plate. The cold plate conducts heat through water cooling, thereby achieving heat dissipation for the processor.
- the number of thermal pads can be the same as the number of processors included in the computing layer.
- the same thermal pad is provided in the computing layer.
- the computing layer includes multiple processors, and by arranging corresponding thermal pads above each processor, the heat dissipation efficiency of the processor can be improved.
- the heat dissipation structure has multiple heat dissipation methods.
- the heat dissipation structure adopts any one, two or three heat dissipation methods of air cooling, liquid cooling, air cooling-assisted liquid cooling.
- the heat dissipation structure is deployed in the heat dissipation layer. Through the logical layering of the computer system, the heat dissipation structure can be separated from the cable routing space. Therefore, the heat dissipation structure adopts various heat dissipation methods such as air cooling, liquid cooling, air cooling-assisted liquid cooling, etc. , can improve the heat dissipation efficiency of the processor.
- the computing layer is located between the thermal layer and the high-speed layer.
- the heat dissipation layer and the high-speed layer can be coupled to the computing layer respectively.
- the computing layer is arranged between the heat dissipation layer and the high-speed layer to solve the problem of spatial interference between heat dissipation and cable routing. This application does not limit the spatial deployment locations of the heat dissipation layer and the high-speed layer.
- the heat dissipation layer is located on the upper layer of the computing layer, and the high-speed layer is located on the lower layer of the computing layer.
- the upper layer and the lower layer refer to the relative positions.
- the high-speed layer is located on an upper layer of the computing layer, and the heat dissipation layer is located on a lower layer of the computing layer.
- the upper layer and the lower layer refer to the relative positions. From top to bottom in the computer system, they are the high-speed layer, the computing layer and the heat dissipation layer.
- the computer system implemented in a logical layered manner solves the problem of spatial interference between heat dissipation and cable routing, and improves the space utilization of the computer system.
- this application also provides a server, which includes the computer system as in the foregoing first aspect or any one of the first aspects.
- the component module of the server can also execute the component structure described in the aforementioned first aspect and various possible implementations. For details, see the aforementioned description of the first aspect and various possible implementations. illustrate.
- the computer system includes: a heat dissipation layer, a computing layer and a high-speed layer, wherein the computing layer includes a processor for generating electrical signals; the heat dissipation layer includes a heat dissipation structure, the heat dissipation structure is coupled to the processor, and the heat dissipation structure is used to To dissipate heat to the processor; the high-speed layer includes a cable module, the cable module is coupled to the processor, and the cable module is used to transmit electrical signals.
- the computer system is logically layered, the heat dissipation structure is coupled to the processor, and the cable module is coupled to the processor, thereby releasing the cable routing space in the cable module and solving the problem of cable routing space. And the problem of heat dissipation interference, improve the heat dissipation of the computer system, and improve the deployment space utilization of the processor.
- Figure 1 is a schematic three-dimensional structural diagram of a computer system provided by this application.
- Figure 2 is a schematic diagram of the logical structure of a computer system provided by this application.
- Figure 3 is a side view of a high-speed layer provided by this application.
- Figure 4 is a schematic diagram of the three-dimensional structure of a high-speed layer provided by this application.
- Figure 5 is a schematic plan view of a high-speed layer provided by this application.
- Figure 6 is a side view of a heat dissipation structure provided by the present application.
- Figure 7 is a schematic three-dimensional structural diagram of a heat dissipation structure provided by this application.
- Figure 8 is a schematic diagram of the logical structure of a computer system provided by this application.
- Figure 9 is a side view of a computing layer provided by this application.
- Figure 10 is a schematic diagram of the three-dimensional structure of a computing layer provided by this application.
- Figure 11 is an exploded view of a computer system provided by this application.
- Figure 12 is a schematic three-dimensional structural diagram of a server provided by this application.
- Computer system 100 motherboard 200 heat dissipation layer 10 computing layer 20
- Processor group 21 bracket 202 output terminal 203 of computing layer external sub-module 30131
- this application provides a computer system that adopts the architectural design idea of water and electricity logical layering to divide the computer system into a heat dissipation layer and a computing layer. and high-speed layer.
- the heat dissipation layer containing the heat dissipation structure and the high-speed layer containing the cable module are respectively placed on both sides of the computing layer containing the processor, making full use of the vertical physical space within the computer system and increasing the heat dissipation structure. area to meet the cooling needs of the increasing number of processors.
- the cable modules and heat dissipation structures are distributed on both sides of the computing layer, crosstalk between signal cables and heat dissipation space is avoided.
- Figure 1 is a three-dimensional structural schematic diagram of a computer system provided by this application.
- Figure 2 is a schematic logical structure diagram of a computer system.
- Figure 11 is an exploded view of a computer system provided by this application. As shown in Figure 1, Figure 2 and Figure 11, this application provides a computer system 100, including: a heat dissipation layer 10, a computing layer 20 and a high-speed layer 30, wherein,
- the computing layer 20 includes a processor 201 for generating electrical signals.
- the processor 201 may be a central processing unit (CPU), a graphics processing unit (GPU), a network processing unit (NPU), or a neural network processing unit. NPU), data processing unit (data processing unit, DPU) and other electronic devices used to perform data processing.
- CPU central processing unit
- GPU graphics processing unit
- NPU network processing unit
- DPU data processing unit
- XPU XPU
- the computer system 100 may include multiple processors 201 , and the multiple processors 201 may be of the same or different types.
- the number and type of the processors 201 do not limit this application.
- the following description of this application takes the connection method of a processor 201 in the computing layer 20 with the heat dissipation layer 10 and the high-speed layer 30 as an example.
- the heat dissipation layer 10 includes a heat dissipation structure 101.
- the heat dissipation structure 101 is coupled to the processor 201.
- the heat dissipation structure 101 is used to dissipate heat from the processor 201.
- the heat dissipation structure 101 can be arranged on the upper layer, lower layer or side of the processor 201.
- the computer system 100 may include one or more heat dissipation structures 101, and the heat dissipation modes of the multiple heat dissipation structures 101 may be the same or different.
- the number and type of the heat dissipation structures 101 do not limit this application.
- the following description of the present application takes the connection method between a heat dissipation structure 101 in the heat dissipation layer 10 and the computing layer 20 as an example.
- coupling includes a direct or indirect connection between two bodies.
- the high-speed layer 30 may include a cable module 301 for transmitting electrical signals, and the cable module 301 is coupled to the processor 201 .
- the computer system is logically layered in water and electricity, so that the heat dissipation layer 10, the computing layer 20 and the high-speed layer 30 can be obtained.
- the heat dissipation layer 10 and the high-speed layer 30 are deployed in the internal space of the computer system's chassis. The method is not limited.
- the computing layer 20 is located between the heat dissipation layer 10 and the high-speed layer 30 .
- the computer system is logically layered, and the heat dissipation layer 10 and the high-speed layer 30 can be coupled to the computing layer 20 respectively.
- the computing layer 20 is provided between the heat dissipation layer 10 and the high-speed layer 30, which can solve the problem of heat dissipation and cable routing. The problem of space interference.
- the spatial deployment positions of the heat dissipation layer 10 and the high-speed layer 30 are not limited in this application.
- the heat dissipation layer 10 is located on the upper layer of the computing layer 20
- the high-speed layer 30 is located on the lower layer of the computing layer 20 .
