# US 2026/0281060 A1 · Heterogeneous server with multiple accelerator cards and resource link reconfiguration method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0281060 A1
- Document ID: US20260281060A1
- Status: Published application
- Kind: A1
- Application number: 19/166,810
- Filed: 2024-05-27
- Published: 2026-09-17
- Earliest claimed: 2024-01-31
- Assignees: Suzhou Metabrain Intelligent Technology Co., Ltd.
- Applicants: Suzhou Metabrain Intelligent Technology Co., Ltd.
- Inventors: Xianyang GAO; Tanlong CI; An WU; Tong XU
- Web page: https://patentgrep.com/patent/US20260281060A1

## CPC

- G06N 20/00: Machine learning
- H04L 49/109: Integrated on microchip, e.g. switch-on-chip
- G06F 15/177: Initialisation or configuration control {(processor initialisation G06F9/4405)}
- Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

The present disclosure discloses a heterogeneous server with multiple accelerator cards. The heterogeneous server includes a switching unit, a heterogeneous accelerated computing pooling unit and a general computing resource pooling unit, where the general computing resource pooling unit includes at least one processor, the heterogeneous accelerated computing pooling unit includes at least two accelerator cards, and the processor and the accelerator cards are heterogeneous to each other; the switching unit includes a management controller and at least two switching chips, any switching chip is connected to at least one other switching chip through an internal port, the processor and one or more accelerator cards are connected to the switching chip; and the management controller is connected to each switching chip, and is configured to reconstruct the resource link according to a preset task resource requirement and chip information of each current switching chip.

## Claims

### Claim 1

A heterogeneous server with multiple accelerator cards, comprising: a switching unit, a heterogeneous accelerated computing pooling unit and a general computing resource pooling unit, wherein
the general computing resource pooling unit comprises at least one processor, and the heterogeneous accelerated computing pooling unit comprises at least two accelerator cards, wherein the at least one processor and the at least two accelerator cards are heterogeneous to each other;
the switching unit comprises a management controller and at least two switching chips, wherein any switching chip in the at least two switching chips is connected to at least one other switching chip in the switching unit through an internal port, the at least one processor is connected to an uplink port of the switching chip, and one or more of the at least two accelerator cards are connected to a downlink port of the switching chip; and
the management controller is connected to each of the at least two switching chips, and is configured to determine a resource link reconfiguration instruction according to a preset task resource requirement and chip information of each current switching chip, and send the resource link reconfiguration instruction to each target switching chip to form a resource link for completing a computing task corresponding to the task resource requirement,
wherein the switching unit further comprises a first substrate controller and a first complex programmable logic chip, wherein the first complex programmable logic chip is respectively connected to the switching chip, the first substrate controller and the management controller, and the management controller is connected to the first substrate controller;
the heterogeneous accelerated computing pooling unit further comprises a second substrate controller and a second complex programmable logic chip, wherein the second complex programmable logic chip is respectively connected to the accelerator card and the second substrate controller;
the general computing resource pooling unit further comprises a third substrate controller and a third complex programmable logic chip, wherein the third complex programmable logic chip is respectively connected to the processor and the third substrate controller;
the third complex programmable logic chip is connected to the first complex programmable logic chip, and the first complex programmable logic chip is connected to the second complex programmable logic chip; and
the second complex programmable logic chip controls the accelerator card and the second substrate controller to be powered on, the first complex programmable logic chip controls the switching chip, the management controller and the first substrate controller to be powered on, and the third complex programmable logic chip controls the processor and the third substrate controller to be powered on, wherein a power off sequence is opposite to a power on sequence.

### Claim 2 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein each of the at least two switching chips comprises an embedded controller, wherein the embedded controller is configured to enumerate chip information of the switching chip where the embedded controller is located, and send the chip information to the management controller.

### Claim 3 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the management controller is configured to search for a target switching chip identification (ID) and a target port ID on the target switching chip based on the preset task resource requirement, a switching chip ID, a port ID of the switching chip, port bandwidth, a port rate, port connection information and port usage information for minimizing link communication overhead, and generate the resource link reconfiguration instruction comprising the target switching chip ID and the target port ID on the target switching chip.

### Claim 4 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the management controller is further configured to display the preset task resource requirement, a switching chip identification (ID), a port ID of the switching chip, port bandwidth, a port rate, port connection information and port usage information on a display device through a graphical interface, and receive the resource link reconfiguration instruction input through the graphical interface.

### Claim 5 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the switching unit comprises two switching chipsets, each of the two switching chipsets comprises at least one switching chip, and a number of switching chips in the switching chipset is less than a total number of ports of the switching chips, any switching chip in a first switching chipset is connected to each switching chip in a second switching chipset through an internal port, and any switching chip in the second switching chipset is connected to each switching chip in the first switching chipset through the internal port, at least one of remaining ports of all switching chips is used as the uplink port to connect the processor, and at least two of remaining ports of all switching chips are used as the downlink ports to connect the accelerator cards.

