WO2024041077A1 - 服务器及数据中心 - Google Patents

服务器及数据中心 Download PDF

Info

Publication number
WO2024041077A1
WO2024041077A1 PCT/CN2023/097797 CN2023097797W WO2024041077A1 WO 2024041077 A1 WO2024041077 A1 WO 2024041077A1 CN 2023097797 W CN2023097797 W CN 2023097797W WO 2024041077 A1 WO2024041077 A1 WO 2024041077A1
Authority
WO
WIPO (PCT)
Prior art keywords
space
server
pcie
chassis
cards
Prior art date
Application number
PCT/CN2023/097797
Other languages
English (en)
French (fr)
Inventor
亓延军
林冰凡
Original Assignee
超聚变数字技术有限公司
厦门创普云科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 超聚变数字技术有限公司, 厦门创普云科技有限公司 filed Critical 超聚变数字技术有限公司
Publication of WO2024041077A1 publication Critical patent/WO2024041077A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
    • G06F13/38—Information transfer, e.g. on bus
    • G06F13/42—Bus transfer protocol, e.g. handshake; Synchronisation
    • G06F13/4282—Bus transfer protocol, e.g. handshake; Synchronisation on a serial bus, e.g. I2C bus, SPI bus
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2213/00—Indexing scheme relating to interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
    • G06F2213/0026—PCI express

