server

Abstract

This application discloses a server, relating to the field of server technology. The server includes multiple power supply units, multiple motherboard computing nodes, a first bus, and a backplane. One power supply unit is disposed on one motherboard computing node, and the motherboard computing node includes a first power expansion interface. The power supply unit is electrically connected to the first power expansion interface. The first bus is electrically connected to the first power expansion interfaces of the multiple motherboard computing nodes. The backplane includes a second power expansion interface, and the first bus is also electrically connected to the second power expansion interface. The first bus acts as the power distribution hub of the server. The power supplied by the multiple power supply units is evenly distributed on the first bus before being allocated to the corresponding loads. The first bus has a simple structure, which helps to reduce costs, effectively achieve thermal management, and improve current balance.

Description

Technical Field

[0001] This application relates to the field of server technology, and more particularly to a server. Background Technology

[0002] In server architecture, the power supply structure is typically centered around the power supply board. Power distribution usually involves the power supply unit transmitting power sequentially to the motherboard, power supply board, and backplane, and finally supplying power to the fans and hard drives, among other loads. However, in high-density, multi-node server architectures, the power supply board-centric architecture suffers from three main problems: high cost, thermal management failure, and unbalanced power distribution, all of which affect server reliability. Summary of the Invention

[0003] This application provides a server based on a thermoelectric co-optimized power supply architecture to solve the problems of high cost, thermal management failure, and power imbalance in related technologies.

[0004] In this application, the server includes multiple power supply units, multiple motherboard computing nodes, a first bus, and a middle backplane.

[0005] One power supply unit is located on a motherboard compute node, which includes a first power expansion interface. The power supply unit is electrically connected to the first power expansion interface. A first bus is electrically connected to the first power expansion interfaces of multiple motherboard compute nodes. The middle backplane includes a second power expansion interface, and the first bus is also electrically connected to the second power expansion interface.

[0006] In the above embodiments of this application, the first bus is a conductive metal strip, such as a copper bus or an aluminum bus. Multiple power supply units are electrically connected to the first bus through corresponding first power expansion interfaces. The backplane is electrically connected to the first bus through a second power expansion interface. That is, the first bus acts as the power distribution hub of the server. The power supplied by the multiple power supply units is evenly distributed on the first bus before being allocated to the corresponding loads. The first bus has a simple structure, excellent conductivity, and minimal heat generation during operation. By setting up the first bus, the power board in related technologies can be eliminated, reducing hardware costs and subsequent maintenance costs. Furthermore, it helps optimize the airflow layout within the server, enhances heat dissipation efficiency, effectively achieves thermal management, and reduces the cost of the cooling system. The first bus itself has low impedance, and the power supply path from each node to the first bus is the same, which helps ensure good current sharing.

[0007] In some embodiments, the motherboard compute node further includes a third power expansion interface, and the power supply unit is also electrically connected to the third power expansion interface. The server further includes a second bus, which is electrically connected to the third power expansion interfaces of the plurality of motherboard compute nodes. The backplane also includes a fourth power expansion interface, and the second bus is also electrically connected to the fourth power expansion interface.

[0008] In some embodiments, multiple motherboard computing nodes are stacked along a first direction to form a first stack structure and a second stack structure, the first stack structure and the second stack structure are arranged along a second direction, and the first direction and the second direction intersect.

[0009] Along the third direction, the first bus and the second bus are both located on the same side of multiple motherboard computing nodes, and the third direction is perpendicular to the first direction and the second direction.

[0010] The first bus includes a first part, a second part, and a third part connected in sequence. The first part is electrically connected to the motherboard computing node and the middle backplane in the first stacked structure, and the third part is electrically connected to the motherboard computing node and the middle backplane in the second stacked structure.

[0011] The second bus includes a fourth part, a fifth part, and a sixth part connected in sequence. The fourth part is electrically connected to the motherboard computing node and the middle backplane in the first stacked structure, and the sixth part is electrically connected to the motherboard computing node and the middle backplane in the second stacked structure.

[0012] In some embodiments, both the first bus and the second bus are U-shaped.

[0013] In some embodiments, along the second direction, the first portion is located between the fourth and sixth portions, and the sixth portion is located between the first and third portions. Along the first direction, the second portion is located on the side of the fifth portion closer to the central backplate.

[0014] In some embodiments, the second portion includes a first segment and a second segment connected together, the second segment being closer to the third portion than the first segment, the first segment extending along a second direction, and the second segment bending toward the direction of the plurality of motherboard computing nodes.

[0015] The fifth part consists of the connected third and fourth segments. The fourth segment is closer to the sixth part than the third segment. The third segment extends along the second direction, while the fourth segment bends away from the multiple motherboard computing nodes.

[0016] In some embodiments, the backplane further includes a support member, which is disposed on the same side of the backplane as the second power expansion interface. Along a third direction, the support member is disposed on one side of a plurality of motherboard computing nodes. Both the first bus and the second bus are fixed to the support member and are located on the side of the support member closest to the plurality of motherboard computing nodes.