- Figure 2 takes the heat dissipation layer 10, the computing layer 20 and the high-speed layer 30 in the computer system from top to bottom as an example.
- the high-speed layer 30 is located on the upper layer of the computing layer 20
- the heat dissipation layer 10 is located on the lower layer of the computing layer 20 .
- the upper layer and the lower layer refer to the relative positions. From top to bottom in the computer system, they are the high-speed layer 30, the computing layer 20 and the heat dissipation layer 10.
- the computer system implemented in a logical layering manner solves the problem of heat dissipation and cables.
- the problem of wiring space interference improves the space utilization of computer systems.
- the heat dissipation layer 10 the computing layer 20 and the high-speed layer 30 are described in detail respectively.
- the computing layer 20 includes a processor 201.
- the processor 201 is disposed on the mainboard.
- the processor 201 can generate electrical signals.
- the processor 201 is coupled to the cable module 301 in the high-speed layer 30. , so that electrical signals can be transmitted to the high-speed layer 30, and heat is generated during the operation of the processor 201. Therefore, the processor 201 has a need for heat dissipation.
- the processor 201 is coupled to the heat dissipation structure 101 in the heat dissipation layer 10, so the processor 201 The heat generated by 201 can be conducted to the heat dissipation layer 10.
- the processor 201 also needs to communicate with other devices in the motherboard (such as memory, network card) or other devices outside the computer system. Therefore, the computing layer 20 is also connected to the high-speed layer 30 .
- the number of processors 201 included in the computing layer 20 is not limited.
- the computing layer 20 includes one processor 201 as an example, which does not limit the application.
- Figure 9 is a side view of a computing layer provided by this application
- Figure 10 is a schematic diagram of the three-dimensional structure of a computing layer provided by this application. 9 to 10 illustrate the computing layer from different perspectives. Next, take Figure 10 as an example for detailed explanation.
- the computing layer 20 includes: a processor group 21;
- the processor group includes: at least two processors 201;
- the processor set is coupled to the heat dissipation structure 101 .
- the processor group 21 shown in Figure 10 can also be called a chip module or a dual XPU chip module.
- eight processor groups 21 are used as an example, and two processes are set in each processor group 21.
- Device 201 is a chip module or a dual XPU chip module.
- Each processor group 21 has an output port 203.
- the output port 203 is coupled to the internal terminal 3011 of the high-speed layer 30.
- the output port 203 can be a UBC connector, through which the output generated by the processor 201 is connected. The electrical signal is output to the high speed layer 30.
- the processor set 21 further includes: a bracket 202;
- At least two processors 201 are fixed on the bracket 202;
- the bracket 202 is coupled to the heat dissipation structure 101 .
- the bracket 202 plays a role in fixing the processor 201 and in conducting heat.
- the computing layer 20 is located in the middle layer of the computer system. It uses a dual XPU chip module design. Two XPU management integrated circuits (ICs) on a single module share the UBC connector to realize a 16P XPU computer. System design plan.
- ICs XPU management integrated circuits
- the high-speed layer 30 may also be called a high-speed signal layer or a signal transmission layer.
- the high-speed layer 30 is used for routing high-speed cables, and the high-speed layer 30 can transmit high-speed electrical signals.
- the high-speed layer 30 can include a cable module 301.
- the cable module 301 is coupled to the processor 201.
- the processor 201 can generate electrical signals.
- the processor 201 can transmit electrical signals to the cable module 301. Signal.
- the high-speed layer 30 may include at least one cable module 301. This application does not limit the number of cable modules 301. In Figure 2, the high-speed layer 30 includes one cable module 301. As an example, it is not intended to limit the application embodiments.
- Figure 3 is a side view of a high-speed layer provided by the present application
- Figure 4 is a side view of a high-speed layer provided by the present application.
- Figure 5 is a schematic plan view of a high-speed layer provided by this application.
- the composition structure of the high-speed layer shown in FIG. 5 is used as an example.
- the cable module 301 includes an internal terminal 3011, a cable 3012, and an external terminal 3013.
- the internal terminal 3011 and the external terminal 3013 are connected through a cable.
- the internal terminal 3011 is used for the output of the processor 201.
- the terminal 203 is coupled, and the external terminal 3013 is used to couple with the motherboard of the computer system.
- the processor 201 in the computing layer 20 has an output terminal 203.
- the output terminal 203 is used to transmit the electrical signal generated by the processor 201.
- the output terminal 203 and the cable module 301 The internal terminal 3011 is coupled, for example, the internal terminal 3011 is plugged into the output terminal 203.
- the internal terminal 3011 is a male port and the output terminal is a female port, and they are plugged together through the male port and the female port.
- the output terminal 203 is plugged into the internal terminal 3011, and the electrical signal is transmitted to the cable module 301 through the internal terminal 3011.
- the electrical signal is then transmitted to the external terminal 3013 through the cable 3012, and the electrical signal is output to the external terminal 3013 through the external terminal 3013.
- Computer system motherboard Computer system motherboard.
- one end of the cable 3012 is connected to the internal terminal 3011, and the other end of the cable 3012 is connected to the external terminal 3013.
- the shape and number of the internal terminal 3011, the cable 3012, and the external terminal 3013 There are no restrictions on the type of terminals.
- the cable module 301 adopts a flexible interface design, adopts high-speed cable routing, and lays the routing in the printed circuit board (PCB) to connect different devices, and uses repeaters ( Retimer) to enhance signal strength, in this application, the connection between devices is realized through the cable module 301 outside the PCB board, which reduces the signal transmission delay caused by signal crosstalk during the signal transmission process.
- the failure rate of the repeater is 150 parts per million (ppm) to 200ppm.
- the failure rate of the cable module 301 used in this application is 30ppm to 50ppm. The failure rate of the device is greatly reduced. Therefore, the reliability of the computer system is improved.
- the internal terminal 3011 may also be called an internal connector.
- the internal terminal 3011 may be a unified bus connector (unified bus connector, UBC).
- the internal terminal 3011 can also be other types of connectors, which are not limited here.
- the external terminal 3013 can be implemented in various ways.
- the external terminal 3013 can be a SHLM connector, or Connectors such as SHDZ connectors or SHDX connectors that implement communication between the motherboard and the cable module are determined based on the interface type of the motherboard that needs to be externally connected to the high-speed layer 30.
- SHLM connector, SHDZ connector and SHDX connector respectively refer to the LM, DZ and DX type connectors developed by Coral Group (SH).
- the high-speed layer 30 is located at the bottom of the computer system.
- the cable module 301 includes internal terminals 3011, cables 3012 and external terminals 3013.
- the internal terminals 3011 are specifically UBC terminals.
- the cable module 301 includes 24 UBC terminals, 24 UBC terminals are connected to the cable 3012, and the external terminal 3013 is specifically a SHLM connector.
- the SHLM connector can also be called an external SHLM terminal.
- the cable module 301 includes 14 SHLM connectors, and the SHLM connector The device is connected to cable 3012.
- the UBC terminal transmits high-speed electrical signals to the SHLM connector through cable 3012
- the SHLM connector transmits high-speed electrical signals to the motherboard of the computer system.
- the external terminal 3013 can also be replaced with other connector types.
- the external terminal 3013 can be a UBC terminal, a SHDZ connector or an SHDX connector. etc., to meet the needs of adapting to different ecosystem interfaces.