Definitions

  • This application relates to the field of server technology, specifically to a server and a data center.
  • the space layout of the existing server is unreasonable.
  • the PCIe card and the motherboard are connected through an adapter card.
  • the adapter card occupies the lower 1U space, so the PCIe card can only be configured in the upper 3U space.
  • For the PCIe card with upper power supply its power supply interface faces upward, and the cable needs to be routed above the PCle card. Since the PCle card itself already occupies the upper 3U space, there is no extra space above to accommodate the interface of the PCIe card and the power supply cable connected to the interface. This makes the server unable to support upper-powered PCIe cards.
  • the purpose of the embodiments of this application is to provide a server and a data center.
  • the server provided by the embodiment of this application can support the upper power supply PCIe card and has strong adaptability to the PCIe card.
  • embodiments of the present application provide a server.
  • the server provided by the embodiment of the present application includes a chassis and multiple PCIe cards.
  • the internal space of the chassis includes a first space and a second space arranged in the height direction.
  • the first space is located above the second space, and the PCIe cards are located in the second space.
  • at least one PCIe card has an interface facing the first space, the interface of the PCIe card is used to connect the power supply line, the first space is used to accommodate the power supply line, the height of the second space is greater than or equal to 3U, and the height of the chassis is greater than or equal to 4U.
  • the first space of the chassis is located above the second space.
  • the second space is used to accommodate the PCIe card.
  • the PCIe card has an interface facing the first space.
  • the interface of the PCIe card is used to connect the power supply line, that is, the PCIe card uses an upper power supply method.
  • the first space is used to accommodate the power supply cable, so that the server can support the upper power supply PCIe card and has strong adaptability to the PCIe card.
  • the height of the second space is greater than or equal to 3U, that is, greater than or equal to 133.35 mm, and can support FHHL PCIe cards (height less than or equal to 111.15mm) and FHFL PCIe cards (height less than or equal to 111.15mm). Supports HHHL PCIe cards (height less than or equal to 68.9mm) and HHFL PCIe cards (height less than or equal to 68.9mm). Therefore, the server provided by the embodiment of the present application can support a variety of PCIe cards and has strong adaptability to PCIe cards.
  • the interface of the PCIe card protrudes from the upper surface of the PCIe card and extends to the first space.
  • the interface of the PCIe card can be located on the upper surface, or can protrude relative to the upper surface.
  • the first space can also accommodate the interface protruding from the upper surface of the PCIe card, so that the server can support both the interface and the upper surface of the PCIe card.
  • PCIe cards on the surface can also support PCIe cards with interfaces protruding from the upper surface, further improving the server's adaptability to PCIe cards.
  • the server also includes a motherboard, which is fixed on the bottom plate of the chassis, and the PCIe card is inserted into the motherboard.
  • the server provided by the embodiment of this application can support seventeen PCIe cards and has strong computing power density.
  • the server provided by the embodiment of this application can support nineteen PCIe cards and has strong computing power density.
  • the multiple PCIe cards include at least eight FHFL double-wide GPU cards, and each FHFL double-wide GPU card occupies two slots.
  • the server includes at least eight FHFL double-width GPU cards, which have strong computing power density and can meet business scenarios such as AI training and/or inference, HPC, and graphics rendering.
  • the server further includes a power supply unit
  • the internal space of the chassis further includes a third space.
  • the third space is parallel to the second space in the horizontal direction, and the power supply unit is located in the third space.
  • the power supply unit is connected to the mainboard and used to supply power to the mainboard.
  • the third space and the second space are juxtaposed in the horizontal direction and do not occupy the height space of the second space and the first space, so that the height of the first space and the height of the second space are equal to the height of the chassis.
  • the third space is located on the left or right side of the second space; or the second space has at least two subspaces, and the third space is located between the two subspaces.
  • the server further includes a first adapter board, and there are multiple power supply units; the multiple power supply units are arranged side by side along the height direction of the chassis and are connected to the motherboard through the first adapter board.
  • multiple power supply units can be arranged along the height direction of the chassis, thereby reducing the occupation of the multiple power supply units in the width direction of the chassis and providing the third
  • the second space leaves more space in the width direction, so that the second space can have a larger width, thereby accommodating a larger number of PCIe cards and increasing the computing power density of the server.
  • the server also includes an I/O interface
  • the internal space of the chassis also includes a fourth space.
  • the fourth space is parallel to the second space in the horizontal direction, and the I/O interface is located in the fourth space.
  • the I/O interface is used to connect to external devices of the server to realize data transmission between the external devices and the server.
  • the third space and the second space are juxtaposed in the horizontal direction and do not occupy the height space of the second space and the first space, so that the height of the first space and the height of the second space are equal to the height of the chassis.
  • the server further includes a second adapter board with multiple I/O interfaces; the multiple I/O interfaces are arranged side by side along the height direction of the chassis and are connected to the motherboard through the second adapter board.
  • multiple I/O interfaces can be arranged along the height direction of the chassis, thereby reducing the occupation of multiple I/O interfaces in the width direction of the chassis. for the second space in width Leaving more space in the horizontal direction allows the second space to have a larger width, thereby accommodating a larger number of PCIe cards and increasing the computing power density of the server.
  • embodiments of the present application also provide a data center.
  • the data center provided by the embodiment of the present application includes multiple servers.
  • the server can support PCIe cards powered by upper parts and has strong adaptability to PCIe cards.
  • Figure 1 is a schematic structural diagram of a server provided by an embodiment of the present application in some embodiments;
  • Figure 2 is a schematic diagram of the internal structure of the server shown in Figure 1 in some embodiments;
  • FIG 3 is a schematic diagram of the regional distribution of the rear panel shown in Figure 1;
  • Figure 4 is a schematic diagram of the area distribution of the rear panel shown in Figure 3 in some other embodiments.
  • Figure 5 is a schematic diagram of the area distribution of the rear panel shown in Figure 3 in some embodiments.
  • Figure 6 is a schematic diagram of the area distribution in the height direction of the first space and the second space corresponding to the first area shown in Figure 3;
  • Figure 7 is a schematic diagram of the area distribution in the width direction of the first space and the second space corresponding to the first area shown in Figure 3;
  • Figure 8 is an exploded schematic diagram of part of the structure of the server shown in Figure 2;
  • Figure 9 is a schematic diagram of the partial structure shown in Figure 8 in an assembled state
  • Figure 10 is an exploded schematic diagram of part of the structure shown in Figure 1;
  • Figure 11 is a structural schematic diagram of the structure shown in Figure 10 from another perspective;
  • Figure 12 is a schematic diagram of the logical topology of the server shown in Figure 1 in some embodiments.
  • Data center A global collaborative network of specific equipment used to transmit, accelerate, display, calculate, and store data information on the Internet infrastructure.
  • a data center includes multiple servers.
  • Server A high-performance computer that runs corresponding application software in a network environment to provide shared information resources and various services to online users, and provides various shared services and other applications to end users.
  • Servers can include central processing units, memory, hard disks, various buses, etc., and have high-speed computing capabilities, long-term reliability, and powerful external data throughput capabilities.
  • Workstation Based on personal computers and distributed network computing, it is oriented to professional application fields and has powerful data computing and graphics and image processing capabilities. It is used to meet the needs of engineering design, animation production, scientific research, software development, financial management, etc. A high-performance computer designed and developed for professional fields such as information services and simulation.
  • the motherboard can also be called the main board, the system board, or the mother board.
  • Various components in the server are connected through the motherboard, which is used to transmit various electronic signals and initially process some peripheral data.
  • BIOS basic input output system
  • BIOS chip equipped with the computer's basic input and output program, post-boot self-test program and system self-startup program, it can provide the lowest and most direct hardware settings and control for the computer.
  • the BIOS chip can also provide some system parameters to the operating system.
  • CPU central processing unit, central processing unit
  • CPU central processing unit, central processing unit
  • DIMM dual inline memory module, dual in-line memory module
  • Memory sticks are used to provide running space for information processing systems and software, and are also used to temporarily store computing data in the CPU and data exchanged with external memories such as hard disks.
  • the CPU can address the memory through the data bus and perform read and write operations on the memory.
  • PCH platform controller hub, bridge chip
  • the south bridge chip is used to process low-speed signals and communicates with the CPU through the north bridge chip.
  • the north bridge chip is set close to the CPU and is mainly responsible for controlling the data exchange between the AGP (accelerated graphics port) graphics card, memory and CPU;
  • the south bridge chip is set close to the PCI (peripheral component interconnect) slot and is mainly responsible for Data exchange of floppy drives, hard drives, keyboards and add-on cards.
  • the south bridge chip and the north bridge chip are connected through the PCI bus to form a motherboard chipset architecture.
  • BMC baseboard management controller
  • BMC and PCH can implement functions such as power on and off, log acquisition, image display, and virtual media.
  • PCIe peripheral component interconnect express
  • PCIe peripheral component interconnect express
  • It is a high-speed serial computer expansion bus standard used to achieve high-speed serial point-to-point, dual-channel, high-bandwidth transmission.
  • PCIe mainly supports active power management, error reporting, end-to-end reliable transmission, and hot-plugging and other functions.
  • PCIe is a multi-layer protocol involving the conversation layer, data exchange layer and physical layer.
  • PCIe card A card that complies with the PCIe standard. You can use HHHL (half height half length), FHHL (full height half length) and FHFL (full height full length) to describe the physical size of the PCIe card. Among them, “half length” means less than or equal to 167.65mm; “full length” means less than or equal to 312.00mm; “half height” means less than or equal to 68.9mm; “full height” means less than or equal to 111.15mm.
  • Link It is a point-to-point communication channel between two PCIe ports, which can also be called an interconnect. Communication between two PCIe ports occurs over a link.
  • PCIe SW switch, switch or switch: Provides expansion or aggregation capabilities and allows more devices to be connected to a PCIe port.
  • PCIe SW acts as a packet router, identifying which path a given packet needs to take based on address or other routing information.
  • GPU graphics processing unit, graphics processor
  • graphics processor a microprocessor used to implement images and graphics-related operations.
  • GPU card It can also be called a graphics card and can include components such as GPU, video memory, circuit board, BIOS firmware, and radiator.
  • IB infiniband, infinite bandwidth: It is a computer network communication standard for high-performance computing, with extremely high throughput and extremely low latency, and is used for data interconnection between computers. IB can also be used for communication between servers and servers, servers and storage devices (such as direct storage of attachments), and servers and networks (such as LANs, WANs, and the Internet).
  • IB card A card that complies with IB standards.
  • Network card A component that can be used to implement communication between the server and the network.
  • RAID redundant arrays of independent disks, redundant array of independent disks: is a method of storing the same data in different places on multiple hard drives. By placing data on multiple hard drives, I/O processing can overlap in a balanced manner, improving I/O performance, thereby increasing data reading efficiency and achieving redundant protection of data.
  • RAID card A component used to implement the RAID method.
  • HBA host bus adapter
  • a circuit board or integrated circuit adapter used to provide I/O processing and physical connections between servers and storage devices.
  • Power supply unit It can also be called a power supply. It is an electric energy conversion power supply (different from battery-powered power supply). It is used to convert standard alternating current into low-voltage stable direct current. And provide DC power to other devices.
  • I/O input/output, input/output: usually refers to the input and output of data between the server's internal information processing system and external devices.
  • I/O interface used to connect external devices to the server’s internal information processing system.
  • Serial interface It can also be called serial interface, serial communication interface or serial communication interface. It is an expansion interface that uses serial communication.
  • Network port It can also be called a network interface, which refers to various interfaces that connect network devices to Ethernet.
  • Common network port types include RJ (registered jack)-45 interface, RJ-11 interface, FC (ferrule connector) fiber optic connector, SC (square connector) fiber optic interface, ST (stab & twist) fiber optic interface, FDDI (fiber distributed data interface, optical fiber distributed data interface), AUI (attachment unit interface, connection unit interface), BNC (bayonet nut connector, bayonet nut connector) interface, etc.
  • U is the unit that represents the external dimensions of the server and is the abbreviation of unit.
  • the specific size represented by 1U is determined by the Electronic Industries Association (EIA). 1U is 44.45 mm, 3U is 3 times 1U at 133.35 mm, and 4U is 4 times 1U at 177.8 mm.
  • EIA Electronic Industries Association
  • 1U is 44.45 mm
  • 3U is 3 times 1U at 133.35 mm
  • 4U is 4 times 1U at 177.8 mm.
  • the actual size of "U” is allowed to deviate, or the specific size of "U” can be determined by other optional server external dimension making units or organizations. This is not restricted.
  • the slot represents the space required for the component to be plugged into the slot.
  • AI artificial intelligence: a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence.
  • HPC high performance computing
  • connection in this article should be understood in a broad sense.