[0017] In some embodiments, the support includes an inner groove, and at least a portion of the connection between the fifth and sixth portions is located within the inner groove.

[0018] In some embodiments, the support member further includes a plurality of hollowed-out portions, which penetrate the support member along a third direction.

[0019] In some embodiments, the first power expansion interface is a bus clamp, and the second power expansion interface is a bus clamp.

[0020] In some embodiments, the motherboard computing node further includes a circuit board and a power connector, a first terminal, a second terminal, and a first power expansion interface disposed on the circuit board. Within the circuit board, the power connector is electrically connected to the first terminal, and the second terminal is electrically connected to the first power expansion interface. A power supply unit is electrically connected to the power connector. The motherboard computing node also includes a motherboard bus, one end of which is electrically connected to the first terminal, and the other end of which is electrically connected to the second terminal.

[0021] In some embodiments, the server includes four power supply units, four motherboard compute nodes, and a mid-backplane. Attached Figure Description

[0022] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A power topology diagram of a server provided in this application embodiment;

[0024] Figure 2 This application provides a schematic diagram of the structure of a server according to an embodiment of the present application.

[0025] Figure 3 A partial enlarged view of a bus provided in an embodiment of this application;

[0026] Figure 4 A partially enlarged view of a power expansion interface provided in an embodiment of this application;

[0027] Figure 5 This is a schematic diagram of the structure of a motherboard computing node provided in an embodiment of this application;

[0028] Figure 6 A partial enlarged view of a power connector provided in an embodiment of this application;

[0029] Figure 7This is a rear view of a server provided in an embodiment of this application. Detailed Implementation

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0031] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0032] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] A multi-node server refers to a hardware architecture that integrates multiple independent server nodes within a single physical chassis (or rack unit). Each node possesses independent computing, storage, and network resources (such as processors, memory, and hard drives), but shares some infrastructure (such as power supplies, cooling systems, and chassis frames), thereby enabling high-density deployment, flexible expansion, and intensive resource management. High-density multi-node servers are a high-density computing solution designed to maximize space utilization and energy efficiency in data centers. Their core architecture and advantages can be summarized as: modular independent design, efficient heat dissipation and redundant power supplies, and intelligent power distribution and interconnect architecture.

[0034] The modular, independent design means that each node is a complete computing unit, typically including core computing components (such as a central processing unit, CPU), a storage subsystem, a network subsystem, a management subsystem, and a power supply and cooling system. It supports independent operation or collaborative work, flexibly adapting to different workloads, such as virtualization, high-performance computing clusters, and storage clusters. Each computing unit's front end is usually hot-swappable, supporting online maintenance and rapid replacement, which helps reduce operational complexity.

[0035] The high-efficiency heat dissipation function is based on its multi-level heat dissipation system. For example, the front high-efficiency fan wall and node-level auxiliary fans support air cooling and cold plate heat dissipation configurations, which helps to ensure stable operation of the system under high load.

[0036] Power supplies typically employ an N+N redundancy design. For example, if N power supply units are sufficient to meet load requirements, high-density multi-node servers are usually configured with 2N power supply units to facilitate failover. Redundant power supply designs typically support 80 Plus Platinum / Titanium certification (a high-end standard in power supply efficiency ratings).

[0037] In related technologies, a backplane is often used as a high-speed communication hub to connect various nodes and expand peripheral modules (such as hard drive backplanes and fan control boards). Furthermore, a high-reliability power board is used to distribute power to redundant power supplies, ensuring stable power supply and efficient current sharing for each subsystem (CPU, storage subsystem, network subsystem).

[0038] For example, in a typical 2U (rack unit) multi-node architecture design, the current transmission path is: power supply unit, motherboard compute node, power board, mid-backplane, fan board and hard drive backplane, fan and hard drive. The connections between the power supply unit and the motherboard compute node, between the motherboard compute node and the power board, and between the power board and the mid-backplane are typically board-to-board or board-to-gold-finger connections to achieve efficient and stable power transmission. The connections between the mid-backplane and the fan board and hard drive backplane are wire-to-board connections for flexible cabling and maintenance.

[0039] In the above power supply architecture, the power distribution within the server is realized with the power board as the core. However, in high-density multi-node servers, this power supply architecture mainly suffers from three major problems: high cost, thermal management failure, and power distribution imbalance.

[0040] This is because power supply boards are manufactured using the Printed Circuit Board Fabrication Process (PCB process). Taking copper as the conductive material of a power supply board as an example, in a conventional PCB process, the safe current carrying capacity per millimeter of trace width on a 1oz copper-thick (approximately 35µm) trace is about 1A. However, in high-density multi-node servers, power supply boards need to carry a total current of up to 300A, which necessitates the use of PCB designs with more than 10 layers, resulting in high hardware costs. The highly integrated power distribution within the server also makes it prone to single-point failures affecting the entire system, requiring the replacement of the entire power supply board in case of failure, thus increasing maintenance costs.

[0041] The power supply board is the core of power distribution and carries a large operating current for a long time. Copper loss leads to a serious temperature rise in the power supply board. The power supply board is large in area and lies between the motherboard computing nodes and the middle backplane, which can also block the heat dissipation airflow in the server, reduce the heat dissipation efficiency of the server's internal heat dissipation system, and increase the heat dissipation burden of the server.