- the external terminal 3013 can be replaced with other types of terminals to adapt to different interfaces, and the computing layer 20 is coupled to the mainboard of the computer system through the external terminal 3013.
- the external terminal 3013 is detachably connected to the main body of the cable module 301 .
- the main part of the cable module 301 is used to indicate the part of the cable module 301 excluding the external terminals.
- the main body of the cable module 301 includes external terminals 3013 , internal terminals 3011 and cables 3012 .
- the external terminal 3013 is detachably connected to the main body of the cable module 301.
- the external terminal 3013 can be separated from the cable module 301, and then the type of the external terminal 3013 can be changed to adapt to the cable module 301. Connection to device or motherboard. What is shown in the dotted frame in FIG. 4 is the external terminal 3013. The part in FIG. 4 other than the external terminal 3013 is the main part of the cable module 301.
- the cable 3012 is a detachable part, and the cable 3012 can be replaced with other cable types according to the needs of the application scenario.
- the external terminal 3013 may also include one or more external sub-modules 30131.
- the external terminal 3013 includes three external sub-modules 30131 for illustration.
- the sub-module 30131 is detachably connected to the main part of the cable module 301.
- One or more external sub-modules 30131 can be separated from the cable module 301, and the type of the external sub-module 30131 can be changed. , the connection between the adapter cable module 301 and the device or motherboard.
- the external terminal 3013 is a blind plug connector.
- the blind plug connector is a form of the external terminal 3013.
- the blind plug connector is used to couple with the motherboard of the computer system.
- the blind plug connector is fixed on the PCB interface board of the computer system, and the interface of the motherboard is plugged into On the PCB interface board, the blind plug connector can adjust itself to the correct connection position according to the self-configured guidance system, thereby facilitating the connection between the high-speed layer 30 and the motherboard of the computer system.
- the external terminal 3013 is detachably connected to the main body of the cable module 301.
- the external terminal 3013 is detachably connected to the main body of the cable module 301.
- the main body of the cable module 301 refers to the collection of other components in the cable module 301 except the external terminals 3013 .
- the external terminal 3013 can be completely detached from the main body of the cable module 301, and other external terminals 3013 can be replaced according to the needs of actual application scenarios.
- the external terminal 3013 includes a plurality of external sub-modules 30131. As shown in Figure 4, each external sub-module 30131 is detachably connected to the main body of the cable module 301.
- the main body of the cable module 301 includes a UBC terminal.
- each 2 external sub-modules The block 30131 is detachably connected to the UBC terminal.
- the external terminal 3013 has an external sub-module 30131 that can be detached from the main body of the cable module 301 and replaced with another external sub-module 30131, thus improving the external interface availability of the computer system. Adaptability. There is no need to make hardware changes in this application, and only the cable module 301 is needed to adapt to the complete machine with different interfaces and sizes.
- the cable module 301 includes one or more cables 3012.
- the cables 3012 are also used to connect internal terminals 3011.
- the internal terminals 3011 are UBC. Terminals, UBC terminals in the same group can be connected through cable 3012. Transmitting electrical signals through the cable 3012 has the advantage of being able to transmit high-speed signals with less signal loss compared to the current solution using repeaters.
- the heat dissipation layer 10 includes a heat dissipation structure.
- the heat dissipation structure is coupled to the processor 201 of the computing layer 20. Heat is generated during the operation of the processor 201. Therefore, The processor 201 has a need for heat dissipation.
- the processor 201 is coupled to the heat dissipation structure 101 in the heat dissipation layer 10 so that the heat generated by the processor 201 can be conducted to the heat dissipation layer 10 .
- the heat dissipation method of the heat dissipation structure 101 includes at least one of air cooling, liquid cooling, and liquid assisted air cooling (Liquid Assisted Air Cooling, LAAC).
- air cooling liquid cooling
- LAAC liquid assisted Air Cooling
- the heat dissipation layer 10 is a separate layer in the computer system.
- the heat dissipation layer 10 can be disposed on a single board of the computing layer 20 so that the heat dissipation layer 10 can dissipate the heat generated by the processor 201 .
- the heat dissipation method of the heat dissipation layer 10 is not limited in this application.
- air-cooling-assisted liquid cooling can also be called liquid-assisted air cooling (LAAC).
- LAAC can be used in high-performance computer systems to control the processors in the computing layer 20 of the computer system.
- 201 for heat dissipation the heat dissipation structure 101 is deployed in the heat dissipation layer 10. Through the logical layering of the computer system, the heat dissipation structure 101 can be separated from the cable routing space. Therefore, the heat dissipation structure 101 adopts air cooling, liquid cooling, and air cooling assistance.
- Various heat dissipation methods such as liquid cooling can improve the heat dissipation efficiency of the processor 201 .
- Figure 6 is a side view of a heat dissipation structure provided by the present application
- Figure 7 is a three-dimensional structure of a heat dissipation structure provided by the present application.
- Figure 8 is a schematic diagram of the logical structure of a computer system provided by this application. Figures 6 to 8 illustrate the heat dissipation structure from different perspectives. Next, Figure 8 is taken as an example for a detailed description.
- the heat dissipation structure 101 includes: a thermal pad 1011, where the thermal pad 1011 is coupled to the processor 201.
- the thermal pad 1011 is coupled to the processor 201, so that the heat generated by the processor 201 can be quickly conducted to the heat dissipation structure 101, and is cooled through the heat dissipation structure 101.
- the thermal pads 1011 can cover each chip.
- the thermal pads 1011 can conduct heat to the cold plate, and a heat-conducting liquid medium is provided in the cold plate, for example, through water.
- the heat generated by the processor 201 is conducted to achieve heat dissipation of the processor 201 .
- the shape and material parameters of the thermal pad 1011 are not limited.
- the number of thermal pads 1011 is the same as the number of processors 201 included in the computing layer 20 , and the thermal pads 1011 correspond to the processors 201 one-to-one.
- the computing layer 20 includes two processors 201 as an example.
- a corresponding thermal pad 1011 is provided above each processor 201 , thus improving the heat dissipation efficiency of the processors 201 .
- the heat dissipation structure 101 when the heat dissipation structure 101 is air-cooled auxiliary liquid cooling, the heat dissipation structure 101 also includes: a thermal pad 1011 and a heat exchanger; the thermal pad 1011 conducts heat to the cold plate, and the heat exchanger is used to conduct the heat of the cold plate, thereby transmitting the heat of the processor 201 to the outside of the computer system.
- the heat dissipation structure 101 when the heat dissipation structure 101 is air-cooled auxiliary liquid cooling, the heat dissipation structure 101 includes: an air-liquid heat exchanger 1015, a main pipeline 1013, a branch pipeline 1014, and a pump 1012; among which, the main pipeline 1013 and the branch pipeline 1014 The heat dissipation layer 10 is arranged up and down; the main pipeline 1013 and the branch pipeline 1014 are both used to connect the air-to-liquid heat exchanger 1015 and the pump 1012.
- the main pipeline 1013 and the branch pipeline 1014 both belong to water pipes, and the water pipes belong to the cooling components in the heat dissipation layer 10.
- the branch pipeline 1014 is connected to the main pipeline 1013, and the main pipeline 1013 is connected to the pump 1012 (the pump 1012 is installed in the chassis), forming a liquid circuit.