  • connection can be a detachable connection or a non-detachable connection; it can be a direct connection or an indirect connection through an intermediary.
  • connection in this article should also be understood in a broad sense.
  • fixing can be direct fixation or indirect fixation through an intermediary.
  • ICT information and communications technology
  • ICT products can refer to equipment that allows people and organizations to enter the digital world, including switches, routers, network transmission equipment, servers, etc.
  • ICT products can have communication functions to realize data transmission; ICT products can also have data processing functions to analyze and process data.
  • the embodiment of this application takes a server as an example for description.
  • An embodiment of the present application also provides a data center.
  • a data center can include multiple servers.
  • multiple servers can be connected through communication to form a data center network for transmitting data.
  • Figure 1 is a schematic structural diagram of the server 100 provided by the embodiment of the present application in some embodiments.
  • Figure 2 is a schematic structural diagram of the server 100 shown in Figure 1 in some embodiments.
  • the server 100 may include a rectangular parallelepiped-shaped chassis 1 .
  • the chassis 1 may include six side panels, and the six side panels may be a front panel 11 , a rear panel 12 arranged oppositely, and a top panel 13 , a bottom panel 14 , a left panel 15 and a right panel connected between the front panel 11 and the rear panel 12 . 16.
  • the top plate 13 and the bottom plate 14 are arranged oppositely, and the left panel 15 and the right panel 16 are arranged oppositely.
  • the front panel 11 , the rear panel 12 , the top panel 13 , the bottom panel 14 , the left panel 15 and the right panel 16 together enclose the internal space of the chassis 1 .
  • one or more of the six side panels of the chassis 1 may adopt an integrated plate structure, or may adopt a plate-like structure formed by laminating and/or splicing multiple plates.
  • the outer surfaces of the six side panels of the chassis 1 are the surfaces of the six side panels facing the external power supply device of the chassis 1 .
  • the outer surface of one or more of the six side panels of the chassis 1 may be flat, or may be provided with a protruding structure and/or a groove structure, but is generally flat.
  • the chassis 1 has a height H, a width W and a length L.
  • the rear panel 12 is parallel to the height direction Z and the width direction X of the chassis 1 .
  • the bottom plate 14 is parallel to the width direction X and the length direction Y of the chassis 1 .
  • the height H of the chassis 1 is the distance between the outer surface of the top plate 13 and the outer surface of the bottom plate 14 of the chassis 1.
  • the width W of the chassis 1 is the distance between the outer surface of the left panel 15 and the outer surface of the right panel 16 of the chassis 1.
  • the distance, the length L of the chassis 1 is the distance between the outer surface of the front panel 11 and the outer surface of the rear panel 12 of the chassis 1 .
  • the thickness of the six side panels of the chassis 1 is thin and can be ignored relative to the size of the chassis 1. That is, the height H, width W and length L of the chassis 1 can be approximately regarded as the height H of the internal space of the chassis 1. , width W and length L.
  • the server 100 may also include a motherboard 10 installed in the internal space of the chassis 1 and functional components connected to the motherboard 10 .
  • the mainboard 10 is fixedly installed on the base plate 14 .
  • the plane where the motherboard 10 is located is parallel to the bottom plate of the chassis 1 14. At least part of the structure is in contact with the bottom plate 14 of the chassis 1 to reduce the space between the motherboard 10 and the bottom plate 14, thereby reducing the motherboard 10's occupation of the height space of the chassis 1 and leaving enough height space for functional components.
  • the motherboard 10 can be attached to the bottom plate 14 of the chassis 1 , thereby further reducing the space between the motherboard 10 and the bottom plate 14 and leaving more height space for functional components.
  • the mainboard 10 has the function of electrically connecting electronic components and supports. Each functional component in the server 100 is connected to the mainboard 10, and various electronic signals are transmitted through the mainboard 10 to realize communication connections.
  • the motherboard 10 can be equipped with a circuit system.
  • the circuit system may include one or more components such as a BIOS chip, an I/O control chip, a keyboard and panel control switch interface, an indicator light connector, an expansion slot, a DC power supply connector of the motherboard 10 and a plug-in card, etc. .
  • the motherboard 10 can control the system memory, storage devices and other I/O devices through the circuit system.
  • the mainboard 10 may be a hard substrate.
  • the hard substrate can be a copper clad laminate (CCL).
  • the copper clad laminate can be made by impregnating the reinforcing material with resin, then covering one or both sides with copper foil and hot pressing.
  • the reinforcing material of the copper-clad laminate can be paper, fiberglass, ceramics, silicon dioxide, boron nitride, metal, resin, composite materials, etc.
  • the material of the motherboard 10 can be selected based on requirements such as demand specifications, high-speed signals, flow capacity, cost, etc., which are not strictly limited in the embodiments of the present application.
  • the mainboard 10 may also be a flexible substrate.
  • the reinforcing plate can be used to provide hard support to the flexible substrate.
  • the flexible substrate may use polyester materials such as polyimide.
  • the reinforcing plate may be metal, polymer material, inorganic non-metallic material, etc., which is not strictly limited in the embodiment of the present application.
  • functional components can include CPU, DIMM, PCH, BMC, PCIe card, PSU, I/O interface and other components.
  • PCIe cards may include RAID cards, network cards, HBAs, IB cards, GPU cards, etc.
  • GPU cards can be divided into data center GPU cards, workstation GPU cards and consumer-grade GPU cards according to their application categories.
  • Data center GPU cards, workstation GPU cards and consumer-grade GPU cards can be applied to application scenarios of data centers, workstations and consumer-grade electronic products respectively. Due to the different requirements for image data processing in application scenarios such as data centers, workstations, and consumer electronics, the performance of GPU cards used in different application scenarios is also different. In order to meet different performance requirements, the structures of GPU cards used in different application scenarios are also different, and they also have different requirements for installation space.
  • the server 100 may have more or fewer functional components than those described above, may combine two or more functional components, or may have different functional component configurations.
  • Various functional components may be implemented in hardware, software, or a combination of hardware and software including one or more signal processing and/or application specific integrated circuits.
  • FIG. 3 is a schematic diagram of the area distribution of the rear panel 12 shown in FIG. 1 .
  • the rear panel 12 of the server 100 may include a first area 121, a second area 122, and a third area 123 for respectively supporting multiple PCIe cards 1211, power supply units 1221, and I/O interfaces 1231.
  • the width W2 of the first area 121 is greater than or equal to 70% of the width W of the chassis 1
  • the height H of the first area 121 is equal to the height H of the chassis 1
  • the height H of the chassis 1 is greater than or equal to 4U, that is, greater than or equal to 177.8 mm.
  • the width W2 of the first area 121 is greater than or equal to 70% of the width W of the chassis 1, occupying a larger space, so that the first area 121 can accommodate a larger number of PCIe cards 1211, and thus Increase the computing power density of server 100.
  • the server 100 can support multiple PCIe cards 1211.
  • the first area 121 has a height as large as possible, which can meet the space requirements of PCIe cards 1211 of different designs; in addition, the width W2 of the first area 121 is large and can accommodate a larger number of PCIe cards.
  • PCIe card 1211 thereby enabling the server 100 to support multiple different types of PCIe cards 1211. Different types of PCIe cards 1211 can be suitable for different application scenarios, and multiple PCIe cards 1211 can be flexibly configured according to needs to meet diverse application scenarios.
  • the height H of the first area 121 can be equal to the height H of the chassis 1 .
  • the embodiment of the present application takes the height H of the chassis 1 as 4U as an example for description.
  • the height H of the first area 121 is also 4U.
  • the internal space of the chassis 1 includes a first space 1240 and a second space 1210.
  • the first space 1240 is located above the second space 1210.
  • the first space 1240 and the second space 1210 are the first space 1240 and the second space 1210.
  • the area 121 extends from the rear panel 12 toward the inside of the chassis 1 .
  • the internal space of the chassis 1 may also include a third space 1220 .
  • the third space 1220 is an area where the second area 122 extends from the rear panel 12 toward the inside of the chassis 1 .
  • the internal space of the chassis 1 may also include a fourth space 1230 .
  • the fourth space 1230 is an area where the third area 123 extends from the rear panel 12 toward the inside of the chassis 1 .
  • the first space 1240 may be the upper side or upper space of the internal space of the chassis 1
  • the second space 1210 may be the lower side or lower space of the internal space of the chassis 1 , or in other words, the chassis 1
  • the space accommodating the PCle card 1211 in the internal space is the second space 1210
  • the space above the second space 1210 is the first space 1240.
  • the power supply interface of at least one PCle card 1211 faces upward, that is, toward the first space 1240.
  • the space 1240 is used to accommodate the power supply line of the PCle card 1211.
  • FIG. 4 is a schematic diagram of the area distribution of the rear panel 12 shown in FIG. 3 in some other embodiments.
  • FIG. 4 is a schematic diagram of the area distribution of the rear panel 12 shown in FIG. 3 in some other embodiments. .
  • the first area 121 may be juxtaposed with other areas of the rear panel 12 in the horizontal direction.
  • the horizontal direction is parallel to the width direction X.
  • the first area 121 may be juxtaposed with the second area 122 and the third area 123 in the horizontal direction, that is, the third space 1220 and the second space 1210 corresponding to the first area 121 may be juxtaposed in the horizontal direction, and the fourth space 1230 is parallel to the second space 1210 corresponding to the first area 121 in the horizontal direction.
  • the third space 1220 and the fourth space 1230 do not occupy the height space of the second space 1210 and the first space 1240, so that the height of the second space 1210 is equal to the height of the first space 1240.
  • the degree is equal to the height H of the chassis 1, so that the first area 121 can fully utilize the height space of the chassis 1.
  • the third area 123 and the second area 122 can be located on both sides of the first area 121 respectively, that is, the third space 1220 and the fourth space 1230 can be located on the second space 1210 respectively. both sides.
  • the third space 1220 may be located on the left or right side of the second space 1210 .
  • the third area 123 and the second area 122 can also be located on the same side of the first area 121 , that is, the third space 1220 and the fourth space 1230 can also be located in the second space. Same side as 1210.
  • the first region 121 may have at least two sub-regions 124 , and the second region 122 and/or the third region 123 may also be located between two adjacent sub-regions 124 , or That is, the second space 1210 has at least two subspaces, and the third space 1220 and/or the fourth space 1230 may also be located between two adjacent subspaces.
  • the second area 122 and the third area 123 may be juxtaposed in the horizontal direction, or may be stacked and distributed along the height direction Z of the chassis 1, that is, the third space 1220 and the fourth space 1230 may be juxtaposed in the horizontal direction. They can also be stacked and distributed along the height direction Z of the chassis 1, which is not limited in the embodiment of the present application.
  • first space 1240 and the second space 1210, the third space 1220 and the fourth space 1230 corresponding to the first area 121, the second area 122 and the third area 123 will be exemplified below with reference to the drawings. sexual description.
  • FIG. 6 is a schematic diagram of the area distribution in the height direction Z of the first space 1240 and the second space 1210 corresponding to the first area 121 shown in FIG. 3 .
  • the height of the second space 1210 is greater than or equal to 3U
  • the height of the chassis 1 is greater than or equal to 4U
  • the heights of the first space 1240 and the height of the second space 1210 are equal to the height H of the chassis 1 .
  • the server 100 may also include a plurality of PCIe cards 1211 located in the second space 1210 . Multiple PCIe cards 1211 are arranged side by side along the width direction X of the chassis 1 and inserted into the motherboard 10 .
  • At least one PCIe card 1211 may have an interface 1214 facing the first space 1240, and the interface 1214 of the PCIe card 1211 is used to connect a power supply line.
  • the first space 1240 is used to accommodate power supply lines.
  • the interface 1214 of the PCIe card 1211 is connected to the motherboard 10 , and the motherboard 10 provides auxiliary power supply to the PCIe card 1211 through the interface 1214 of the PCIe card 1211 .
  • the first space 1240 of the chassis 1 is located above the second space 1210 .
  • the second space 1210 is used to accommodate the PCIe card 1211.
  • the PCIe card has an interface facing the first space 1240.
  • the interface 1214 of the PCIe card 1211 is used to connect the power supply line, that is, the PCIe card 1211 uses an upper power supply method.
  • the first space 1240 is used to accommodate the power supply line, so that the server 100 can support the upper power supply PCIe card 1211 and has strong adaptability to the PCIe card 1211.
  • the height of the second space 1210 is greater than or equal to 3U, that is, greater than or equal to 133.35 mm, and can support FHHL PCIe cards (height less than or equal to 111.15mm) and FHFL PCIe cards (height less than or equal to 111.15mm). Can support HHHL PCIe cards (height less than or equal to 68.9mm) and HHFL PCIe cards (height less than or equal to 68.9mm). Therefore, the server 100 provided by the embodiment of the present application can Supports a variety of PCIe cards 1211 and has strong adaptability to PCIe cards 1211.
  • the plurality of PCIe cards 1211 may include a GPU card 1212, and the GPU card 1212 is located in the second space 1210.
  • the GPU card 1212 that is, the interface 1214 of the PCIe card 1211 protrudes from the upper surface of the PCIe card 1211 and extends to the first space 1240 , or at least part of the interface 1214 of the PCIe card 1211 is located in the first space 1240 .
  • the schematic illustration of the position of the structure of the GPU card 1212 in FIG. 6 does not represent the actual shape of the structure of the GPU card.
  • the GPU card may have a different shape than that shown in Figure 6, which is not limited in the embodiments of the present application.
  • the power consumption of GPU card 1212 is relatively large, generally exceeding 75w, such as: 300w, 375w, 400w, etc.
  • the bottom end of the GPU card 1212 is plugged into the motherboard 10 and is powered by the motherboard 10.
  • the maximum power that the motherboard 10 can provide is 75w.
  • the GPU card 1212 needs to be auxiliary powered through an auxiliary power supply structure such as the interface 1214 to meet the requirements of the GPU card 1212. power consumption requirements.