[0042] Furthermore, considering the routing design on the power supply board and the avoidance relationships between traces, it is difficult to ensure that the power supply path length and wire diameter for each node on the power supply board are completely consistent. This results in different impedances for different power supply paths, leading to significant current distribution deviations. Consequently, this causes wasted power margin, reduced energy efficiency, and accelerated local temperature rise, affecting device reliability. If thick copper technology and high-precision impedance control are considered to optimize the power supply path impedance during the power supply board design process, it will further increase the complexity and pressure of PCB manufacturing, as well as increase hardware costs.

[0043] Based on this, embodiments of this application provide a server, such as... Figure 1 As shown, Figure 1 This is a power topology diagram of a server provided in an embodiment of this application.

[0044] like Figure 1 As shown, the server includes multiple power supply units 1, multiple motherboard computing nodes 2, a first bus 31, and a middle backplane 4.

[0045] In this configuration, a power supply unit 1 is mounted on a motherboard computing node 2. The motherboard computing node 2 includes a first power expansion interface 21, and the power supply unit 1 is electrically connected to the first power expansion interface 21. A first bus 31 is electrically connected to the first power expansion interfaces 21 of multiple motherboard computing nodes 2.

[0046] The backplane 4 includes a second power expansion interface 41, and the first bus 31 is also electrically connected to the second power expansion interface 41.

[0047] The first busbar 31 is a conductive metal strip used to concentrate, distribute and transmit electrical energy. It is usually made of copper, aluminum or their alloys and has good conductivity and mechanical strength, such as copper busbar or aluminum busbar.

[0048] In the above embodiments of this application, multiple power supply units 1 are electrically connected to the first busbar 31 through corresponding first power expansion interfaces 21, and the middle backplane 4 is electrically connected to the first busbar 31 through the second power expansion interface 41. That is, the first busbar 31 acts as the power distribution hub of the server, and the power provided by the multiple power supply units 1 is distributed to the corresponding loads on the first busbar 31.

[0049] For example, the first power expansion interface 21 and the first bus 31, and the second power expansion interface 41 and the first bus 31 adopt a hot-swappable design. The corresponding components can be connected or removed without shutting down the system or cutting off the power. This supports online maintenance and quick replacement, which is beneficial for maintaining and expanding system functions without interrupting services and helps reduce the complexity of operation and maintenance.

[0050] In this embodiment, by providing a first bus 31, the power supply board in related technologies can be eliminated. The first bus 31 provides efficient and reliable centralized power supply, meeting the stringent requirements of servers in terms of energy consumption, density, and reliability. Compared to the power supply board in related technologies, the first bus 31 has a simpler structure and lower manufacturing cost, which facilitates standardized production and the availability of components.

[0051] In the event of a failure in the first busbar 31, it can be quickly replaced, preventing a single point of failure from affecting the entire system and helping to reduce maintenance costs.

[0052] Furthermore, by appropriately selecting the specifications of the first bus 31, it can achieve a larger current carrying capacity. In other words, compared to the power supply board, a smaller increase in the size of the first bus 31 can significantly increase its current carrying capacity. The larger the current carrying capacity of the first bus 31, the less heat generation it experiences during operation. Even under high load and heat generation, the first bus 31 is directly exposed to the air, and since the conductive material is not encased in an insulating substrate, it can directly dissipate heat to the air, resulting in high heat dissipation efficiency and minimal temperature rise. Moreover, the first bus 31 has a simple structure and occupies a small area, which, compared to the power supply board, is more conducive to optimizing the airflow layout within the server. It does not cause airflow blockage within the server, further enhancing heat dissipation efficiency, reducing the thermal management pressure on the server, and helping to lower the cost of the cooling system.

[0053] Furthermore, the first bus 31 has a relatively low impedance. The power supply path from the power supply unit 1 to the first bus 31 on each motherboard computing node 2 depends on the design of the motherboard computing node 2. When each motherboard computing node 2 adopts a standardized, modular process, the power supply path from the power supply unit 1 to the first power expansion interface 21 is the same on different motherboard computing nodes 2. That is, this application helps ensure that the power supply path from each power supply unit 1 to the first bus 31 is identical. Combined with the low impedance of the first bus 31, when multiple power supply units 1 are powered in parallel, the identical power supply paths from the power supply unit 1 to the first bus 31 help ensure that current is evenly distributed to each path, resulting in low current distribution deviation and good current sharing effect, which is beneficial for ensuring the safety and stability of the server power supply system. Under sudden high load conditions, such as CPU acceleration, the power supply scheme of this application will not cause response delay or voltage dips, which helps avoid CPU protective frequency throttling.