- the air-to-liquid heat exchanger 1015 can also be called an air-cooled heat exchanger. It has the advantages of centralized heat dissipation, compact structure, reduced noise, and improved energy efficiency.
- the heat dissipation layer 10 is located on the top layer of the computer system.
- LAAC technology is used to eliminate the impact of multi-chip cascading and increase the heat exchange area of the computer system.
- the cable routing space and the heat dissipation space are decoupled in the system structure, so that The heat dissipation layer 10 can support more water pipe layouts, thereby improving the system's heat dissipation capacity.
- the computer system includes: a heat dissipation layer 10, a computing layer 20 and a high-speed layer 30.
- the computing layer 20 includes a processor 201 for generating electrical signals;
- the heat dissipation layer 10 includes a heat dissipation structure 101.
- the structure 101 is coupled to the processor 201, and the heat dissipation structure 101 is used to dissipate heat from the processor 201;
- the high-speed layer 30 includes a cable module 301, the cable module 301 is coupled to the processor 201, and the cable module 301 is used for transmission. electric signal.
- the computer system is logically layered, the heat dissipation structure 101 is coupled to the processor 201, and the cable module 301 is coupled to the processor 201, thereby releasing the cable routing space in the cable module 301 and solving the problem.
- the problem of cable routing space and heat dissipation interference improves the heat dissipation of the computer system and improves the deployment space utilization of the processor 201.
- the server includes a motherboard and a computer system as shown in any one of Figures 1 to 10.
- the computer system provided by this application adopts a water and electricity logical layered system architecture design and a flexible adaptation interface design to achieve flexible configuration of multiple processors through layered system design.
- this application uses cable modules 301 to achieve Flexible interface adaptation design.
- An example is as follows.
- LAAC technology is used, and the computer system adopts a multi-layer system architecture design.
- the height space is divided into a heat dissipation layer 10, a computing layer 20, and a high-speed layer 30.
- the computing power of the computer system is increased by 100%.
- the processor is based on the industry's 8 processor (also called 8P) architecture design.
- 8P industry's 8 processor
- connection methods and other contents between the modules/units of the above-mentioned server are based on the same concept as the computer system provided by this application, and the technical effects they bring are the same as those of the previous embodiments of this application.
- the specific content can be Please refer to the descriptions in the foregoing embodiments of this application, which will not be described again here.
- the device embodiments described above are only illustrative.
- the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. , that is, it can be located in one place, or it can be distributed across multiple units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
- the device embodiments provided in this application In the drawings, the connection relationship between modules indicates that there are communication connections between them, which can be implemented as one or more buses or signal lines.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Power Engineering (AREA)
- Computer Hardware Design (AREA)
- Cooling Or The Like Of Electrical Apparatus (AREA)
Abstract
一种计算机***(100)和服务器,用于提高计算机***(100)的散热性,以及提高处理器(201)的部署空间利用率。计算机***(100)包括:散热层(10)、计算层(20)和高速层(30),其中,计算层(20)包括用于产生电信号的处理器(201);散热层(10)包括散热结构(101),散热结构(101)与处理器(201)耦接,散热结构(101)用于对处理器(201)散热;高速层(30)包括用于传输电信号的线缆模组(301),线缆模组(301)与处理器(201)耦接。
Description