  • the interface 1214 of the consumer-grade GPU card can be located on the upper surface of the consumer-grade GPU card, using a top charging method, and the power supply line of the consumer-grade GPU card can be accommodated in the first space 1240.
  • the interfaces of the data center GPU card and the workstation GPU card can be located at the tail of the data center GPU card and the workstation GPU card, that is, the rear surface of the data center GPU card and the workstation GPU card, using the tail charging method.
  • the server 100 can support consumer-grade GPU cards using top charging, and can also support data center GPU cards and workstation GPU cards using tail charging, thereby supporting consumer-grade GPU cards, data center GPU cards, and workstation GPUs.
  • GPU cards such as cards, with strong adaptability to be suitable for more application scenarios and meet diverse needs.
  • the second space 1210 can support an active cooling GPU card 1212 or a passive cooling GPU card 1212, so that the server 100 has strong heat dissipation capability.
  • FIG. 7 is a schematic diagram of the area distribution of the first space 1240 and the second space 1210 corresponding to the first area 121 shown in FIG. 3 in the width direction X.
  • the width W of the chassis 1 is 447 mm.
  • the width W2 of the second space 1210 may be greater than or equal to 312.9 mm, such as: 325.12 mm, 386.08 mm, etc.
  • PCIe card 1211 occupies at least one slot.
  • the FHFL double-width GPU card 1215 occupies two slots, and other types of PCIe cards 1216 occupy a single slot.
  • other types of PCIe cards may be network cards, IB cards, etc., or they may be GPU cards occupying a single slot.
  • the second space 1210 can support eight FHFL double-width GPU cards 1215 and leaves one slot for plugging in a PCIe card 1216 occupying a single slot.
  • the second space 1210 can support seventeen PCIe cards 1216 occupying a single slot.
  • PCIe card 1216 has strong computing power density, which can meet business scenarios such as AI training and/or reasoning, HPC, and graphics rendering.
  • the second space 1210 can support at least eight FHFL double-width GPU cards 1215, and leaves three slots for plugging in PCIe cards 1216 occupying a single slot; alternatively, the second space 1210 can support nine FHFL double-width GPU card 1215, and one slot is reserved for plugging in a PCIe card 1216 occupying a single slot; alternatively, the second space 1210 can support nineteen PCIe cards 1216 occupying a single slot.
  • the server 100 provided by the embodiment of the present application can support the configuration of nine FHFL double-width GPU cards 1215 and one other type of PCIe card 1216, further improving the computing power density of the server 100.
  • the GPU card 1212 may also occupy three slots or four slots, etc., which is not limited in the embodiments of this application.
  • FIG. 8 is an exploded schematic diagram of a partial structure of the server 100 shown in FIG. 2 .
  • FIG. 8 illustrates the structure of the motherboard 10 and some functional devices plugged into the motherboard 10 .
  • the motherboard 10 may include a plurality of PCIe slots 125 , and the plurality of PCIe slots 125 are arranged at intervals along the width direction X of the chassis 1 .
  • the PCIe card 1211 is connected to the motherboard 10 by being inserted into the PCIe slot 125 .
  • the PCIe card 1211 can be connected to the motherboard 10 in a direct plug-in manner, thereby reducing the distance between the PCIe card 1211 and the motherboard 10 so that the first area 121 has a height space of 4U, thereby being able to accommodate the PCIe card 1211.
  • a 1U second space 1210 is left at the top of the first space 1240 to provide installation space for the interface 1214 of the PCIe card 1211 and the power supply cable, so that the server 100 supports the upper power supply GPU card 1212 and can be applied to more application scenarios. Meet diverse needs.
  • the bottom of the GPU card 1212 can be plugged into the motherboard 10 and powered by the motherboard 10; the top can be connected to the motherboard 10 through a power supply line, and the motherboard 10 provides auxiliary power. Therefore, the server 100 provided by the embodiment of the present application can support bottom-powered and top-powered GPU cards 1212, and has strong adaptability of the GPU card 1212.
  • the distance S between two adjacent PCIe slots 125 can be approximately equal to two slots, thereby reserving sufficient space to meet the space requirements of different PCIe cards 1211 and increasing the number of slots that the server 100 can support.
  • the types of PCIe cards 1211 enable the server 100 to meet a variety of application scenarios, expand the scope of application, and have strong adaptability to the PCIe cards 1211.
  • the distance S between the plurality of PCIe slots 125 may be the distance in the width direction X between the structural centers of the two PCIe slots.
  • Figure 9 is a schematic diagram of the partial structure shown in Figure 8 in an assembled state.
  • FIG. 9 illustrates the structure of multiple power supply units 1221 , the first adapter board 1224 and the mainboard 10 in an assembled state.
  • the server 100 further includes a power supply unit 1221 and a first adapter board 1224.
  • the power supply unit 1221 is located in the third space 1220.
  • the power supply unit 1221 is connected to the mainboard 10 through the first adapter board 1224 and is used to provide power to the mainboard 10 .
  • the power supply unit 1221 may include a power supply interface 1222 and pins 1223. As shown in FIG. 2 , the power supply interface 1222 of the power supply unit 1221 is located on the rear panel 12 . The power supply interface 1222 of the power supply unit 1221 is used to connect to an external power supply device and receive current from the external power supply device. As shown in Figures 8 and 9, the pins 1223 of the power supply unit 1221 are used to connect to the first adapter board 1224. The power supply unit 1221 and The first adapter board 1224 transmits current through the pin 1223 .
  • the power supply unit 1221 can also be directly connected to the motherboard 10 , which is not limited in the embodiments of the present application.
  • the first adapter board 1224 is arranged vertically relative to the motherboard 10 and is plugged into the motherboard 10 .
  • the plurality of power supply units 1221 are arranged side by side along the height direction Z of the chassis 1 and are connected to the motherboard 10 through the first adapter board 1224 .
  • the plurality of power supply units 1221 are used to receive current from external power supply devices of the chassis 1 and transmit the current to the motherboard 10 through the first adapter board 1224 to provide power to the motherboard 10 .
  • multiple power supply units 1221 can be arranged along the height direction Z of the chassis 1 , thereby reducing the space between the multiple power supply units 1221 in the chassis.
  • the occupation of 1 in the width direction X leaves more space for the second space 1210 in the width direction computing power density.
  • the width W3 of the third space 1220 may be less than or equal to 10% of the width W of the chassis 1 .
  • the width W of the chassis 1 is 447 mm
  • the width W3 of the third space 1220 may be less than or equal to 44.7 mm, such as 40.5 mm, 39 mm, 37 mm, etc.
  • the multiple power supply units 1221 in the second area 122 are arranged along the height direction Z of the chassis 1, thereby reducing the occupation of the multiple power supply units 1221 in the width direction X of the chassis 1, so that The width W3 of the third space 1220 may be less than or equal to 10% of the width W of the chassis 1 .
  • the server 100 may include two power supply units 1221.
  • Two power supply units 1221 are arranged side by side along the height direction Z of the chassis 1 .
  • two power supply units 1221 provide power to the mainboard 10 to meet the power demand of the mainboard 10 .
  • the server 100 may also include three, five or other numbers of power supply units 1221, which is not limited in the embodiments of this application.
  • the first adapter board 1224 may be a circuit board made of the same material as the mainboard 10 , or may be a circuit board made of a different material.
  • the materials of the main board 10 and the first adapter board 1224 can be selected based on requirements such as demand specifications, high-speed signals, flow capacity, cost, etc., which are not strictly limited in the embodiments of the present application.
  • FIG. 9 also illustrates the structure of multiple I/O interfaces 1231, the second adapter board 1232, and the mainboard 10 in an assembled state.
  • the server 100 may further include a plurality of I/O interfaces 1231 and a second adapter board 1232, and the plurality of I/O interfaces 1231 are located in the fourth space 1230.
  • the I/O interface 1231 may be one or more of a VGA interface, a USB interface, a serial port, a network port, etc.
  • the number of each interface may be one or multiple, which is not limited in the embodiments of this application.
  • the I/O interface 1231 may be located on the rear panel 12 for connecting to external devices of the server 100 to implement data transmission between the external devices and the server 100 .
  • the fourth space 1230 that is, the width of the third area 123 W1 may be less than or equal to 10% of the width W of chassis 1.
  • the width W of the chassis 1 is 447 mm, then the width W1 of the fourth space 1230 may be less than or equal to 44.7 mm, such as: 40.5 mm, 39 mm, 37 mm, etc.
  • the second adapter board 1232 is arranged vertically relative to the motherboard 10 and is plugged into the motherboard 10 .
  • a plurality of I/O interfaces 1231 are provided on the side of the second adapter board 1232 .
  • Multiple I/O interfaces 1231 are arranged side by side along the height direction Z of the chassis 1 and are connected to the motherboard 10 through the second adapter board 1232 .
  • the external device is connected to the mainboard 10 via the I/O interface 1231 and the second adapter board 1232 to implement data transmission with the mainboard 10 .
  • the second adapter board 1232 may be a circuit board made of the same material as the main board 10 , or may be a circuit board made of a different material.
  • the materials of the main board 10 and the second adapter board 1232 can be selected based on requirements such as demand specifications, high-speed signals, flow capacity, cost, etc., which are not strictly limited in the embodiments of the present application.
  • the fourth space 1230 corresponding to the third area 123 may also be located at the front end of the server 100, and the I/O interface 1231 may also be located on the front panel 11, which is not limited in the embodiments of the present application.
  • FIG. 10 is an exploded schematic diagram of part of the structure shown in FIG. 1
  • FIG. 11 is a schematic structural diagram of the structure shown in FIG. 10 from another perspective.
  • Figure 10 illustrates the structure of the PCIe card 1211, the connecting piece 127 and the fixing piece 126.
  • the server 100 may further include a fixing member 126 fixed to the side wall of the chassis 1 .
  • the fixing part 126 is connected between the chassis 1 and the PCIe card 1211.
  • the PCIe card 1211 is fixed on the fixing part 126 and is fixedly installed on the chassis 1 through the fixing part 126, which reduces the shaking of the PCIe card 1211 relative to the chassis 1 and improves the reliability of the PCIe card 1211. sex.
  • the fixing member 126 may be a part of the rear panel 12 , and the PCIe card 1211 is fixed to the rear panel 12 .
  • the fixing member 126 may also be a rigid structure with certain mechanical strength, such as a rod-shaped or plate-shaped metal part, a plastic part, or a composite material structural part. Both ends of the fixing member 126 can be respectively fixed on two opposite side panels of the chassis 1, such as the left side panel and the right side panel, thereby providing support for the components fixed on the fixing member 126 and avoiding being fixed on the fixing member 126. The components shake relative to the chassis 1.
  • multiple PCIe cards 1211 are arranged along the extension direction of the fixing part 126.
  • the multiple PCIe cards 1211 may be located on the same side of the fixing part 126, or may be located on the front and rear sides of the fixing part 126. This application The embodiment does not limit this.
  • the server 100 may further include a connector 127 fixedly connected between the PCIe card 1211 and the fixing member 126 .
  • the connector 127 can be fixed to the PCIe card 1211 first, and then the connector 127 can be fixed to the fixing member 126 to achieve the fixation of the PCIe card 1211.
  • PCIe cards 1211 may have different structures. For example, different types of PCIe cards 1211 have different heights H. Some of the PCIe cards 1211 cannot match the fastener 126 .
  • the PCIe card 1211 is fixed to the fixing part 126 through the connecting part 127, which can solve the problem of PCIe card 1211
  • the structure does not match the fixing part 126 and cannot be directly fixed to the fixing part 126.
  • the distance between the fixing member 126 and the base plate 14 is 3U. Even if the height H of the PCIe card 1211 is less than 3U, it can be fixed to the fixing member 126 through the connecting member 127 .
  • the number of connectors 127 is multiple. In some embodiments, the number of connectors 127 may be equal to the number of PCIe cards 1211 , and the connectors 127 are fixedly connected between a single PCIe card 1211 and the fixing member 126 . In some other embodiments, the number of connectors 127 may also be smaller than the number of PCIe cards 1211 . For example, PCIe cards 1211 with the same height H among multiple PCIe cards 1211 can be fixed to the same connector 127 and then fixed to the fixing member 126 through the connector 127 .
  • the structure of the connecting piece 127 may also be different.
  • the structure of the connector 127 can be designed according to the structure of the PCIe card 1211 to match PCIe cards 1211 with different structures.
  • the logical topology structure of the server 100 will be exemplified below with reference to the accompanying drawings, and the transmission path of image data in the logical topology structure of the server 100 will be exemplified to illustrate the manner in which the server 100 implements the image processing function.
  • Figure 12 is a schematic diagram of the logical topology of the server 100 shown in Figure 1 in some embodiments.
  • the logical topology of the server 100 may include two CPUs, two DIMMs, PCH, BMC, two PCIe SWs, two PCIe cards, and multiple GPU cards.
  • the two CPUs may be a first CPU and a second CPU.
  • the first CPU and the second CPU are connected to each other for data transmission.
  • the CPU operates on the data and stores the operation results in the hard disk; the CPU reads data from the DIMM and can also store the data in the DIMM.
  • the CPU receives the request signal from the PCH and performs related functions.
  • two DIMMs are connected to two CPUs respectively.
  • the CPU can store data in the DIMM and read data from the DIMM.
  • the PCIe card can be a network card or an IB card.
  • PCIe cards can be connected to the network and used to receive data from the network.
  • PCH is connected to BMC.
  • the interaction between BMC and PCH is mainly some management signals, such as IPMI, USB, VGA, etc.
  • the main functions are to realize power on and off, log acquisition, image display, virtual media, etc.
  • the logical topology structure shown in Figure 12 only illustrates a possible logical topology structure of the server 100.
  • the server 100 may also include other components, and the server 100 may also have other logical topology structures.
  • the application examples do not limit this.
  • the server 100 can implement image processing functions through PCIe cards, CPUs, DIMMs, and GPU cards.
  • the PCIe card stores the received image data into the memory via the first CPU.
  • the GPU card reads image data from the DIMM through the PCIe link, performs operations on the image data, and stores the operation results in the hard disk to implement the image processing function of the server 100 .
  • the image data may also be transmitted in other components or may have other transmission paths, which is not limited in the embodiments of the present application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Power Sources (AREA)