[0054] For example, in a certain related technology, the power supply board uses a PCB design with more than 10 layers, thick copper processing, and high-precision impedance control. The cost of the motherboard and power supply board-related PCB components alone exceeds 1,000 yuan per unit, accounting for approximately 15% to 20% of the total material cost, far exceeding the 10% upper limit in the ideal solution. Its total power consumption exceeds 6,000W, and the board-level temperature rise caused by copper loss reaches 20℃ to 35℃. The installation of the power supply board also cuts off 30% to 40% of the air intake area, disrupting the airflow path and reducing heat dissipation efficiency by about 25%. The combined effects of the power supply board's own heat generation, blocked airflow, and the high temperatures of adjacent heat-generating components (e.g., the motherboard chipset on compute node 2 (approximately 80°C~95°C), the middle backplane 4 (approximately 60°C~75°C), and the server's internal hard drive backplane (approximately 50°C~65°C) create localized hot spots exceeding 100°C. This accelerates the degradation of electrolytic capacitors and causes the junction temperatures of power devices such as metal-oxide-semiconductor field-effect transistors (MOSFETs) within the server to exceed their limits, leading to frequency reduction or system crashes. Furthermore, the routing of traces on the power supply board requires careful avoidance of each other, making it difficult to precisely ensure that the power supply path from the power supply board to each node is identical. This results in a current distribution deviation of 10%~15% on the power supply board, leading to a response delay greater than 100μs and a voltage sag of approximately 5%~8% under sudden high loads at nodes.

[0055] This application reduces production costs and power loss by about 10% by using the first busbar 31 as the power distribution center, avoids local high-temperature hot spots, reduces heat dissipation system costs by 15%, and reduces the overall system cost by more than 5%. Furthermore, by using copper busbar impedance matching, it can achieve ±2% current balance, avoiding the problem of insufficient dynamic response.

[0056] In some embodiments, the motherboard compute node 2 further includes a third power expansion interface 22, and the power supply unit 1 is also electrically connected to the third power expansion interface 22. The server further includes a second bus 32, which is electrically connected to the third power expansion interfaces 22 of the plurality of motherboard compute nodes 2. The backplane 4 further includes a fourth power expansion interface 42, and the second bus 32 is also electrically connected to the fourth power expansion interface 42.

[0057] For example, the server's internal power supply system uses DC power. One of the first bus 31 and the second bus 32 is used to connect to the positive power in the DC power supply system, and the other is used to connect to the negative power in the DC power supply system. In other words, one is energized and the other is grounded.

[0058] For example, on a motherboard computing node 2, a first power expansion interface 21 is used for power connection, and a third power expansion interface 22 is used for grounding. Both are electrically connected to the positive and negative terminals of the power supply unit 1, respectively, to form a power supply loop. The first power expansion interface 21 is electrically connected to the first bus 31, thereby transmitting power to the first bus 31. Downstream of the power supply from the first bus 31, a second power expansion interface 41 on the backplane 4 is electrically connected to the first bus 31, thereby further transmitting power to the backplane 4. Similarly, a fourth power expansion interface 42, the second bus 32, and the third power expansion interface 22 are electrically connected to form a common grounding loop, thus ensuring the integrity of the power supply loop.

[0059] Similar to the first bus 31, the second bus 32 has a simple structure and low manufacturing cost. Together with the first bus 31, it completely eliminates the power board in related technologies, which helps to solve the problems of "high cost, thermal management failure and unbalanced power distribution" in related technologies, and realizes the optimization of the server power supply system.

[0060] In some embodiments, such as Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of a server provided in an embodiment of this application.

[0061] Multiple motherboard computing nodes 2 are stacked along the first direction Z to form a first stacking structure A1 and a second stacking structure A2. The first stacking structure A1 and the second stacking structure A2 are arranged along the second direction X, and the first direction Z intersects the second direction X.

[0062] Along the third direction Y, the first bus 31 and the second bus 32 are both located on the same side of the multiple motherboard computing nodes 2, and the third direction Y is perpendicular to the first direction Z and the second direction X.

[0063] Understandably, arranging multiple motherboard compute nodes 2 in an array within the server is beneficial for saving space, increasing the number of compute nodes, and improving server computing power. The motherboard compute nodes 2 can be formed using standardized, modular processes. On each motherboard compute node 2, the first power expansion interface 21 and the third power expansion interface 22 are located in the same position. Based on this, the first power expansion interface 21 and the third power expansion interface 22 on the multiple motherboard compute nodes 2 after array arrangement also form an array arrangement. This neat arrangement facilitates electrical connection with the bus.

[0064] like Figure 2 As shown, the first bus 31 includes a first part 311, a second part 312 and a third part 313 connected in sequence. The first part 311 is electrically connected to the motherboard computing node 2 and the middle backplane 4 in the first stacked structure A1, and the third part 313 is electrically connected to the motherboard computing node 2 and the middle backplane 4 in the second stacked structure A2.

[0065] The second bus 32 includes a fourth part 321, a fifth part 322 and a sixth part 323 connected in sequence. The fourth part 321 is electrically connected to the motherboard computing node 2 and the middle backplane 4 in the first stacked structure A1, and the sixth part 323 is electrically connected to the motherboard computing node 2 and the middle backplane 4 in the second stacked structure A2.