本申请要求于2022年06月27日提交中国专利局、申请号为202210742690.2、发明名称为“一种计算机***”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本申请要求于2022年07月22日提交中国专利局、申请号为202210869176.5、发明名称为“一种计算机***和服务器”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本申请涉及计算机技术领域,尤其涉及一种计算机***和服务器。
随着人工智能(artificial intelligence,AI)、高性能计算(highperformancecomputing,HPC)等算力密集型应用技术的快速发展,计算机***中处理器的数量也在不断增加。与此同时,由于处理器的算力不断增强,功率也随之增大,这也对计算机***的散热能力带来了极大考验。
目前的计算机***中每个处理器通常单独部署散热器,单个处理器的换热面积有限。此外,由于计算机***内需要增加信号线缆实现处理器间和处理器与其他器件间的通信连接,在有限的计算机***的整机空间内,信号线缆与散热器之间存在干涉,而随着处理器数量不断增加,信号线缆的走线空间与散热空间冲突问题也越发严重。受限于计算机***的机箱尺寸及散热能力的约束,现有的计算机***无法满足不断增加的高功率处理器的散热需求。因此,如何提供一种满足有限空间的计算机***中不断增加的处理器的散热需求的解决方案成为亟待解决的技术问题。
发明内容
本申请提供了一种计算机***和服务器,用于提高计算机***的散热性,以及提高处理器的部署空间利用率。
第一方面,提供一种计算机***,该计算机***按照逻辑分层,可以分为如下三个层,分别是散热层、计算层和高速层,其中,计算层具有用于产生电信号的处理器,散热层具有散热结构,高速层具有用于传输电信号的线缆模组,散热结构与处理器耦接,线缆模组与处理器耦接。在该计算机***中,该散热结构可以对处理器散热,线缆模组可以传输电信号。
通过对计算机***进行逻辑分层,散热结构与处理器耦接(也可以称为耦合),线缆模组与处理器耦接,从而可以释放线缆模组中的线缆走线空间,解决线缆走线空间和散热干涉的问题,提高计算机***的散热性,提高处理器的部署空间利用率。
在一种可能的实现方式中,线缆模组包括内接端子、线缆、外接端子,内接端子和外接端子通过线缆连接,处理器具有输出端子,处理器产生的电信号由输出端口输出,内接端子与输出端子耦接,电信号通过内接端子传输至线缆,线缆再将电信号输出到外接端子,
外接端子与算机***的主板耦接。
计算层中的处理器具有输出端子,该输出端子用于输出处理器产生的电信号,该输出端子和线缆模组中的内接端子耦接,从而电信号通过内接端子传输至线缆模组中,电信号通过线缆再传输至外接端子,通过外接端子将电信号输出至计算机***的主板。
在另一种可能的实现方式中,外接端子与线缆模组的主体可拆卸连接。线缆模组的主体是指线缆模组中除外接端子以外的其它器件的集合体,例如,可将内接端子、外接端子、线缆视为线缆模组的主体。外接端子可以整体的从线缆模组的主体上拆卸掉,以根据实际应用场景的需要更换别的外接端子。另外,外接端子中包括多个子模块,每个子模块与线缆模组的主体可拆卸连接,例如线缆模组的主体包括UBC端子,则每个子模块与UBC端子可拆卸连接,例如外接端子有一个子模块可以从线缆模组的主体上拆卸,以更换为别的子模块,因此能够提升计算机***的外接接口可适配性。本申请中不需要做硬件变更,只需要使用线缆模组就能适配不同接口及尺寸的整机。
在另一种可能的实现方式中,外接端子为盲插连接器。盲插连接器是外接端子的具体实现一种形式,该盲插连接器与计算机***的主板耦接,例如,盲插连接器固定在计算机***的PCB接口板上,主板的接口插在该PCB接口板上,盲插式连接器能依据自配的引导***将自身调整到正确的连接位置,从而方便高速层与计算机***的主板的连接。
在另一种可能的实现方式中,散热结构包括:导热垫,其中,导热垫与处理器耦接(例如,导热垫通过处理器的输出端子与处理器连接)。导热垫与处理器耦接,处理器在输出电信号时会产生热量,通过导热垫,处理器产生的热量能够快速传导至散热结构中,通过散热结构进行降温。例如,导热垫可以覆盖在芯片的上层,导热垫可以传导热量至冷板,冷板通过水冷传导热量,从而可以实现对处理器的散热处理。
在另一种可能的实现方式中,导热垫的个数与计算层包括的处理器的个数可以相同,在计算机***中散热层中每设置一个导热垫,就在计算层设置与该导热垫相对应的处理器。计算层包括多个处理器,通过在每个处理器的上方都设置相应的导热垫,因此能够提高处理器的散热效率。
在另一种可能的实现方式中,散热结构的散热方式有多种,例如该散热结构采用风冷、液冷、风冷辅助液冷中的任意一种或者两种或者三种散热方式。散热结构部署在散热层中,通过对计算机***的逻辑分层,该散热结构能够与线缆走线空间相分离,因此散热结构采用风冷、液冷、风冷辅助液冷等各种散热方式,都可以提高对处理器的散热效率。
在另一种可能的实现方式中,计算层位于散热层和高速层之间。对计算机***进行逻辑分层,能够将散热层与高速层分别与计算层耦接,该计算层设置在散热层与高速层之间,解决散热与线缆走线空间干涉的问题。本申请中不限定散热层和高速层的空间部署位置。
在另一种可能的实现方式中,散热层位于计算层的上层,高速层位于计算层的下层。上层和下层指的是相对位置,后续以计算机***中从上至下依次为散热层、计算层和高速层进行示例说明,通过采用逻辑分层方式实现的计算机***,解决了散热与线缆走线空间干涉的问题,提高了计算机***的空间利用率。
在另一种可能的实现方式中,高速层位于计算层的上层,且散热层位于计算层的下层。上层和下层指的是相对位置,在计算机***中从上至下依次为高速层、计算层和散热层,
通过采用逻辑分层方式实现的计算机***,解决了散热与线缆走线空间干涉的问题,提高了计算机***的空间利用率。
第二方面,本申请还提供一种服务器,所述服务器包括如前述第一方面或第一方面中任一项的计算机***。
在本申请的第二方面中,服务器的组成模块还可以执行前述第一方面以及各种可能的实现方式中所描述的组成结构,详见前述对第一方面以及各种可能的实现方式中的说明。
从以上技术方案可以看出,本申请具有以下优点:
在本申请中,计算机***包括:散热层、计算层和高速层,其中,计算层包括用于产生电信号的处理器;散热层包括散热结构,散热结构与处理器耦接,散热结构用于对处理器散热;高速层包括线缆模组,线缆模组与处理器耦接,线缆模组用于传输电信号。本申请中对计算机***进行逻辑分层,散热结构与处理器耦接,线缆模组与处理器耦接,从而可以释放线缆模组中的线缆走线空间,解决线缆走线空间和散热干涉的问题,提高计算机***的散热性,提高处理器的部署空间利用率。
图1为本申请提供的一种计算机***的三维结构示意图;
图2为本申请提供的一种计算机***的逻辑结构示意图;
图3为本申请提供的一种高速层的侧视图;
图4为本申请提供的一种高速层的三维结构示意图;
图5为本申请提供的一种高速层的平面示意图;
图6为本申请提供的一种散热结构的侧视图;
图7为本申请提供的一种散热结构的三维结构示意图;
图8为本申请提供的一种计算机***的逻辑结构示意图;
图9为本申请提供的一种计算层的侧视图;
图10为本申请提供的一种计算层的三维结构示意图;
图11为本申请提供的一种计算机***的***图;
图12为本申请提供的一种服务器的三维结构示意图。
本申请附图中涉及到的附图标记如下:
计算机***100主板200散热层10计算层20
高速层30散热结构101处理器201线缆模组301
内接端子3011线缆3012外接端子3013导热垫1011
泵1012主管路1013支管路1014风液换热器1015
处理器组21托架202计算层的输出端子203外接子模块30131
下面结合附图,对本申请进行描述。
本申请的说明书和权利要求书及上述附图中的术语“上层”、“下层”等是用于区别类
似的对象在空间位置上的相对关系,而不必用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换,这仅仅是描述本申请中对相同属性的对象在描述时所采用的区分方式。此外,术语“包括”和“具有”以及他们的任何变形,意图在于覆盖不排他的包含,以便包含一系列单元的过程、方法、***、产品或设备不必限于那些单元,而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它单元。
为解决有限空间的计算机***中不断增加的处理器的散热需求的技术问题,本申请提供了一种计算机***,采用水电逻辑分层的架构设计思路,将计算机***内划分为散热层、计算层和高速层,将包含散热结构的散热层和包含线缆模组的高速层分别设置在包含处理器的计算层的两侧,充分利用计算机***整机内的纵向物理空间,增加了散热结构的面积,可满足不断增加的处理器数量的散热需求。另一方面,由于线缆模组和散热结构分布在计算层的两侧,避免了信号线缆和散热空间的串扰。
接下来,结合附图详细介绍本申请提供的计算机***。
图1为本申请提供的一种计算机***的三维结构示意图,图2为一种计算机***的逻辑结构示意图,图11为本申请提供的一种计算机***的***图。如图1、图2和图11所示,本申请提供一种计算机***100,包括:散热层10、计算层20和高速层30,其中,