Abstract

本申请实施例公开一种服务器及数据中心。本申请实施例提供的服务器包括机箱和多个PCIe卡，机箱的内部空间包括沿高度方向上排列的第一空间和第二空间，第一空间位于第二空间的上方，PCIe卡位于第二空间，至少一个PCIe卡具有朝向第一空间的接口，PCIe卡的接口用于连接供电线，第一空间用于容纳供电线，第二空间的高度大于或等于3U，机箱的高度大于或等于4U。本申请实施例提供的服务器能够支持上部供电的PCIe卡，对PCIe卡具有较强适配能力，能支持多种PCIe卡。

Description

服务器及数据中心
本申请要求在2022年8月24日提交中国国家知识产权局、申请号为202211019373.4，发明名称为“服务器及数据中心”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。
技术领域
本申请涉及服务器技术领域，具体涉及一种服务器及数据中心。
背景技术
现有的服务器的空间布局不合理，PCIe卡和主板之间通过转接卡连接，转接卡占据下1U空间，使得PCIe卡只能配置在上3U空间。对于上部供电的PCIe卡，其供电接口向上，需要在PCle卡的上方走线，由于PCle卡本身已经占据了上3U空间，上方没有多余空间容纳该PCIe卡的接口和与接口连接的供电线，使得服务器无法支持上部供电的PCIe卡。
发明内容
本申请实施例的目的在于提供了一种服务器及数据中心。本申请实施例提供的服务器能够支持上部供电的PCIe卡，对PCIe卡具有较强适配能力。
第一方面，本申请实施例提供一种服务器。本申请实施例提供的服务器包括机箱和多个PCIe卡，机箱的内部空间包括沿高度方向上排列的第一空间和第二空间，第一空间位于第二空间的上方，PCIe卡位于第二空间，至少一个PCIe卡具有朝向第一空间的接口，PCIe卡的接口用于连接供电线，第一空间用于容纳供电线，第二空间的高度大于或等于3U，机箱的高度大于或等于4U。
在本申请实施例中，机箱的第一空间位于第二空间的上方。第二空间用于容纳PCIe卡，PCIe卡具有朝向第一空间的接口，PCIe卡的接口用于连接供电线，也即PCIe卡采用上部供电的方式。第一空间用于容纳供电线，使得服务器能够支持上部供电的PCIe卡，对PCIe卡具有较强适配能力。
此外，第二空间的高度大于或等于3U，也即大于或等于133.35毫米，能够支持FHHL的PCIe卡(高度小于或等于111.15mm)及FHFL的PCIe卡(高度小于或等于111.15mm)，也可以支持HHHL的PCIe卡(高度小于或等于68.9mm)及HHFL的PCIe卡(高度小于或等于68.9mm)。因此，本申请实施例提供的服务器能够支持多种PCIe卡，对PCIe卡具有较强适配能力。
一些实现方式中，PCIe卡的接口凸出于PCIe卡的上表面并延伸至第一空间。
在本实现方式中，PCIe卡的接口可以位于上表面，也可以相对于上表面凸出。第一空间还能够容纳凸出于PCIe卡的上表面的接口，使得服务器既能够支持接口位于上 表面的PCIe卡，也能够支持接口相对上表面凸出的PCIe卡，进一步提升服务器对PCIe卡的适配能力。
一些实现方式中，服务器还包括主板，主板固定于机箱的底板，PCIe卡插在主板上。
在本实现方式中，由于主板至少存在部分结构与机箱的底板接触，从而为第一区域留出了较大的高度空间，使得第一区域的高度能够等于机箱的高度。
一些实现方式中，第二空间的宽度方向上至少有十七个槽位，PCIe卡至少占据一个槽位。
在本实现方式中，本申请实施例提供的服务器能够支持十七个PCIe卡，具有较强的算力密度。
一些实现方式中，第一区域的宽度方向上至多有十九个槽位。
在本实现方式中，本申请实施例提供的服务器能够支持十九个PCIe卡，具有较强的算力密度。
一些实现方式中，多个PCIe卡包括至少八个FHFL双宽GPU卡，每个FHFL双宽GPU卡占据两个槽位。
在本实现方式中，服务器包括至少八个FHFL双宽GPU卡，具有较强的算力密度，从而能够满足AI训练和/或推理、HPC、图形渲染等业务场景。
一些实现方式中，服务器还包括电源供应单元，机箱的内部空间还包括第三空间，第三空间与第二空间在水平方向上并列，电源供应单元位于第三空间。
在本实现方式中，电源供应单元与主板连接，用于为主板供电。此外，第三空间与第二空间在水平方向上并列，不会占用第二空间和第一空间的高度空间，使得第一空间的高度和第二空间的高度等于机箱的高度。
一些实现方式中，第三空间位于第二空间的左侧或右侧；或第二空间具有至少两个子空间，第三空间位于两个子空间之间。
一些实现方式中，服务器还包括第一转接板，电源供应单元为多个；多个电源供应单元沿机箱的高度方向并排设置且通过第一转接板连接到主板。
在本实现方式中，通过设置相对第一转接板，能够使得多个电源供应单元沿机箱的高度方向排布，从而减小了多个电源供应单元在机箱的宽度方向上的占用，为第二空间在宽度方向上留出更多空间，使得第二空间可以具有较大的宽度，进而容纳更多数量的PCIe卡，提升服务器的算力密度。
一些实现方式中，服务器还包括I/O接口，机箱的内部空间还包括第四空间，第四空间与第二空间在水平方向上并列，I/O接口位于第四空间。
在本实现方式中，I/O接口用于与服务器外部设备连接，实现外部设备与服务器之间的数据传输。此外，第三空间与第二空间在水平方向上并列，不会占用第二空间和第一空间的高度空间，使得第一空间的高度和第二空间的高度等于机箱的高度。
一些实现方式中，服务器还包括第二转接板，I/O接口为多个；多个I/O接口沿机箱的高度方向并排设置、且通过第二转接板连接到主板。
在本实现方式中，通过设置相对第二转接板，能够使得多个I/O接口沿机箱的高度方向排布，从而减小了多个I/O接口在机箱的宽度方向上的占用，为第二空间在宽 度方向上留出更多空间，使得第二空间可以具有较大的宽度，进而容纳更多数量的PCIe卡，提升服务器的算力密度。
第二方面，本申请实施例还提供一种数据中心。本申请实施例提供的数据中心包括多个服务器。
在本申请实施例中，服务器能够支持上部供电的PCIe卡，对PCIe卡具有较强适配能力。
附图说明
图1是本申请实施例提供的服务器在一些实施例中的结构示意图；
图2是图1所示服务器在一些实施例中的内部结构示意图；
图3是图1所示后面板的区域分布示意图；
图4是图3所示后面板在其它一些实施例中的区域分布示意图；
图5是图3所示后面板在又一些实施例中的区域分布示意图；
图6是图3所示第一区域对应的第一空间和第二空间在高度方向上的区域分布示意图；
图7是图3所示第一区域对应的第一空间和第二空间在宽度方向上的区域分布示意图；
图8是图2所示服务器的部分结构的分解示意图；
图9是图8所示的部分结构处于装配状态时的示意图；
图10是图1所示部分结构的分解示意图；
图11是图10所示结构在另一视角的结构示意图；
图12是图1所示服务器的逻辑拓扑结构在一些实施例中的示意图。
具体实施方式
为方便理解，下面先对本申请实施例所涉及的英文简写和有关技术术语进行解释和描述。
数据中心：全球协作的特定设备网络，用来在互联网基础设施上传递、加速、展示、计算、存储数据信息。数据中心包括多个服务器。
服务器(server)：在网络环境下运行相应的应用软件，为网上用户提供共享信息资源和各种服务的一种高性能计算机，为终端用户提供各种共享服务以及其它方面应用。服务器可以包括中央处理器、内存、硬盘、各种总线等，具有高速的运算能力、长时间的可靠性、强大的外部数据吞吐能力。
工作站(workstation)：以个人计算机和分布式网络计算为基础，面向专业应用领域，具备强大的数据运算与图形、图像处理能力，为满足工程设计、动画制作、科学研究、软件开发、金融管理、信息服务、模拟仿真等专业领域而设计开发的高性能计算机。
主板，又可以称为主机板(main board)、系统板(system board)、或母板(mother board)。服务器中的各个部件通过主板连接，主板用于传输各种电子信号和初步处理一些外围数据。
BIOS(basic input output system，基本输入输出系统)芯片：搭载计算机的基本输入输出的程序、开机后自检程序和系统自启动程序，能够为计算机提供最底层的、最直接的硬件设置和控制。BIOS芯片还能够向作业系统提供一些系统参数。
CPU(central processing unit，中央处理器)：作为信息处理系统的运算和控制核心，是信息处理、程序运行的执行单元。
DIMM(dual inline memory module，双列直插式存储模块)：是一种内存条。内存条用于提供信息处理系统和软件的运行空间，还用于暂时存放CPU中的运算数据以及与硬盘等外部存储器交换的数据。CPU可通过数据总线对内存寻址，对内存进行读写操作。
PCH(platform controller hub，桥片)，也可以称为南桥芯片。南桥芯片用于处理低速信号，通过北桥芯片与CPU联系。北桥芯片靠近CPU设置，主要负责控制AGP(accelerated graphics port，加速图形端口)显卡、内存与CPU之间的数据交换；南桥芯片靠近PCI(peripheral component interconnect，外围组件互连)槽设置，主要负责软驱、硬盘、键盘以及附加卡的数据交换。南桥芯片和北桥芯片通过PCI总线连接，组成一种主板芯片组架构。
BMC(baseboard management controller，基板管理控制器)，用于在机器未开机的状态下，对机器进行固件升级、查看机器设备等操作。
BMC和PCH之间传输管理类信号，比如IPMI(intelligent platform management interface，智能平台管理接口)信号、USB(universal serial bus，通用串行总线)信号、VGA(video graphics array，视频图形阵列)信号等。BMC和PCH能够实现上下电、日志获取、图像显示、虚拟媒体等功能。
PCIe(peripheral component interconnect express，外围组件快速互连)：是一种高速串行计算机扩展总线标准，用于实现高速串行点对点、双通道、高带宽传输。PCIe主要支持主动电源管理，错误报告，端对端的可靠性传输以及热插拔等功能。PCIe是一个多层协议，涉及对话层，数据交换层和物理层。
PCIe卡：符合PCIe标准的卡。可以使用HHHL(half height half length，半高半长)、FHHL(full height half length，全高半长)及FHFL(full height full length，全高全长)来描述PCIe卡的物理尺寸。其中，“半长”表示小于或等于167.65mm；“全长”表示小于或等于312.00mm；“半高”表示小于或等于68.9mm；“全高”表示小于或等于111.15mm。
链路：是两个PCIe端口之间的点对点通信通道，也可以称为互连。两个PCIe端口之间通过链路进行通信。
PCIe SW(switch，开关或交换机)：提供扩展或聚合能力，并允许更多的设备连接到一个PCIe端口。PCIe SW充当包路由器，根据地址或其他路由信息识别给定包需要走哪条路径。
GPU(graphics processing unit，图形处理器)：用于实现图像和图形相关运算工作的微处理器。
GPU卡：也可以称为显卡，可以包括GPU、显存、电路板、BIOS固件以及散热器等部件。
IB(infiniband，无限带宽)：是用于高性能计算的计算机网络通信标准，具有极高的吞吐量和极低的延迟，用于计算机与计算机之间的数据互连。IB也可以用于服务器与服务器，服务器和存储设备(比如直接存储附件)以及服务器和网络(比如局域网，广域网和互联网)之间的通信。
IB卡：符合IB标准的卡。
网卡：可以用于实现服务器和网络之间通信的部件。
RAID(redundant arrays of independent disks，独立磁盘冗余阵列):是一种把相同的数据存储在多个硬盘的不同的地方的方法。通过把数据放在多个硬盘上，I/O处理能以平衡的方式交叠，改良I/O性能，从而提升数据读取效率，并实现对数据的冗余保护。
RAID卡：用于实现RAID方法的部件。
HBA(host bus adapter，主机总线适配器)：用于在服务器和存储装置间提供I/O处理和物理连接的电路板或集成电路适配器。
电源供应单元(power supply unit，PSU)：也可以称为电源供应器，是一种电能转换类的电源(有别于电池供电类的电源)，用于将标准交流电转成低压稳定的直流电，并将直流电提供给其它的器件使用。
I/O(input/output，输入/输出)：通常指数据在服务器的内部信息处理系统和外部设备之间的输入和输出。
I/O接口：用于实现外部设备与服务器的内部信息处理系统的连接。
串口(serial interface)：也可以称为串行接口、串行通信接口或串行通讯接口，是采用串行通信方式的扩展接口。
网口：也可以称为网络接口，指网络设备与以太网连接的各种接口。常见的网口类型有RJ(registered jack)-45接口，RJ-11接口，FC(ferrule connector)光纤接头，SC(square connector)光纤接口，ST(stab & twist)光纤接口，FDDI(fiber distributed data interface，光纤分布式数据接口)，AUI(attachment unit interface，连接单元接口)，BNC(bayonet nut connector，刺刀螺母连接器)接口等。
“U”是表示服务器外部尺寸的单位，是unit的缩略语，1U所表示的具体尺寸由美国电子工业协会(electronic industries association，EIA)所决定。1U为44.45毫米，3U则是1U的3倍为133.35毫米，4U则是1U的4倍为177.8毫米。本领域技术人员可以理解的是，在实际制造过程中，“U”的实际尺寸允许有偏差，或者，“U”的具体尺寸可以由其他可选的服务器外部尺寸制定单位或组织所决定，在此不受限制。
槽位(slot)：在本申请实施例中槽位表示插接在插槽上的元件所需的空间。
AI(artificial intelligence，人工智能)：研究、开发用于模拟、延伸和扩展人的智能的理论、方法、技术及应用系统的一门新的技术科学。
HPC(high performance computing，高性能计算)：一般个人电脑无法处理的大资料量与高速运算。