[0066] Based on the standardized and modular process of the motherboard computing node 2 and its array arrangement within the server, the first bus 31 connecting the first power expansion interface 21 and the second bus 32 connecting the third power expansion interface 22 can adopt similar design and process requirements. For example, in a certain embodiment, the performance parameters of the first bus 31 and the second bus 32 are: continuous operating current ≥272A, copper thickness of 3mm, and height along the first direction Z of 73.8mm. During the connection process, care should be taken to avoid short circuits between the first bus 31 and the second bus 32. In actual production and assembly, to facilitate differentiation, the first bus 31 and the second bus 32 can be marked with different colors, for example, the first bus 31 can be gray to indicate being charged, and the second bus 32 can be black to indicate being grounded.

[0067] For example, such as Figure 2 As shown, both the first bus 31 and the second bus 32 are U-shaped. Alternatively, depending on the viewing angle, the orthographic projection of the first bus 31 and the second bus 32 on the stacking plane (XZ plane) of the motherboard computing node 2 is a gate shape.

[0068] The orthographic projection of the first bus 31 and the second bus 32 on the XZ plane is gate-shaped. For the airflow along the third direction Y, the shading area is negligible, preventing airflow blockage. Compared to related technologies, this is more beneficial for the heat dissipation of the motherboard compute node 2, preventing severe localized temperature rise in the motherboard compute node 2 due to obstruction of its heat dissipation airflow, which could lead to system throttling or crashes. Based on the U-shaped structure of the first bus 31 and the second bus 32, the server's thermal management is easier, facilitating further optimization of the server's internal cooling system and reducing cooling costs.

[0069] In some embodiments, such as Figure 2 As shown, along the second direction X, the first part 311 is located between the fourth part 321 and the sixth part 323, and the sixth part 323 is located between the first part 311 and the third part 313. Along the first direction Z, the second part 312 is located on the side of the fifth part 322 near the middle back plate 4.

[0070] As mentioned above, the motherboard computing node 2 is formed using a standardized and modular process, which helps to reduce production costs and complexity, and improves the adaptability of the motherboard computing node 2. Based on this, along the second direction X, the first power expansion interface 21 and the third power expansion interface 22 are arranged alternately. In order to realize the electrical connection of multiple first power expansion interfaces 21 to the same first bus 31, and the electrical connection of multiple third power expansion interfaces 22 to the same second bus 32, the orthographic projections of the first bus 31 and the second bus 32 on the XZ plane will intersect.

[0071] To ensure electrical isolation between the first busbar 31 and the second busbar 32 and to prevent air back breakdown, a certain distance needs to be maintained between the first busbar 31 and the second busbar 32.

[0072] Taking the first busbar 31 as an example, the width of the first busbar 31 is its dimension along the third direction Y, and the thickness of the first busbar 31 is the dimension of the first part 311 along the second direction X (or the dimension of the second part 312 along the first direction Z, or the dimension of the third part 313 along the second direction X). To reduce production costs and improve adaptability, the second busbar 32 can adopt the same size design as the first busbar 31.

[0073] Since the thickness of the bus is less than its width, under the scheme of staggered arrangement of the second part 312 and the fifth part 322 along the first direction Z, the total space occupied by the first bus 31 and the second bus 32 in the server is less than the total space occupied by the staggered arrangement of the two along the third direction Y.

[0074] In other words, the above-mentioned configuration scheme of this application can make full use of the space occupied by the motherboard computing nodes 2 stacked along the first direction Z, and avoid increasing the size of the server along the third direction Y in order to achieve electrical isolation between the first bus 31 and the second bus 32.

[0075] For example, the second bus 32 is used for grounding. The second bus 32 is located outside most of the first bus 31 and can also serve as a shield.

[0076] In some embodiments, combined with Figures 2-3 As shown, Figure 3 This is a partially enlarged view of a bus provided in an embodiment of this application. Figure 3 The perspective corresponds to the perspective of "pointing from the motherboard compute node to the bus".

[0077] The second part 312 includes a first segment 3121 and a second segment 3122 connected together. The second segment 3122 is closer to the third part 313 than the first segment 3121. The first segment 3121 extends along the second direction X, and the second segment 3122 bends toward the direction of the multiple motherboard computing nodes 2.

[0078] The fifth part 322 includes the connected third segment 3221 and the fourth segment 3222. The fourth segment 3222 is closer to the sixth part 323 than the third segment 3221. The third segment 3221 extends along the second direction X, and the fourth segment 3222 bends away from the multiple motherboard computing nodes 2.

[0079] Combination Figure 3 Understandably, since the second segment 3122 is connected to the third part 313, if the second segment 3122 is not bent, its path connecting to the third part 313 will conflict with the second busbar 32. Alternatively, the fourth segment 3222 needs to be bent at a large angle to avoid this conflict. However, an excessively large bending angle would affect the mechanical reliability of the busbar and could also cause localized overheating due to uneven impedance.