### Claim 6 (depends on 5)

The heterogeneous server with multiple accelerator cards according to claim 5, wherein the switching chipset comprises four switching chips, each of the four switching chips has nine ports, any switching chip in the first switching chipset is respectively connected to the four switching chips in the second switching chipset through four internal ports, and any switching chip in the second switching chipset is respectively connected to the four switching chips in the first switching chipset through the four internal ports, eight of the remaining ports of all switching chips are used as the uplink ports to connect eight processors, and the remaining thirty-two ports are used as the downlink ports to connect thirty-two accelerator cards.

### Claim 7 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the switching unit comprises a plurality of switching chips, a number of switching chips is less than a number of ports of the switching chips, any switching chip is connected to remaining switching chips through the internal ports, at least one of remaining ports of all switching chips is used as the uplink port to connect the processor, and at least two of remaining ports of all switching chips are used as the downlink ports to connect the accelerator cards.

### Claim 8 (depends on 7)

The heterogeneous server with multiple accelerator cards according to claim 7, wherein the switching unit comprises six switching chips, each of the six switching chips has nine ports, any switching chip is connected to the remaining switching chips through five internal ports, sixteen of the remaining ports of all switching chips are used as the downlink ports to connect the accelerator cards, and the remaining at least four ports are used as the uplink ports to connect four processors.

### Claim 9 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein at least one processor is connected to the uplink ports of different switching chips.

### Claim 10 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein
the first substrate controller is configured to write the preset task resource requirement into the management controller, and receive link information of the resource link fed back by the management controller, wherein the link information comprises a processor identification (ID) occupied by the resource link, an accelerator card ID, a switching chip ID and a port ID on the switching chip.

### Claim 11 (depends on 10)

The heterogeneous server with multiple accelerator cards according to claim 10, wherein the management controller comprises a first management controller and a second management controller which are mutually redundant, wherein the first management controller is connected to the second management controller through a first switching chip, the first switching chip is connected to the first substrate controller, the first substrate controller is further configured to receive fault information sent by the first management controller or the second management controller, after receiving the fault information from the first management controller, control the first switching chip to disconnect the first management controller from the switching chips, and switch the second management controller to be connected to the switching chips; or after receiving the fault information from the second management controller, control the first switching chip to disconnect the second management controller from the switching chips, and switch the first management controller to be connected to the switching chips.

### Claim 12 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the second substrate controller is configured to obtain status information of the accelerator cards.

### Claim 13 (depends on 12)

The heterogeneous server with multiple accelerator cards according to claim 12, wherein the heterogeneous accelerated computing pooling unit further comprises second switching chips, and the accelerator cards are connected to the second substrate controller, the second complex programmable logic chip and the switching chips through the second switching chips.

### Claim 14 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the third substrate controller is configured to obtain status information of the processor.

### Claim 15 (depends on 1)

The heterogeneous server with multiple accelerator cards according to claim 1, wherein the heterogeneous accelerated computing pooling unit further comprises an interface board and an adapter substrate, wherein the adapter substrate is provided with adapter interfaces and a slot,
the accelerator card is connected to the interface board through an adapter card, and the interface board is connected to the downlink port of the switching chip; or
the accelerator card is connected to the adapter interface through the adapter card, the adapter interface is connected to the interface board, and the interface board is connected to the downlink port of the switching chip; or
the accelerator card is connected to the slot, the slot is connected to the adapter interface, the adapter interface is connected to the interface board, and the interface board is connected to the downlink port of the switching chip.

### Claim 16

(canceled)

### Claim 17 (depends on 1)

A resource link reconfiguration method, implemented based on the heterogeneous server with multiple accelerator cards according to claim 1, and applied to a management controller, comprising:
obtaining a task resource requirement and chip information of each current switching chip;
determining a resource link reconfiguration instruction according to the task resource requirement and the chip information of each current switching chip; and
sending the resource link reconfiguration instruction to each target switching chip to form a resource link for completing a computing task corresponding to the task resource requirement.

### Claim 18 (depends on 17)

The resource link reconfiguration method according to claim 17, wherein determining the resource link reconfiguration instruction according to the task resource requirement and the chip information of each current switching chip comprises:
searching for a target switching chip identification (ID) and a target port ID on the target switching chip based on the preset task resource requirement, a switching chip ID, a port ID of the switching chip, port bandwidth, a port rate, port connection information and port usage information for minimizing link communication overhead, and generating the resource link reconfiguration instruction comprising the target switching chip ID and the target port ID on the target switching chip.

### Claim 19 (depends on 17)

The resource link reconfiguration method according to claim 17, wherein determining the resource link reconfiguration instruction according to the task resource requirement and the chip information of each current switching chip comprises:
displaying the preset task resource requirement, a switching chip identification (ID), a port ID of the switching chip, port bandwidth, a port rate, port connection information and port usage information on a display device through a graphical interface; and
receiving the resource link reconfiguration instruction input through the graphical interface.

### Claim 20

(canceled)

### Claim 21 (depends on 17)

An electronic device, comprising: a memory, a processor and computer readable instructions stored in the memory and executable on the processor, wherein the processor implements the resource link reconfiguration method according to claim 17 when executing the program.

### Claim 22 (depends on 17)

A non-transitory computer readable storage medium having computer readable instructions stored thereon, wherein the computer readable instructions, when executed by one or more processors, cause the processors to perform the resource link reconfiguration method according to claim 17.

### Claim 23

(canceled)