下面结合本申请实施例中的附图对本申请实施例进行描述。其中，在本申请实施例的描述中，除非另有说明，“多个”是指两个或多于两个。术语“第一”、“第二” 等用词仅用于描述目的，而不能理解为暗示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征。
此外，本文中的“连接”应做广义理解，例如，“连接”可以是可拆卸地连接，也可以是不可拆卸地连接；可以是直接连接，也可以通过中间媒介间接连接。本文中的“固定”也应做广义理解，例如，“固定”可以是直接固定，也可以通过中间媒介间接固定。
再者，本申请实施例中所提到的方位用语，例如，“前”、“后”、“左”、“右”、“顶”、“底”等，仅是参考附图的方向，因此，使用的方位用语是为了更好、更清楚地说明及理解本申请实施例，而不是指示或暗指所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本申请实施例的限制。
本申请实施例提供一种ICT(information and communications technology，信息与通信技术)产品。ICT产品可以是指使人们和组织进入数字世界的设备，包括交换机、路由器、网络传送设备、服务器等。ICT产品可以具有通信功能，以实现数据的传输；ICT产品还可以具有数据处理功能，能够对数据进行分析和处理。
本申请实施例以服务器为例进行说明。
本申请实施例还提供一种数据中心。数据中心可以包括多个服务器。示例性的，多个服务器之间可以通信连接，组成数据中心网络，用于传输数据。
请参阅图1和图2，图1是本申请实施例提供的服务器100在一些实施例中的结构示意图,图2是图1所示服务器100在一些实施例中的内部结构示意图。
示例性的，服务器100可以包括长方体形状的机箱1。机箱1可以包括六个侧板，六个侧板可以是相对设置的前面板11、后面板12以及连接于前面板11和后面板12之间的顶板13、底板14、左面板15以及右面板16。顶板13和底板14相对设置，左面板15和右面板16相对设置。前面板11、后面板12、顶板13、底板14、左面板15以及右面板16共同围出机箱1的内部空间。
示例性的，机箱1的六个侧板中的一个或多个可以采用一体的板材结构，也可以采用由多个板材层叠和/或拼接成的板状结构。机箱1的六个侧板的外表面为六个侧板朝向机箱1外部供电设备的表面。机箱1的六个侧板中的一个或多个的外表面可以是平面，也可以设有凸起结构和/或凹槽结构，但大致呈平面状。
其中，机箱1具有高度H、宽度W和长度L。后面板12平行于机箱1的高度方向Z和宽度方向X。底板14平行于机箱1的宽度方向X和长度方向Y。机箱1的高度H是机箱1的顶板13的外表面和底板14的外表面之间的距离，机箱1的宽度W是机箱1的左面板15的外表面和右面板16的外表面之间的距离，机箱1的长度L是机箱1的前面板11的外表面和后面板12的外表面之间的距离。需要说明的是，机箱1的六个侧板的厚度较薄，相对机箱1的尺寸可以忽略不计，也即机箱1的高度H、宽度W和长度L可以近似视为机箱1内部空间的高度H、宽度W和长度L。
示例性的，如图2所示，服务器100还可以包括安装于机箱1内部空间的主板10以及与主板10连接的功能部件。
示例性的，主板10固定安装于底板14。主板10所在的平面平行于机箱1的底板 14，至少存在部分结构与机箱1的底板14接触，以缩减主板10与底板14之间的空间，从而减少主板10对机箱1的高度空间的占用，为功能部件留出足够大的高度空间。
示例性的，主板10可以与机箱1的底板14贴合，从而进一步缩减主板10与底板14之间的空间，为功能部件留出更大的高度空间。
其中，主板10具有电连接电子元件和支撑的功能，服务器100中的各个功能部件均与主板10连接，通过主板10传输各种电子信号，实现通信连接。主板10可以搭载电路系统。电路系统可以包括BIOS芯片、I/O控制芯片、键盘和面板控制开关接口、指示灯插接件、扩充插槽、主板10及插卡的直流电源供电接插件等元件中的一种或多种。主板10可以通过电路系统实现对系统内存、存储设备和其他I/O设备的控制。
一些实施例中，主板10可以为硬质基板。示例性的，硬质基板可以为覆铜箔层压板(copper clad laminate，CCL)，覆铜箔层压板可以是将增强材料浸以树脂之后，一面或双面覆以铜箔并经热压而制成的板状材料。示例性的，覆铜箔层压板的增强材料可以为纸、玻璃纤维、陶瓷、二氧化硅、氮化硼、金属、树脂以及复合材料等。可以基于需求规格、高速信号、通流能力、成本等方面的需求选择主板10的材料，本申请实施例对此不作严格限定。
在其它一些实施例中，主板10也可以为柔性基板。当主板10为柔性基板时，可以通过补强板对柔性基板提供硬质支撑。示例性的，柔性基板可使用聚酰亚胺等聚酯材料。补强板可以为金属、高分子材料、无机非金属材料等，本申请实施例对此不作严格限定。
其中，功能部件可以包括CPU、DIMM、PCH、BMC、PCIe卡、PSU、I/O接口等部件。
示例性的，PCIe卡可以包括RAID卡、网卡、HBA、IB卡、GPU卡等。
其中，GPU卡根据其所在的应用端类别可以分为：数据中心GPU卡、工作站GPU卡及消费级GPU卡。数据中心GPU卡、工作站GPU卡及消费级GPU卡可以分别适用于数据中心、工作站和消费级电子产品的应用场景。由于数据中心、工作站和消费级电子产品等应用场景对图像数据处理的需求不同，使得应用于不同应用场景的GPU卡的性能也不同。为了满足不同的性能需求，应用于不同应用场景的GPU卡的结构也不同，也对安装空间有不同的需求。
其中，服务器100可以具有比上文描述的更多的或者更少的功能部件，可以组合两个或多个的功能部件，或者可以具有不同的功能部件配置。各种功能部件可以在包括一个或多个信号处理和/或专用集成电路在内的硬件、软件、或硬件和软件的组合中实现。
在本申请实施例中，多个功能部件安装于机箱1的内部空间，通过对安装空间的设计，能够改变服务器100所能支持的功能部件的数量和种类，从而使得服务器100获得不同的性能。本申请实施例将结合附图对机箱1后端空间的布局进行示例性说明。
请结合参阅图1和图3，图3是图1所示后面板12的区域分布示意图。
示例性的，服务器100的后面板12可以包括第一区域121、第二区域122和第三区域123，用于分别支持多个PCIe卡1211、电源供应单元1221和I/O接口1231。
示例性的，第一区域121的宽度W2大于或等于机箱1的宽度W的70％，第一区域121的高度H等于机箱1的高度H。机箱1的高度H大于或等于4U，也即大于或等于177.8毫米。
在本申请实施例中，第一区域121的宽度W2大于或等于机箱1的宽度W的70％，占据了较大的空间，从而使得第一区域121能够容纳较多数量的PCIe卡1211，进而增加服务器100的算力密度。
此外，由于机箱1的高度H大于或等于177.8毫米，也即第一区域121的高度H大于或等于177.8毫米，能够支持FHHL的PCIe卡(高度小于或等于111.15mm)及FHFL的PCIe卡(高度小于或等于111.15mm)，也可以支持HHHL的PCIe卡(高度小于或等于68.9mm)及HHFL的PCIe卡(高度小于或等于68.9mm)。因此，本申请实施例提供的服务器100能够支持多种PCIe卡1211。
在本申请实施例中，第一区域121具有尽可能大的高度，能够满足不同设计式样的PCIe卡1211的空间需求；再加上第一区域121的宽度W2较大，能够容纳较多数量的PCIe卡1211，从而使得服务器100能够支持多个不同种类的PCIe卡1211。不同种类的PCIe卡1211能够适用于不同的应用场景，可以根据需求灵活配置多种PCIe卡1211，以满足多样化的应用场景。
此外，由于主板10至少存在部分结构与机箱1的底板14接触，从而为第一区域121留出了较大的高度空间，使得第一区域121的高度H能够等于机箱1的高度H。
本申请实施例以机箱1的高度H为4U为例进行说明，相应地，第一区域121的高度H也为4U。
示例性的，如图2所示，机箱1的内部空间包括第一空间1240和第二空间1210，第一空间1240位于第二空间1210的上方，第一空间1240和第二空间1210为第一区域121从后面板12向机箱1内侧延伸的区域。在一些可选实施例中，机箱1的内部空间还可以包括第三空间1220，第三空间1220为第二区域122从后面板12向机箱1内侧延伸的区域。在一些可选实施例中，机箱1的内部空间还可以包括第四空间1230，第四空间1230为第三区域123从后面板12向机箱1内侧延伸的区域。
本领域技术人员可以理解的是，第一空间1240可以是机箱1的内部空间的上侧或上部空间，第二空间1210可以是机箱1的内部空间的下侧或下部空间，或者说，机箱1的内部空间中容纳PCle卡1211的空间为第二空间1210，第二空间1210的上方空间为第一空间1240，至少一个PCle卡1211的供电接口朝上，即朝向第一空间1240，该第一空间1240用于容纳该PCle卡1211的供电线。
请结合参阅图3至图5，图4是图3所示后面板12在其它一些实施例中的区域分布示意图，图4是图3所示后面板12在又一些实施例中的区域分布示意图。
示例性的，第一区域121可以与后面板12的其他区域在水平方向上并列。在本申请实施例中，水平方向平行于宽度方向X。例如：第一区域121可以与第二区域122和第三区域123在水平方向上并列，也即：第三空间1220与第一区域121对应的第二空间1210在水平方向上并列，第四空间1230与第一区域121对应的第二空间1210在水平方向上并列。在本实施例中，第三空间1220和第四空间1230不占用第二空间1210和第一空间1240的高度空间，使得第二空间1210的高度和第一空间1240的高 度等于机箱1的高度H，从而使得第一区域121能够充分利用机箱1的高度空间。
一些实施例中，如图3所示，第三区域123和第二区域122可以分别位于第一区域121的两侧，也即第三空间1220和第四空间1230可以分别位于第二空间1210的两侧。第三空间1220可以位于第二空间1210的左侧或右侧。
在其他一些实施例中，如图4所示，第三区域123和第二区域122也可以位于第一区域121的同侧，也即第三空间1220和第四空间1230也可以位于第二空间1210的同侧。
在又一些实施例中，如图5所示，第一区域121可以具有至少两个子区域124，第二区域122和/或第三区域123也可以位于相邻的两个子区域124之间，也即第二空间1210具有至少两个子空间，第三空间1220和/或第四空间1230也可以位于相邻的两个子空间之间。
示例性的，第二区域122和第三区域123可以在水平方向上并列，也可以沿机箱1的高度方向Z层叠分布，也即第三空间1220和第四空间1230可以在水平方向上并列，也可以沿机箱1的高度方向Z层叠分布，本申请实施例对此不作限定。
以下将结合附图对第一区域121、第二区域122及第三区域123所对应的第一空间1240和第二空间1210、第三空间1220及第四空间1230中所包括的具体结构进行示例性说明。
首先对第一区域121对应的第一空间1240和第二空间1210中所包括的具体结构进行示例性说明。
请结合参阅图2和图6，图6是图3所示第一区域121对应的第一空间1240和第二空间1210在高度方向Z上的区域分布示意图。
示例性的，如图2所示，第二空间1210的高度大于或等于3U，机箱1的高度大于或等于4U，第一空间1240的高度和第二空间1210的高度等于机箱1的高度H。服务器100还可以包括多个PCIe卡1211，多个PCIe卡1211位于第二空间1210。多个PCIe卡1211沿机箱1的宽度方向X并排设置、且插在主板10上。
示例性的，至少一个PCIe卡1211可以具有朝向第一空间1240的接口1214，PCIe卡1211的接口1214用于连接供电线。第一空间1240用于容纳供电线。其中，PCIe卡1211的接口1214与主板10连接，由主板10通过PCIe卡1211的接口1214对PCIe卡1211进行辅助供电。
在本申请实施例中，机箱1的第一空间1240位于第二空间1210的上方。第二空间1210用于容纳PCIe卡1211，PCIe卡具有朝向第一空间1240的接口，PCIe卡1211的接口1214用于连接供电线，也即PCIe卡1211采用上部供电的方式。第一空间1240用于容纳供电线，使得服务器100能够支持上部供电的PCIe卡1211，对PCIe卡1211具有较强适配能力。
此外，第二空间1210的高度大于或等于3U，也即大于或等于133.35毫米，能够支持FHHL的PCIe卡(高度小于或等于111.15mm)及FHFL的PCIe卡(高度小于或等于111.15mm)，也可以支持HHHL的PCIe卡(高度小于或等于68.9mm)及HHFL的PCIe卡(高度小于或等于68.9mm)。因此，本申请实施例提供的服务器100能够 支持多种PCIe卡1211，对PCIe卡1211具有较强适配能力。
示例性的，多个PCIe卡1211可以包括GPU卡1212，GPU卡1212位于第二空间1210。如图6所示，GPU卡1212，也即PCIe卡1211的接口1214凸出于PCIe卡1211的上表面并延伸至第一空间1240，或者PCIe卡1211的接口1214的至少部分位于第一空间1240。需要说明的是，图6对GPU卡1212结构的位置的示意性说明，并不表示GPU卡的结构的实际形状。GPU卡可以具有与图6所示不同的形状，本申请实施例对此不作限定。
其中，GPU卡1212的功耗较大，一般会超过75w，例如：300w，375w，400w等。GPU卡1212的底端插接于主板10，由主板10供电，但主板10能够提供的最大功率为75w，需要通过接口1214等辅助供电结构对GPU卡1212进行辅助供电，以满足GPU卡1212的功耗需求。例如：现有的产品中，消费级GPU卡的接口1214可以位于消费级GPU卡的上表面，采用顶部充电的方式，消费级GPU卡的供电线可以容纳于第一空间1240。而数据中心GPU卡和工作站GPU卡的接口可以位于数据中心GPU卡和工作站GPU卡的尾部，也即数据中心GPU卡和工作站GPU卡的后表面，采用尾部充电的方式。
在本实施例中，服务器100能够支持采用顶部充电的消费级GPU卡，也能够支持采用尾部充电的数据中心GPU卡和工作站GPU卡，从而能够支持消费级GPU卡、数据中心GPU卡和工作站GPU卡等较多种类的GPU卡，具有较强的适配能力，以适用于更多的应用场景，满足多样化的需求。