[0080] In this application, the second segment 3122 and the fourth segment 3222 are bent in different directions, with both having small bending angles, to achieve avoidance between the first busbar 31 and the second busbar 32, thereby ensuring the corresponding functions of the busbars. The small bending angles of both the second segment 3122 and the fourth segment 3222 not only facilitate process implementation but also help avoid the problem of uneven impedance at the bend caused by excessively large bending angles, which could lead to localized heating. This helps ensure the impedance uniformity of the busbars themselves and improves their current sharing capability.

[0081] In some embodiments, such as Figure 3As shown, the backplane 4 also includes a support member 5, which is located on the same side as the second power expansion interface 41. Along the third direction Y, the support member 5 is located on one side of multiple motherboard computing nodes 2.

[0082] like Figure 4 As shown, Figure 4 This is a partially enlarged view of a power expansion interface provided in an embodiment of this application.

[0083] In some embodiments, the first power expansion interface 21 and the second power expansion interface 41 are both bus clamps, which support full-load hot-swapping and facilitate the implementation of hot-swapping operations. In the event of a failure in the motherboard computing node 2, the first bus 31, or the middle backplane 4, it is beneficial to achieve rapid replacement.

[0084] For example, the bus clamp is provided with pins for electrical connection to the busbar. The spacing between two adjacent pins in the bus clamp is 5.8mm, and the bus clamp is configured to have a current carrying capacity of ≥200A and a total transmission power of ≥2400W.

[0085] The first power expansion interface 21, the third power expansion interface 22, the second power expansion interface 41, and the fourth power expansion interface 42 can use the same bus clamp configuration, which is beneficial for achieving high reusability and simplified design.

[0086] Combination Figure 3 As can be understood, the support component 5 can be a precision die-cast part for Computer Numerical Control (CNC), balancing cost, efficiency, and precision requirements. Both the first bus 31 and the second bus 32 can be fixed to the support component 5 with screws, and both are located on the side of the support component 5 closest to the multiple motherboard computing nodes 2.

[0087] Combination Figure 2 and Figure 3 As shown, in the first stacked structure A1, the motherboard computing nodes 2 clamp the third part 313 of the first bus 31 through their respective first power expansion interfaces 21; in the second stacked structure A2, the motherboard computing nodes 2 clamp the first part 311 of the first bus 31 through their respective first power expansion interfaces 21, thereby achieving electrical connection between multiple motherboard computing nodes 2 and the first bus 31. Similarly, in the first stacked structure A1, the motherboard computing nodes 2 clamp the sixth part 323 of the second bus 32 through their respective third power expansion interfaces 22; in the second stacked structure A2, the motherboard computing nodes 2 clamp the fourth part 321 of the second bus 32 through their respective third power expansion interfaces 22, thereby achieving electrical connection between multiple motherboard computing nodes 2 and the second bus 32.

[0088] Similarly, the middle backplate 4 is provided with two second power expansion interfaces 41, one of which clamps the third part 313 and the other clamps the first part 311. The middle backplate 4 is also provided with two fourth power expansion interfaces 42, one of which clamps the sixth part 323 and the other clamps the fourth part 321.

[0089] In some embodiments, such as Figure 3 As shown, the support member 5 has a certain thickness along the third direction Y to provide fixed support for the busbar. In order to achieve the avoidance of the first busbar 31 and the second busbar 32 without increasing the size of the server, the support member 5 includes an inner groove 51, and at least part of the connection between the fifth part 322 and the sixth part 323 is located in the inner groove 51.

[0090] That is, by partially reducing the thickness of the support member 5 to allow for the installation of the second busbar 32, the connection between the second part 312 and the third part 313 on the first busbar 31 is bent at a small angle away from the support member 5. This achieves the avoidance of the first busbar 31 and the second busbar 32, ensuring the support member 5 can effectively support and fix the first busbar 31 and the second busbar 32, avoiding the increase in server size, and ensuring electrical isolation between the first busbar 31 and the second busbar 32.

[0091] In some embodiments, such as Figure 3 As shown, the support member 5 also includes multiple hollowed-out portions 52, which penetrate the support member 5 along the third direction Y.

[0092] The perforated section 52 ensures unobstructed airflow within the server, preventing blockage by the support component 5. When airflow passes through the perforated section 52, the flow velocity increases due to the smaller cross-sectional area of ​​the airflow, creating a localized low-pressure zone. This draws in surrounding air, enhancing airflow and acting as an air guide, accelerating airflow along the Y-axis. This improves the cooling efficiency of the heat dissipation system, making the server easier to manage in terms of heat dissipation, preventing severe localized temperature rises, avoiding system throttling or crashes, and ultimately reducing the cost of the cooling system.

[0093] In some embodiments, such as Figures 5-6 As shown, Figure 5 This is a schematic diagram of a motherboard computing node provided in an embodiment of this application. Figure 6 This is a partial enlarged view of a power connector provided in an embodiment of this application.

[0094] The motherboard computing node 2 also includes a circuit board 20 and a power connector 23, a first terminal 24, a second terminal 25, and a first power expansion interface 21 disposed on the circuit board 20. Within the circuit board 20, the power connector 23 is electrically connected to the first terminal 24, and the second terminal 25 is electrically connected to the first power expansion interface 21. The power supply unit 1 is electrically connected to the power connector 23. The motherboard computing node 2 also includes a motherboard bus 26, one end of which is electrically connected to the first terminal 24, and the other end of which is electrically connected to the second terminal 25.