计算层20包括用于产生电信号的处理器201。其中,处理器201可以是中央处理单元(central processing unit,CPU)、图形处理单元(graphics processing unit,GPU)、网络处理单元(network process unit,NPU)、神经网络处理单元(neural network processing unit,NPU)、数据处理单元(data processing unit,DPU)等用于执行数据处理的电子器件中的至少一种。为了便于描述,上述执行数据处理的电子器件也可以被称为XPU。
值得说明的是,计算机***100中可包括多个处理器201,且多个处理器201的类型可相同或不同,处理器201的数量和类型不构成对本申请的限定。为了便于描述,本申请的以下描述以计算层20中一个处理器201与散热层10和高速层30的连接方式为例进行说明。
散热层10包括散热结构101,散热结构101与处理器201耦接,散热结构101用于对处理器201散热;散热结构101设置方式可以是在处理器201的上层,或者下层或者侧面等。
值得说明的是,计算机***100中可包括一个或多个散热结构101,且多个散热结构101的散热方式可相同或不同,散热结构101的数量和类型不构成对本申请的限定。为了便于描述,本申请的以下描述以散热层10中一个散热结构101与计算层20的连接方式为例进行说明。此外,耦接包括两个主体间直接或间接连接。
高速层30可以包括用于传输电信号的线缆模组301,线缆模组301与处理器201耦接。
其中,本申请中对计算机***进行水电逻辑分层,从而可以得到散热层10、计算层20和高速层30,本申请中对于散热层10和高速层30在计算机***的机箱内部空间中的部署方式不做限定。
在本申请的一些实施例中,如图1和图2所示,计算层20位于散热层10和高速层30之间。
其中,对计算机***进行逻辑分层,能够将散热层10与高速层30分别与计算层20耦接,计算层20设置在散热层10和高速层30之间,能够解决散热与线缆走线空间干涉的问题。本申请中不限定散热层10和高速层30的空间部署位置。
在本申请的一些实施例中,如图1和图2所示,散热层10位于计算层20的上层,且高速层30位于计算层20的下层。
其中,上层和下层指的是相对位置,图2中以计算机***中从上至下依次为散热层10、计算层20和高速层30进行示例说明,通过采用逻辑分层方式实现的计算机***,解决了散热与线缆走线空间干涉的问题,提高了计算机***的空间利用率。
在本申请的一些实施例中,高速层30位于计算层20的上层,且散热层10位于计算层20的下层。
其中,上层和下层指的是相对位置,在计算机***中从上至下依次为高速层30、计算层20和散热层10,通过采用逻辑分层方式实现的计算机***,解决了散热与线缆走线空间干涉的问题,提高了计算机***的空间利用率。
接下来,分别对上述散热层10、计算层20和高速层30进行详细说明。
一、计算层20
如图2所示,计算层20包括处理器201,处理器201设置在主板(mainboard)中,该处理器201可以产生电信号,处理器201和高速层30中的线缆模组301耦接,从而可以向高速层30传输电信号,在该处理器201工作过程中产生热量,因此处理器201有散热的需求,该处理器201与散热层10中的散热结构101耦接,从而处理器201产生的热量能够传导至散热层10。与此同时,处理器201还需与主板中其他器件(如内存、网卡)或计算机***外的其他设备进行通信,因此,计算层20还与高速层30相连。
需要说明的是,本申请中计算层20包括的处理器201的个数不做限定,图2中以计算层20中包括一个处理器201为例,并不作为对申请的限定。
在本申请的一些实施例中,如图9和图10所示,图9为本申请提供的一种计算层的侧视图,图10为本申请提供的一种计算层的三维结构示意图,图9至图10以不同的视角对计算层进行了说明,接下来以图10为例进行详细的说明。
具体的,如图10所示,计算层20包括:处理器组21;
处理器组包括:至少两个处理器201;
处理器组与散热结构101耦接。
其中,图10所示的处理器组21又可以称为芯片模组或者双XPU芯片模组,图10中以8个处理器组21进行示例,在每个处理器组21中设置两个处理器201。
每个处理器组21具有一个输出端口203,该输出端口203与高速层30的内接端子3011耦接,例如该输出端口203可以是UBC连接器,通过该UBC连接器将处理器201产生的电信号输出至高速层30。
在本申请的一些实施例中,处理器组21还包括:托架202;
至少两个处理器201固定在托架202上;
托架202与散热结构101耦接。
其中,托架202起到对处理器201的固定作用以及导热作用。
举例说明如下,计算层20位于计算机***的中层,使用双XPU芯片模组的设计方案,单个模组上两个XPU管理集成电路(integrated circuit,IC)与UBC连接器共用,实现16P XPU的计算机***的设计方案。
二、高速层30
高速层30,也可以称为高速信号层,或者信号传输层。高速层30用于高速线缆的走线,高速层30可以传输高速的电信号。如图2所示,该高速层30可以包括线缆模组301,线缆模组301与处理器201耦接,处理器201可以产生电信号,处理器201可以向线缆模组301传输电信号。
需要说明的是,本申请中高速层30可包括至少一个线缆模组301,本申请对线缆模组301的个数不做限定,图2中以高速层30包括一个线缆模组301为例,并不作为对申请实施例的限定。
在本申请的一些实施例中,如图3、图4和图5所示,其中,图3为本申请提供的一种高速层的侧视图,图4为本申请提供的一种高速层的三维结构示意图,图5为本申请提供的一种高速层的平面示意图。后续实施例中以图5所示的高速层的组成结构进行举例说明。
如图5所示,线缆模组301包括内接端子3011、线缆3012、外接端子3013,内接端子3011和外接端子3013通过线缆连接,内接端子3011用于与处理器201的输出端子203耦接,外接端子3013用于与计算机***的主板耦接。
如图9所示的一种计算层的侧视图,计算层20中的处理器201具有输出端子203,输出端子203用于传输处理器201产生的电信号,输出端子203和线缆模组301中的内接端子3011耦接,例如内接端子3011与输出端子203插接,又如,内接端子3011为公口,输出端子为母口,通过公口和母口插接在一起。通过输出端子203与内接端子3011插接,电信号通过内接端子3011传输至线缆模组301中,电信号通过线缆3012再传输至外接端子3013,通过外接端子3013将电信号输出至计算机***的主板。
本申请中,图5所示,线缆3012的一端连接内接端子3011,线缆3012的另一端连接外接端子3013,对于内接端子3011、线缆3012、外接端子3013的形状和个数,以及端子的类型都不做限定。
本申请中,线缆模组301采用灵活接口设计的方式,采用高速线缆走线,与在印制电路板(printed circuit board,PCB)中布设走线连接不同器件,并利用中继器(Retimer)增强信号强度的方案相比,本申请中通过PCB板外部的线缆模组301实现器件间的连接,减少了信号传输过程中的信号串扰带来的信号传输延迟。另一方面,由于无需利用中继器增强电信号,也降低了计算机***的成本。其中,采用中继器的失效率为150百万分之(parts permillion,ppm)至200ppm,本申请采用线缆模组301的失效率为30ppm至50ppm,器件的失效率有极大的降低,因此提升了计算机***的可靠性。
本申请中内接端子3011也可以称为内接连接器,例如内接端子3011可以是统一总线连接器(unifiedbusconnector,UBC)。可选地,内接端子3011还可以是其它类型的连接器,此处不做限定。
外接端子3013具有多种实现方式,例如,外接端子3013可以为SHLM连接器,或者
SHDZ连接器或SHDX连接器等实现主板与线缆模组通信的连接器,具体结合高速层30需要外接的主板的接口类型确定。其中,SHLM连接器、SHDZ连接器和SHDX连接器分别指的是珊瑚组(SH)开发的LM、DZ、DX类型连接器。
作为一种示例,高速层30位于计算机***的底层,线缆模组301包括内接端子3011、线缆3012和外接端子3013,该内接端子3011具体为UBC端子,例如线缆模组301包括24个UBC端子,24个UBC端子与线缆3012连接,外接端子3013具体为SHLM连接器,SHLM连接器也可以称为对外SHLM端子,例如线缆模组301包括14个SHLM连接器,SHLM连接器与线缆3012连接。UBC端子将高速电信号通过线缆3012传输至SHLM连接器,SHLM连接器将高速电信号传输至计算机***的主板。
值得说明的是,上述举例中外接端子3013除了可以是SHLM连接器,还可以将外接端子3013进行更换其他连接器类型,例如该外接端子3013具体可以是UBC端子、或者SHDZ连接器或者SHDX连接器等,从而满足适配不同生态***接口的需求。