示例性的，第二空间1210可以支持主动散热GPU卡1212，也可以支持被动散热GPU卡1212，使得服务器100具有较强的散热能力。
请参阅图7，图7是图3所示第一区域121对应的第一空间1240和第二空间1210在宽度方向X上的区域分布示意图。
一些实施例中，机箱1的宽度W为447毫米。相应地，第二空间1210的宽度W2可以大于或等于312.9毫米，例如：325.12毫米，386.08毫米等。
示例性的，第二空间1210的宽度方向X上至少可以有十七个槽位，单个槽位为20.32毫米，也即第二空间1210的宽度W2可以大于或等于345.44毫米。PCIe卡1211至少占据一个槽位。在现有的产品中，FHFL双宽GPU卡1215占据两个槽位，其他类型的PCIe卡1216占据单个槽位。在本申请实施例中，其他类型的PCIe卡可以是网卡或IB卡等，也可以是占据单个槽位的GPU卡。
示例性的，第二空间1210的宽度方向X上可以有十七个槽位。在本实施例中，如图7所示，第二空间1210能够支持八个FHFL双宽GPU卡1215，并留有一个槽位用于插接占据单个槽位的PCIe卡1216。或者，第二空间1210能够支持十七个占据单个槽位的PCIe卡1216。
在本实施例中，第二空间1210的宽度方向X上至少可以有十七个槽位，使得本申请实施例提供的服务器100能够支持八个FHFL双宽GPU卡1215和一个占据单个槽位的PCIe卡1216的配置，具有较强的算力密度，从而能够满足AI训练和/或推理、HPC、图形渲染等业务场景。
在其它一些实施例中，第二空间1210的宽度方向X上至多可以有十九个槽位。 在本实施例中，第二空间1210能够支持至少八个FHFL双宽GPU卡1215，并留有三个槽位用于插接占据单个槽位的PCIe卡1216；或者，第二空间1210能够支持九个FHFL双宽GPU卡1215，并留有一个槽位用于插接占据单个槽位的PCIe卡1216；或者，第二空间1210能够支持十九个占据单个槽位的PCIe卡1216。
在本实施例中，本申请实施例提供的服务器100能够支持九个FHFL双宽GPU卡1215和一个其他类型的PCIe卡1216的配置，进一步提升服务器100的算力密度。
其他一些实施例中，GPU卡1212还可以占据三个槽位或四个槽位等，本申请实施例对此不作限定。
请参阅图8，图8是图2所示服务器100的部分结构的分解示意图，图8示意出了主板10及插接于主板10的部分功能器件的结构。
示例性的，主板10可以包括多个PCIe插槽125，多个PCIe插槽125沿机箱1的宽度方向X间隔排布。PCIe卡1211通过插接在PCIe插槽125的方式与主板10连接。
示例性的，PCIe卡1211可以采用直插的方式与主板10连接，从而缩小PCIe卡1211与主板10之间的距离，使得第一区域121具有4U的高度空间，从而能够在容纳PCIe卡1211的第一空间1240的顶部留出1U的第二空间1210，为PCIe卡1211的接口1214和供电线提供安装空间，使得服务器100支持上部供电的GPU卡1212，进而能够适用于更多的应用场景，满足多样化的需求。
在本实施例中，如图6和图8所示，GPU卡1212的底部可以插接在主板10上，由主板10供电；顶部可以通过供电线与主板10连接，由主板10进行辅助供电，因此，本申请实施例提供的服务器100能够支持底部供电及顶部供电的GPU卡1212，具有较强的GPU卡1212的适配能力。
示例性的，相邻的两个多个PCIe插槽125之间的距离S可以大致等于两个槽位，从而预留充足的空间，以满足不同PCIe卡1211的空间需求，增加服务器100能够支持的PCIe卡1211的种类，使得服务器100能够满足多种应用场景，并扩大适用范围，具有较强的PCIe卡1211的适配能力。在本申请实施例中，多个PCIe插槽125之间的距离S可以是两个PCIe插槽的结构中心在宽度方向X上的距离。
以下将结合附图对第二区域122对应的第三空间1220中所包括的具体结构进行示例性说明。
请结合参阅图2、图8和图9，图9是图8所示的部分结构处于装配状态时的示意图。图9示意出了多个电源供应单元1221、第一转接板1224和主板10处于装配状态时的结构。
示例性的，服务器100还包括电源供应单元1221和第一转接板1224，电源供应单元1221位于第三空间1220。电源供应单元1221通过第一转接板1224与主板10连接，用于为主板10供电。
示例性的，电源供应单元1221可以包括供电接口1222和引脚1223。如图2所示，电源供应单元1221的供电接口1222位于后面板12，电源供应单元1221的供电接口1222用于与外部供电设备连接，接收来自外部供电设备的电流。如图8和图9所示，电源供应单元1221的引脚1223用于与第一转接板1224连接，电源供应单元1221和 第一转接板1224通过引脚1223传输电流。
在其它一些实施例中，电源供应单元1221还可以直接与主板10连接，本申请实施例对此不作限定。
一些实施例中，如图9所示，第一转接板1224相对主板10垂直设置、且插接于主板10。电源供应单元1221可以为多个，多个电源供应单元1221沿机箱1的高度方向Z并排设置、且通过第一转接板1224连接到主板10。多个电源供应单元1221用于接收来自机箱1外部供电设备的电流，并通过第一转接板1224将电流传输至主板10，向主板10供电。
在本实施例中，通过设置相对主板10垂直的第一转接板1224，能够使得多个电源供应单元1221沿机箱1的高度方向Z排布，从而减小了多个电源供应单元1221在机箱1的宽度方向X上的占用，为第二空间1210在宽度方向X上留出更多空间，使得第二空间1210可以具有较大的宽度，进而容纳更多数量的PCIe卡1211，提升服务器100的算力密度。
示例性的，如图1和图3所示，第三空间1220，也即第二区域122的宽度W3可以小于或等于机箱1的宽度W的10％。一些实施例中，机箱1的宽度W为447毫米，则第三空间1220的宽度W3可以小于或等于44.7毫米，例如：40.5毫米、39毫米、37毫米等。
在本实施例中，第二区域122中的多个电源供应单元1221沿机箱1的高度方向Z排布，从而减小了多个电源供应单元1221在机箱1的宽度方向X上的占用，使得第三空间1220的宽度W3可以小于或等于机箱1的宽度W的10％。
示例性的，服务器100可以包括两个电源供应单元1221。两个电源供应单元1221沿机箱1的高度方向Z并排设置。本实施例通过两个电源供应单元1221为主板10供电，以满足主板10的电量需求。在其它一些实施例中，服务器100还可以包括三个、五个或其他数量个电源供应单元1221，本申请实施例对此不作限定。
示例性的，第一转接板1224可以采用与主板10相同材质的电路板，也可以采用不同材质的电路板。可以基于需求规格、高速信号、通流能力、成本等方面的需求选择主板10和第一转接板1224的材料，本申请实施例对此不作严格限定。
以下将结合附图对第三区域123对应的第四空间1230中所包括的具体结构进行示例性说明。
请结合参阅图2、图8和图9。图9还示意出了多个I/O接口1231、第二转接板1232和主板10处于装配状态时的结构。
示例性的，服务器100还可以包括多个I/O接口1231和第二转接板1232，多个I/O接口1231位于第四空间1230。
示例性的，I/O接口1231可以为VGA接口、USB接口、串口、网口等中的一种或多种。每种接口的数量可以为一个，也可以为多个，本申请实施例对此不作限定。
示例性的，如图1所示，I/O接口1231可以位于后面板12，用于与服务器100外部设备连接，实现外部设备与服务器100之间的数据传输。
一些实施例中，如图1和图3所示，第四空间1230，也即第三区域123的宽度 W1可以小于或等于机箱1的宽度W的10％。一些实施例中，机箱1的宽度W为447毫米，则第四空间1230的宽度W1可以小于或等于44.7毫米，例如：40.5毫米、39毫米、37毫米等。
示例性的，如图9所示，第二转接板1232相对主板10垂直设置、且插接于主板10。多个I/O接口1231设置于第二转接板1232的侧边。多个I/O接口1231沿机箱1的高度方向Z并排设置、且通过第二转接板1232连接到主板10。外部设备经由I/O接口1231和第二转接板1232与主板10连接，实现与主板10之间的数据传输。
在本实施例中，通过设置相对主板10垂直的第二转接板1232，能够使得多个I/O接口1231沿机箱1的高度方向Z排布，从而减小了多个I/O接口1231在机箱1的宽度方向X上的占用，为第二空间1210在宽度方向X上留出更多空间，使得第二空间1210可以具有较大的宽度，进而容纳更多数量的PCIe卡1211，提升服务器100的算力密度。
示例性的，第二转接板1232可以采用与主板10相同材质的电路板，也可以采用不同材质的电路板。可以基于需求规格、高速信号、通流能力、成本等方面的需求选择主板10和第二转接板1232的材料，本申请实施例对此不作严格限定。
在其他一些实施例中，与第三区域123对应的第四空间1230也可以位于服务器100的前端，I/O接口1231也可以位于前面板11，本申请实施例对此不作限定。
请结合参阅图10和图11，图10是图1所示部分结构的分解示意图，图11是图10所示结构在另一视角的结构示意图。图10示意出了PCIe卡1211、连接件127及固定件126的结构。
示例性的，服务器100还可以包括固定件126，固定件126固定于机箱1的侧壁。固定件126连接于机箱1和PCIe卡1211之间，PCIe卡1211固定于固定件126，并通过固定件126固定安装于机箱1，减少PCIe卡1211相对机箱1的晃动，提升PCIe卡1211的可靠性。
一些实施例中，如图11所示，固定件126可以是后面板12的一部分，PCIe卡1211固定于后面板12。在其它一些实施例中，固定件126也可以是具有一定机械强度的刚性结构，例如：杆状或板状的金属件、塑料件或复合材料结构件。固定件126的两端可以分别固定于机箱1相对的两个侧板，例如：左侧板和右侧板，从而能够对固定于固定件126上的部件提供支撑，避免固定于固定件126上的部件相对机箱1发生晃动。
示例性的，多个PCIe卡1211沿固定件126的延伸方向排布，示例性的，多个PCIe卡1211可以位于固定件126的同一侧，也可以位于固定件126的前后两侧，本申请实施例对此不做限定。
一些实施例中，服务器100还可以包括连接件127，连接件127固定连接于PCIe卡1211和固定件126之间。在安装PCIe卡1211时，可以先将连接件127固定于PCIe卡1211，之后将连接件127固定于固定件126，实现PCIe卡1211的固定。
在现有的产品中，不同种类的PCIe卡1211可能具有不同结构，例如：不同种类的PCIe卡1211的高度H不同。其中部分PCIe卡1211无法与固定件126匹配。在本实施例中，通过连接件127将PCIe卡1211固定于固定件126，能够解决PCIe卡1211 的结构与固定件126不匹配，无法直接固定于固定件126的问题。例如：固定件126与底板14之间的距离为3U，即使PCIe卡1211的高度H小于3U，也可以通过连接件127固定于固定件126。
示例性的，连接件127的数量为多个。一些实施例中，连接件127的数量可以等于PCIe卡1211的数量，连接件127固定连接于单个PCIe卡1211和固定件126之间。其他一些实施例中，连接件127的数量也可以小于PCIe卡1211的数量。例如：多个PCIe卡1211中高度H相同的PCIe卡1211可以固定于同一个连接件127，再通过连接件127固定于固定件126。
示例性的，连接件127的结构也可以不同。连接件127的结构可以根据PCIe卡1211的结构进行设计，以匹配具有不同结构的PCIe卡1211。
以下将结合附图对服务器100的逻辑拓扑结构进行示例性说明，并举例说明图像数据在服务器100的逻辑拓扑结构中的传输路径，以说明服务器100实现图像处理功能的方式。
请参阅图12，图12是图1所示服务器100的逻辑拓扑结构在一些实施例中的示意图。
示例性的，服务器100的逻辑拓扑结构可以包括两个CPU、两个DIMM、PCH、BMC、两个PCIe SW、两个PCIe卡及多个GPU卡。
其中，两个CPU可以为第一CPU和第二CPU。第一CPU和第二CPU相互连接，进行数据传输。CPU对数据进行运算，并将运算结果存入硬盘；CPU从DIMM中读取数据，也可以将数据存入DIMM。CPU接收来自PCH的请求信号，并执行相关功能。
其中，两个DIMM分别与两个CPU连接，CPU能够将数据存入DIMM，也可以从DIMM中读取数据。
其中，PCIe卡可以是网卡或IB卡。PCIe卡可以与网络连接，用于接收来自网络的数据。
其中，PCH与BMC连接。BMC和PCH交互主要是一些管理类信号，比如IPMI、USB、VGA等，功能主要是实现上下电、日志获取、图像显示、虚拟媒体等。
图12中所示的逻辑拓扑结构仅仅示意出了服务器100可能具有的一种逻辑拓扑结构，在其他一些实施例中，服务器100也可以包括其他部件，服务器100也可以具有其他逻辑拓扑结构，本申请实施例对此不作限定。
以下将根据附图中所示的逻辑拓扑结构，示例性说明图像信息在服务器100的逻辑拓扑结构中的传输路径，以说明服务器100实现图像处理功能的方式。
示例性的，服务器100可以通过PCIe卡、CPU、DIMM及GPU卡实现图像处理功能。PCIe卡将接收到的图像数据经第一CPU存入内存中。GPU卡通过PCIe链路从DIMM中读取图像数据，对图像数据进行运算并将运算结果存入硬盘中，以实现服务器100的图像处理功能。
在其它一些实施例中，图像数据还可以在其他部件中传输，也可以具有其他传输路径，本申请实施例对此不做限定。
以上描述，仅为本申请实施例的具体实施方式，但本申请实施例的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请实施例揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请实施例的保护范围之内；在不冲突的情况下，本申请实施例的实施例及实施例中的特征可以相互组合。因此，本申请实施例的保护范围应以权利要求的保护范围为准。