[0095] That is, on the motherboard computing node 2, the power supply path is as follows: power supply unit 1, power connector 23, first terminal 24, motherboard bus 26, second terminal 25, and first power expansion interface 21. This power supply path is relatively long (especially between the first terminal 24 and the second terminal 25) and the power supply current is relatively large. If all power supply paths are set in the circuit board 20, the process pressure and cost pressure of the circuit board 20 will be relatively large, and it will be accompanied by a relatively serious heat generation problem.

[0096] In this application, by setting a main board bus 26 outside the circuit board 20, 97% of the power load can be transferred to the main board bus 26, and the number of conductive layers inside the circuit board 20 can be optimized and reduced by at least 2 layers, which can reduce the manufacturing cost by about 25%.

[0097] Furthermore, this application only separates a portion of the power transmission path to the outside of the circuit board 20, while the signal transmission path remains inside the circuit board 20. Based on this, the transmission paths of signals and power are separated, which helps to reduce interference and avoid temperature rise. For example, in a certain product, the temperature rise can be reduced by 10 degrees Celsius compared to the traditional design.

[0098] There are clearly visible nodes on the power supply path described above in this application. In the event of a fault in the power supply path, it is possible to inspect the fault in sections to quickly determine the location of the fault point and then carry out corresponding repairs and maintenance. This helps to reduce repair time and improve maintainability. For example, in a certain product, the mean time to repair (MTTR) is reduced by 30%.

[0099] Power Supply Unit 1 (PSU) is used to convert external AC power into the DC power required by the server. The redundant power supply design can be understood as the server supporting two hot-swappable PSUs; if one fails, the other seamlessly takes over, ensuring uninterrupted server operation. For example, PSU 1 can also be integrated with a Baseboard Management Controller (BMC) to enable remote monitoring, power control, and energy-saving strategies.

[0100] In some embodiments, the power supply unit 1 outputs 12V, 270A electrical energy to the power connector 23, such as Figure 6 As shown, the power connector 23 includes power terminals and signal terminals arranged in an array. For example, the power terminals include two rows of 18 pins, with a spacing of 2.54 mm between adjacent pins, and each pin has a current carrying capacity of ≥33A. The signal terminals include two rows of 10 pins, with a spacing of 1.27 mm between adjacent pins, and each pin has a current carrying capacity of ≥1A. The total power carrying capacity of this power connector 23 is approximately 3564W.

[0101] Combination Figure 5 It is understandable that the power supply unit 1 is located at the rear end of the circuit board 20, while the first power expansion interface 21 and the third power expansion interface 22 are located at the front end of the circuit board 20, resulting in a longer transmission path.

[0102] In this application, the power connector 23 transmits system power to the first terminal 24 via a short distance on the circuit board 20, and then efficiently transmits it to the second terminal 25 via the main board bus 26. Finally, it transmits system power to the first power expansion interface 21 and the third power expansion interface 22 via another short distance on the circuit board 20, thereby realizing the transmission of power from the rear end to the front end of the circuit board 20.

[0103] Combination Figure 5 The circuit board 20 also has two CPU chips 27. The transmission path from the back end to the front end of the circuit board 20 must cross these two CPU chips 27. If the transmission path still relies on the internal wiring of the circuit board 20, it will not only put great pressure on the process and cost of the circuit board 20, but its serious heat generation problem will also affect the CPU chips 27, and the serious local temperature rise will cause the system to crash.

[0104] This application utilizes a motherboard bus 26, spanning the area corresponding to two CPU chips 27 above the circuit board 20. For example, the motherboard bus 26 can adopt an L-shaped copper bus structure, with both ends soldered to the terminals on the circuit board 20 furthest from the CPU chips 27. The motherboard bus 26 stands on the circuit board 20, providing a large spatial isolation between the power supply path and the CPU chips 27. The excellent heat dissipation capability of the motherboard bus 26 allows even minor heat generation to be dissipated into the air in a timely manner, preventing heat accumulation within the circuit board 20 and thus avoiding overall temperature rise and impact on the CPU chips 27. For example, the motherboard bus 26 is configured with an operating current ≥227A, and can be designed with structural parameters of 17.5mm thickness and 17.5mm board height.

[0105] In some embodiments, combined with Figure 1 and Figure 2 As shown, the server includes four power supply units 1, four motherboard compute nodes 2, and a mid-backplane 4.

[0106] For example, such as Figure 7 As shown, Figure 7 This is a rear view of a server provided in an embodiment of this application. Figure 7 This corresponds to a 2U4N multi-node server, which integrates four independent motherboard compute nodes 2 within a standard 2U chassis height. These four independent motherboard compute nodes 2 can be designated Node0, Node1, Node2, and Node3, respectively. The corresponding four power supply units 1 are designated PSU0, PSU1, PSU2, and PSU3, respectively. The four power supply units 1 employ a 2+2 redundancy configuration to ensure power supply reliability. Each power supply unit 1 can independently support one motherboard compute node 2. A single power supply unit 1 has a width of 68mm, a rated power of 3200W, and a conversion efficiency of ≥96% under typical operating conditions, achieving high efficiency and energy saving. Alternatively, it can be flexibly configured for on-demand power supply, with current sharing switching between nodes to optimize energy usage efficiency.