需要说明的是,本申请中,外接端子3013可替换为其它类型的端子,以适配不同接口,通过外接端子3013实现计算层20与计算机***的主板耦合。另外,外接端子3013与线缆模组301的主体部分为可拆卸连接。其中,线缆模组301的主体部分用于指示线缆模组301中除对外端子的部分。示例地,如图4所示,线缆模组301的主体包括外端子3013、内接端子3011和线缆3012。外接端子3013与线缆模组301的主体部分之间为可拆卸连接,即可将外接端子3013与线缆模组301之间分离,进而更换对外端子3013的类型,适配线缆模组301与器件或主板的连接。图4中虚线框所示的是为外接端子3013,图4中除该外接端子3013以外的部分为线缆模组301的主体部分。
进一步的,在本申请一些实施例中,如图4所示,线缆3012为可拆卸部分,线缆3012可以根据应用场景的需要更换为其它线缆的类型。
在本申请的一些实施例中,如图4所示,外接端子3013还可以包括一个或多个的外接子模块30131,图4中以外接端子3013包括3个外接子模块30131进行示例说明,外接子模块30131与线缆模组301的主体部分之间为可拆卸连接,即可将某个或多个的外接子模块30131与线缆模组301之间分离,进而更换外接子模块30131的类型,适配线缆模组301与器件或主板的连接。
在本申请的一些实施例中,外接端子3013为盲插连接器。
其中,盲插连接器是外接端子3013的一种形式,该盲插连接器用于与计算机***的主板耦接,例如,盲插连接器固定在计算机***的PCB接口板上,主板的接口插在该PCB接口板上,盲插式连接器能依据自配的引导***将自身调整到正确的连接位置,从而方便高速层30与计算机***的主板的连接。
在本申请的一些实施例中,外接端子3013与线缆模组301的主体可拆卸连接。
其中,外接端子3013与线缆模组301的主体可拆卸连接。线缆模组301的主体是指线缆模组301中除外接端子3013以外的其它器件的集合体。外接端子3013可以整体的从线缆模组301的主体上拆卸掉,以根据实际应用场景的需要更换别的外接端子3013。另外,外接端子3013中包括多个外接子模块30131,如图4所示,每个外接子模块30131与线缆模组301的主体可拆卸连接,例如线缆模组301的主体包括UBC端子,则每个2外接子模
块30131与UBC端子可拆卸连接,例如外接端子3013有一个外接子模块30131可以从线缆模组301的主体上拆卸,以更换为别的外接子模块30131,因此能够提升计算机***的外接接口可适配性。本申请中不需要做硬件变更,只需要使用线缆模组301就能适配不同接口及尺寸的整机。
另外,本申请中,如图5所示,线缆模组301中包括一根或者多根的线缆3012,线缆3012还用于内接端子3011间连接,例如内接端子3011具体为UBC端子,通过线缆3012可以将同一组的UBC端子连接。通过线缆3012传输电信号,相比于目前采用中继器的方案,具有能够传输高速信号,信号损失少的优点。
三、散热层10
接下来介绍本申请提供的散热层10,如图2所示,散热层10包括散热结构,该散热结构与计算层20的处理器201耦接,在该处理器201工作过程中产生热量,因此处理器201有散热的需求,该处理器201与散热层10中的散热结构101耦接,从而处理器201产生的热量能够传导至散热层10。
散热结构101的散热方式包括风冷、液冷、液冷辅助风冷散热(Liquid Assisted Air Cooling,LAAC)中至少一种。
本申请中,散热层10为计算机***中的单独一个层,该散热层10可以设置在计算层20的单板上,从而散热层10能够对处理器201产生的热量进行散热。本申请中对于散热层10的散热方式不做限定。
其中,风冷辅助液冷也可以称为液冷辅助风冷散热(liquid assisted air cooling,LAAC),LAAC可用于高性能的计算机***中,用于对计算机***中的计算层20中的处理器201进行散热。本申请中散热结构101部署在散热层10中,通过对计算机***的逻辑分层,该散热结构101能够与线缆走线空间相分离,因此散热结构101采用风冷、液冷、风冷辅助液冷等各种散热方式,都可以提高对处理器201的散热效率。
在本申请的一些实施例中,如图6、图7和图8所示,图6为本申请提供的一种散热结构的侧视图,图7为本申请提供的一种散热结构的三维结构示意图,图8为本申请提供的一种计算机***的逻辑结构示意图。图6至图8以不同的视角对散热结构进行了说明,接下来以图8为例进行详细的说明。
具体的,散热结构101包括:导热垫1011,其中,导热垫1011与处理器201耦接。
导热垫1011与处理器201耦接,从而处理器201产生的热量,能够快速传导至散热结构101中,通过散热结构101进行降温。例如,当散热结构101为风冷辅助液冷时,导热垫1011可以逐个芯片覆盖,此时,导热垫1011可以传导热量至冷板,冷板中设置有可导热的液体介质,例如,通过水传导处理器201产生的热量,进而可以实现对处理器201的散热处理。本申请中对于导热垫1011的形状以及材料参数均不作限定。
在本申请的一些实施例中,如图8所示,导热垫1011的个数与计算层20包括的处理器201的个数相同,且导热垫1011和处理器201一一对应。
其中,图8中以计算层20包括两个处理器201为例,则在每个处理器201的上方都设置相应的导热垫1011,因此能够提高处理器201的散热效率。
在本申请的一些实施例中,如图7所示,当散热结构101为风冷辅助液冷时,散热结
构101还包括:导热垫1011和换热器;导热垫1011将热量传导至冷板,换热器用于传导冷板热量,进而将处理器201的热量传到至计算机***外部。
如图7所示,当散热结构101为风冷辅助液冷时,散热结构101包括:风液换热器1015,主管路1013、支管路1014、泵1012;其中,主管路1013和支管路1014在散热层10中为上下布局;主管路1013和支管路1014,都用于连接风液换热器1015和泵1012。
其中,主管路1013和支管路1014都属于水管,水管属于散热层10中的冷却部件,当计算层20中的处理器201数量增加,水管的数量随着增加。例如,支管路1014连接至主管路1013,主管路1013连接至泵1012(该泵1012设置在机箱内),形成液体回路。风液换热器1015也可以称为风冷换热器,具有集中散热,结构紧凑,降低噪音,提升能效的优点。
举例说明如下,散热层10位于计算机***的顶层,使用LAAC技术,消除多芯片级联影响,提升计算机***的换热面积;并在***结构上将线缆走线空间与散热空间解耦,使散热层10能支持更多的水管布局,从而提高***散热能力。
通过前述实施例的举例说明可知,计算机***包括:散热层10、计算层20和高速层30,其中,计算层20包括用于产生电信号的处理器201;散热层10包括散热结构101,散热结构101与处理器201耦接,散热结构101用于对处理器201散热;高速层30包括线缆模组301,线缆模组301与处理器201耦接,线缆模组301用于传输电信号。本申请中对计算机***进行逻辑分层,散热结构101与处理器201耦接,线缆模组301与处理器201耦接,从而可以释放线缆模组301中的线缆走线空间,解决线缆走线空间和散热干涉的问题,提高计算机***的散热性,提高处理器201的部署空间利用率。
接下来介绍本申请提供的一种服务器,如图11所示,服务器包括主板和如图1至图10中任一项的计算机***。
本申请提供的计算机***采用水电逻辑分层的***架构设计,以及采用灵活适配接口设计,通过分层***设计实现多个处理器的灵活配置,另外,本申请使用线缆模组301,实现灵活接口适配设计。
举例说明如下,本申请中,使用LAAC技术,计算机***采用多层***架构设计,将高度空间分为散热层10、计算层20和高速层30,通过空间逻辑分层,将计算机***的机箱高度及深度空间极致利用,实现部署的处理器的数量翻倍。计算机***的算力提升100%,按照处理器数量计算,处理器基于业界8个处理器(也可以称为8P)的架构设计,本申请中计算机***中的处理器数量提升翻倍,算力翻倍,提高计算机***的散热性,提高处理器的部署空间利用率。
需要说明的是,上述服务器的各模块/单元之间的连接方式等内容,由于与本申请提供的计算机***基于同一构思,其带来的技术效果与本申请的前述实施例相同,具体内容可参见本申请前述所示的实施例中的叙述,此处不再赘述。
另外需说明的是,以上所描述的装置实施例仅仅是示意性的,其中作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。另外,本申请提供的装置实施例
附图中,模块之间的连接关系表示它们之间具有通信连接,具体可以实现为一条或多条总线或信号线。
通过以上的实施方式的描述,所属领域的技术人员可以清楚地了解到本申请可借助必需的通用硬件的方式来实现,当然也可以通过专用硬件包括专用集成电路、专用CPU、专用存储器、专用元器件等来实现。在上述实施例中,可以全部或部分地通过硬件、固件或者其任意组合来实现。
Claims (11)
- 一种计算机***,其特征在于,所述计算机***包括:散热层、计算层和高速层,其中,所述计算层包括用于产生电信号的处理器;所述散热层包括散热结构,所述散热结构与所述处理器耦接,所述散热结构用于对所述处理器散热;所述高速层用于传输所述电信号的线缆模组,所述线缆模组与所述处理器耦接。