Claims (12)

  1. 一种服务器，其特征在于，包括机箱和多个PCIe卡，所述机箱的内部空间包括沿高度方向上排列的第一空间和第二空间，所述第一空间位于所述第二空间的上方，所述PCIe卡位于所述第二空间，至少一个所述PCIe卡具有朝向所述第一空间的接口，所述PCIe卡的接口用于连接供电线，所述第一空间用于容纳所述供电线，所述第二空间的高度大于或等于3U，所述机箱的高度大于或等于4U。
  2. 如权利要求1所述的服务器，其特征在于，所述PCIe卡的接口凸出于所述PCIe卡的上表面并延伸至所述第一空间。
  3. 如权利要求1所述的服务器，其特征在于，所述服务器还包括主板，所述主板固定于所述机箱的底板，所述PCIe卡插在所述主板上。
  4. 如权利要求1所述的服务器，其特征在于，所述第二空间的宽度方向上至少有十七个槽位，所述PCIe卡至少占据一个所述槽位。
  5. 如权利要求4所述的服务器，其特征在于，所述第二空间的宽度方向上至多有十九个槽位。
  6. 如权利要求4所述的服务器，其特征在于，所述多个PCIe卡包括至少八个FHFL双宽GPU卡，每个所述FHFL双宽GPU卡占据两个所述槽位。
  7. 如权利要求3至6中任一项所述的服务器，其特征在于，所述服务器还包括电源供应单元，所述机箱的内部空间还包括第三空间，所述第三空间与所述第二空间在水平方向上并列，所述电源供应单元位于所述第三空间。
  8. 如权利要求7所述的服务器，其特征在于，所述第三空间位于所述第二空间的左侧或右侧；或所述第二空间包括至少两个子空间，所述第三空间位于两个所述子空间之间。
  9. 如权利要求7所述的服务器，其特征在于，所述服务器还包括第一转接板，所述电源供应单元为多个；多个所述电源供应单元沿所述机箱的高度方向并排设置且通过所述第一转接板连接到所述主板。
  10. 如权利要求9所述的服务器，其特征在于，所述服务器还包括I/O接口，所述机箱的内部空间还包括第四空间，所述第四空间与所述第二空间在水平方向上并列，所述I/O接口位于所述第四空间。
  11. 如权利要求10所述的服务器，其特征在于，所述服务器还包括第二转接板，所述I/O接口为多个；多个所述I/O接口沿所述机箱的高度方向并排设置、且通过所述第二转接板连接到所述主板。
  12. 一种数据中心，其特征在于，包括多个如权利要求1至11中任一项所述的服务器。
PCT/CN2023/097797 2022-08-24 2023-06-01 服务器及数据中心 WO2024041077A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211019373.4 2022-08-24
CN202211019373.4A CN115481068B (zh) 2022-08-24 2022-08-24 服务器及数据中心

Publications (1)

Publication Number Publication Date
WO2024041077A1 true WO2024041077A1 (zh) 2024-02-29

Family

ID=84421333

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/097797 WO2024041077A1 (zh) 2022-08-24 2023-06-01 服务器及数据中心

Country Status (2)

Country Link
CN (2) CN118069574A (zh)
WO (1) WO2024041077A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118069574A (zh) * 2022-08-24 2024-05-24 超聚变数字技术有限公司 服务器及数据中心

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150347345A1 (en) * 2014-04-30 2015-12-03 Cirrascale Corporation Gen3 pci-express riser
CN105607709A (zh) * 2015-07-22 2016-05-25 加弘科技咨询(上海)有限公司 电子装置
US20170024347A1 (en) * 2015-07-22 2017-01-26 Celestica Technology Consultancy (Shanghai) Co. Ltd. Electronic apparatus
CN208538046U (zh) * 2018-06-29 2019-02-22 深圳市同泰怡信息技术有限公司 基于4U机箱安装多个硬盘和PCIe卡的四路服务器
CN209543240U (zh) * 2019-04-23 2019-10-25 深圳市阿普奥云科技有限公司 4u高度的gpu服务器机箱及gpu服务器
US20190370203A1 (en) * 2018-06-01 2019-12-05 Wiwynn Corporation Switch Board for Expanding Peripheral Component Interconnect Express Compatibility
CN112148086A (zh) * 2020-09-29 2020-12-29 深圳市国鑫恒运信息安全有限公司 一种基于2u机箱的多pcie、多硬盘的服务器
CN215987126U (zh) * 2021-02-04 2022-03-08 深圳市国鑫恒运信息安全有限公司 一种优化空间设计的服务器主板
CN216145153U (zh) * 2021-08-18 2022-03-29 中移(苏州)软件技术有限公司 服务器主板、pcie设备以及服务器
CN115481068A (zh) * 2022-08-24 2022-12-16 超聚变数字技术有限公司 服务器及数据中心

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101639712A (zh) * 2008-07-31 2010-02-03 英业达股份有限公司 服务器
CN102495663B (zh) * 2011-11-24 2014-10-29 曙光信息产业股份有限公司 服务器和服务器机箱
TWI517780B (zh) * 2013-03-27 2016-01-11 廣達電腦股份有限公司 伺服器機櫃及其伺服器裝置
CN112748774A (zh) * 2019-10-31 2021-05-04 中兴通讯股份有限公司 边缘服务器机箱

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150347345A1 (en) * 2014-04-30 2015-12-03 Cirrascale Corporation Gen3 pci-express riser
CN105607709A (zh) * 2015-07-22 2016-05-25 加弘科技咨询(上海)有限公司 电子装置
US20170024347A1 (en) * 2015-07-22 2017-01-26 Celestica Technology Consultancy (Shanghai) Co. Ltd. Electronic apparatus
US20190370203A1 (en) * 2018-06-01 2019-12-05 Wiwynn Corporation Switch Board for Expanding Peripheral Component Interconnect Express Compatibility
CN208538046U (zh) * 2018-06-29 2019-02-22 深圳市同泰怡信息技术有限公司 基于4U机箱安装多个硬盘和PCIe卡的四路服务器
CN209543240U (zh) * 2019-04-23 2019-10-25 深圳市阿普奥云科技有限公司 4u高度的gpu服务器机箱及gpu服务器
CN112148086A (zh) * 2020-09-29 2020-12-29 深圳市国鑫恒运信息安全有限公司 一种基于2u机箱的多pcie、多硬盘的服务器
CN215987126U (zh) * 2021-02-04 2022-03-08 深圳市国鑫恒运信息安全有限公司 一种优化空间设计的服务器主板
CN216145153U (zh) * 2021-08-18 2022-03-29 中移(苏州)软件技术有限公司 服务器主板、pcie设备以及服务器
CN115481068A (zh) * 2022-08-24 2022-12-16 超聚变数字技术有限公司 服务器及数据中心

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
GREGG T. A., CRADDOCK D., STIGLIANI D. J., BOSCO F. E., CRUZ E. E., SCANLON M. F., SCIUTO P., BAYER G., JUNG M., RAISCH C.: "Overview of IBM zEnterprise 196 I/O Subsystem with Focus on New PCI Express Infrastructure", IBM JOURNAL OF RESEARCH AND DEVELOPMENT., INTERNATIONAL BUSINESS MACHINES CORPORATION, NEW YORK, NY., US, vol. 56, no. 1.2, 29 February 2012 (2012-02-29), US , pages 8:1 - 8:14, XP009552814, ISSN: 0018-8646, DOI: 10.1147/JRD.2011.2178278 *

Also Published As

Publication number Publication date
CN118069574A (zh) 2024-05-24
CN115481068A (zh) 2022-12-16
CN115481068B (zh) 2023-12-01

Similar Documents

Publication Publication Date Title
US10206297B2 (en) Meshed architecture rackmount storage assembly
US7821792B2 (en) Cell board interconnection architecture
US20020124128A1 (en) Server array hardware architecture and system
US8116332B2 (en) Switch arbitration
US7533210B2 (en) Virtual communication interfaces for a micro-controller
US20080259555A1 (en) Modular blade server
WO2012089128A1 (zh) 服务器、服务器组件及控制风扇转速方法
TW200401200A (en) Modular server architecture with ethernet routed across a backplane utilizing an integrated ethernet switch module
KR20020041281A (ko) 네트워크 스위치가 내장된 고밀도 멀티 서버 시스템
WO2007062593A1 (fr) Type de serveur
JP3157935U (ja) サーバー
KR20160032274A (ko) 큰 확장성 프로세서 설치에 있어서 유연한 저장 및 네트워크 프로비저닝을 위한 시스템 및 방법
WO2024041077A1 (zh) 服务器及数据中心
US20150092788A1 (en) Input-output module
CN110134206B (zh) 一种计算板卡
EP1222550A2 (en) Modular computing architecture having common communication interface
CN106371530A (zh) 服务器
WO2024045752A1 (zh) 服务器及电子设备
CN113741642B (zh) 一种高密度gpu服务器
CN216352292U (zh) 服务器主板及服务器
CN215932518U (zh) 一种云计算超融合一体机设备
CN217034655U (zh) 一种双路服务器
US20240232119A9 (en) Scaling midplane bandwidth between storage processors via network devices
CN216310674U (zh) 一种便于供电连接的双层主板及其服务器
US20240134814A1 (en) Scaling midplane bandwidth between storage processors via network devices

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: 23856169

Country of ref document: EP

Kind code of ref document: A1