[0107] Each independent motherboard compute node 2 may also include two high-speed serial computer expansion bus standard interfaces (Peripheral Component Interconnect Express, PCIe), designated PCIe1 and PCIe2, which can be used to connect expansion devices such as graphics cards, network cards, solid-state drives (SSDs), and sound cards. Each motherboard compute node 2 also includes a network card expansion interface OCP0 compliant with the OCP3.0 specification (Open Compute Project 3.0), which can be used to install dedicated high-speed Ethernet cards or fiber optic network cards.

[0108] Each motherboard compute node 2 typically also includes a Unit Identification Button (UID button) 61, a micro USB port 62, a reset button 63, and a standardized network interface 64. The UID button 61 is used for quickly locating and managing the server's hardware functions. The micro USB port 62 is used to connect the server's debug signals externally to view logs or execute debug commands. The reset button 63 is used to restart the server. The standardized network interface 64 can be an RJ45 network interface for Ethernet network connectivity.

[0109] The server provided in this application has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A server, characterized in that, include: Multiple power supply units and multiple motherboard computing nodes are provided, with one power supply unit disposed on one motherboard computing node; the motherboard computing node includes a first power expansion interface and a third power expansion interface, the power supply unit is electrically connected to the first power expansion interface, and the power supply unit is also electrically connected to the third power expansion interface; The first bus is electrically connected to the first power expansion interface of the plurality of motherboard computing nodes; The second bus is electrically connected to the third power expansion interface of the plurality of motherboard computing nodes; The backplane includes a second power expansion interface and a fourth power expansion interface, and the first bus is also electrically connected to the second power expansion interface; the second bus is also electrically connected to the fourth power expansion interface; The plurality of motherboard computing nodes are stacked along a first direction to form a first stacking structure and a second stacking structure, the first stacking structure and the second stacking structure are arranged along a second direction, and the first direction and the second direction intersect. Along the first direction, the first power expansion interface and the third power expansion interface are alternately arranged; Along a third direction, the first bus and the second bus are both located on the same side of the plurality of motherboard computing nodes, and the third direction is perpendicular to the first direction and the second direction; The first bus includes a first part, a second part, and a third part connected in sequence. The first part is electrically connected to the motherboard computing node and the middle backplane in the first stacked structure, and the third part is electrically connected to the motherboard computing node and the middle backplane in the second stacked structure. The second bus includes a fourth part, a fifth part, and a sixth part connected in sequence. The fourth part is electrically connected to the motherboard computing node in the first stacked structure and the middle backplane. The sixth part is electrically connected to the motherboard computing node in the second stacked structure and the middle backplane.

2. The server according to claim 1, characterized in that, Both the first bus and the second bus are U-shaped.

3. The server according to claim 1, characterized in that, Along the second direction, the first portion is located between the fourth portion and the sixth portion, and the sixth portion is located between the first portion and the third portion; Along the first direction, the second portion is located on the side of the fifth portion closer to the middle back plate.

4. The server according to claim 1, characterized in that, The second part includes a first segment and a second segment connected together, the second segment being closer to the third part than the first segment; the first segment extends along the second direction, and the second segment bends toward the plurality of motherboard computing nodes; The fifth part includes a third segment and a fourth segment connected together, the fourth segment being closer to the sixth part than the third segment; the third segment extends along the second direction, and the fourth segment bends away from the plurality of motherboard computing nodes.

5. The server according to claim 1, characterized in that, The back panel also includes a support member, and the support member and the second power expansion interface are located on the same side of the back panel. Along the third direction, the support member is disposed on one side of the plurality of motherboard computing nodes; Both the first bus and the second bus are fixed to the support member and are located on the side of the support member closer to the plurality of motherboard computing nodes.

6. The server according to claim 5, characterized in that, The support member includes an inner groove; At least a portion of the connection between the fifth part and the sixth part is located within the inner groove.

7. The server according to claim 5, characterized in that, The support member also includes multiple hollowed-out portions, which penetrate the support member along the third direction.

8. The server according to claim 1, characterized in that, The first power expansion interface is a busbar clamp; the second power expansion interface is a busbar clamp.

9. The server according to claim 1, characterized in that, The motherboard computing node also includes a circuit board and a power connector, a first terminal, a second terminal, and a first power expansion interface disposed on the circuit board; within the circuit board, the power connector is electrically connected to the first terminal, and the second terminal is electrically connected to the first power expansion interface; The power supply unit is electrically connected to the power connector; The motherboard computing node also includes a motherboard bus, one end of which is electrically connected to the first terminal and the other end of which is electrically connected to the second terminal.

10. The server according to claim 1, characterized in that, The server includes four power supply units, four motherboard compute nodes, and a mid-backplane.

Images