- 根据权利要求1所述的计算机***,其特征在于,所述线缆模组包括内接端子、线缆、外接端子,所述内接端子和所述外接端子通过所述线缆连接,所述内接端子用于与所述处理器的输出端子耦接,所述外接端子用于与所述计算机***的主板耦接。
- 根据权要求2所述的计算机***,其特征在于,所述外接端子与所述线缆模组的主体可拆卸连接。
- 根据权利要求3所述的计算机***,其特征在于,所述外接端子为盲插连接器。
- 根据权利要求1至4中任一所述的计算机***,其特征在于,所述散热结构包括:导热垫,其中,所述导热垫与所述处理器耦接。
- 根据权利要求5所述的计算机***,其特征在于,所述导热垫的个数与所述计算层包括的处理器的个数相同,且所述导热垫和所述处理器一一对应。
- 根据权利要求5所述的计算机***,其特征在于,所述散热结构采用风冷、液冷、风冷辅助液冷中至少一种散热方式。
- 根据权利要求1至7中任一所述的计算机***,其特征在于,所述计算层位于所述散热层和所述高速层之间。
- 根据权利要求1至8中任一所述的计算机***,其特征在于,所述散热层位于所述计算层的上层,且所述高速层位于所述计算层的下层。
- 根据权利要求1至8中任一所述的计算机***,其特征在于,所述高速层位于所述计算层的上层,且所述散热层位于所述计算层的下层。
- 一种服务器,其特征在于,所述服务器包括如权利要求1至10中任一项的计算机***。
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210742690.2 | 2022-06-27 | ||
CN202210742690 | 2022-06-27 | ||
CN202210869176.5 | 2022-07-22 | ||
CN202210869176.5A CN117331410A (zh) | 2022-06-27 | 2022-07-22 | 一种计算机***和服务器 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2024001769A1 true WO2024001769A1 (zh) | 2024-01-04 |
Family
ID=89292145
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2023/100071 WO2024001769A1 (zh) | 2022-06-27 | 2023-06-14 | 一种计算机***和服务器 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN117331410A (zh) |
WO (1) | WO2024001769A1 (zh) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110069454A1 (en) * | 2009-09-24 | 2011-03-24 | International Business Machines Corporation | Liquid-cooled electronics apparatus and methods of fabrication |
CN104597973A (zh) * | 2014-12-31 | 2015-05-06 | 曙光云计算技术有限公司 | 云服务器*** |
CN106774724A (zh) * | 2016-12-30 | 2017-05-31 | 中国科学院计算技术研究所 | 一种基于水冷散热的多处理器计算机***结构及实现方法 |
US20170290202A1 (en) * | 2014-01-08 | 2017-10-05 | Tango Tech, Inc. | Multiple server architecture, server module, and cooler module systems and architecture |
CN108445997A (zh) * | 2018-02-13 | 2018-08-24 | 华为技术有限公司 | 散热***及服务器 |
CN110989787A (zh) * | 2019-12-02 | 2020-04-10 | 北京比特大陆科技有限公司 | 液冷服务器 |
-
2022
- 2022-07-22 CN CN202210869176.5A patent/CN117331410A/zh active Pending
-
2023
- 2023-06-14 WO PCT/CN2023/100071 patent/WO2024001769A1/zh unknown
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110069454A1 (en) * | 2009-09-24 | 2011-03-24 | International Business Machines Corporation | Liquid-cooled electronics apparatus and methods of fabrication |
US20170290202A1 (en) * | 2014-01-08 | 2017-10-05 | Tango Tech, Inc. | Multiple server architecture, server module, and cooler module systems and architecture |
CN104597973A (zh) * | 2014-12-31 | 2015-05-06 | 曙光云计算技术有限公司 | 云服务器*** |
CN106774724A (zh) * | 2016-12-30 | 2017-05-31 | 中国科学院计算技术研究所 | 一种基于水冷散热的多处理器计算机***结构及实现方法 |
CN108445997A (zh) * | 2018-02-13 | 2018-08-24 | 华为技术有限公司 | 散热***及服务器 |
CN110989787A (zh) * | 2019-12-02 | 2020-04-10 | 北京比特大陆科技有限公司 | 液冷服务器 |
Also Published As
Publication number | Publication date |
---|---|
CN117331410A (zh) | 2024-01-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8116332B2 (en) | Switch arbitration | |
JP3619521B2 (ja) | 情報処理装置及びその部品配置方法 | |
CN210627192U (zh) | Vpx信号处理*** | |
WO2013078828A1 (zh) | 内存液冷散热方法、装置及*** | |
US11100040B2 (en) | Modular remote direct memory access interfaces | |
CN110134205B (zh) | 一种ai计算服务器 | |
US9954295B2 (en) | Midplane interconnect system with conductor twist mitigation | |
CN110753473A (zh) | 电路板组合以及电子设备 | |
CN110134206B (zh) | 一种计算板卡 | |
WO2024001769A1 (zh) | 一种计算机***和服务器 | |
JP2005055885A (ja) | 光電子アセンブリ | |
CN115268581A (zh) | 一种高性能计算力的ai边缘服务器***架构 | |
CN115639880A (zh) | 服务器 | |
CN216930643U (zh) | 一种浸没冷却模组以及电子设备 | |
CN113741642B (zh) | 一种高密度gpu服务器 | |
CN209879377U (zh) | 一种计算板卡 | |
CN109992060B (zh) | 层叠式多路服务器***及服务器 | |
CN217847021U (zh) | 一种高性能计算力的ai边缘服务器***架构 | |
CN217983744U (zh) | 双板型智能网卡 | |
WO2023093096A1 (zh) | 一种正交***架构以及网络设备 | |
US11452208B2 (en) | Electronic devices packaged on wing boards | |
WO2024001749A1 (zh) | 一种光模块的液冷结构及光模块 | |
CN101547550B (zh) | 伺服器***和其电路板 | |
CN219611954U (zh) | 基于VPX前后IO架构的10Gbps光电转换互连*** | |
WO2024152586A1 (zh) | 一种信号传输电路及计算设备 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23829951 Country of ref document: EP Kind code of ref document: A1 |