Common rack server

Common rack server

AI work force

AI work force

The XWR7260 X5 server supports up to two cpus for up to a 52% improvement in processor floating-point performance and a 42% increase in the number of cores, combined with 8-channel 3200MT/s DDR4 memory technology to provide users with up to a 60% increase in bandwidth. Flexible scalability with support for 17 PCIe slots and up to 45 hard disk slots. 96% power efficiency and a standard operating temperature design of 5 ° C to 45 ° C provide users with higher energy efficiency returns.

The new generation XWR7260 X6 server supports up to two cpus and provides up to 50% bandwidth improvement with 8-channel 5600MT/sDDR5 memory technology. Flexible scalability with support for 22 PCIe slots and up to 49 hard disk slots. 96% power efficiency and a standard operating temperature design of 5 ° C to 45 ° C provide users with higher energy efficiency returns.

The G7466 X5 server, which supports up to two cpus, can achieve up to 52% improvement in processor floating-point performance and 42% increase in the number of cores, combined with 8-channel 3200MT/s DDR4 memory technology, providing users with up to 60% bandwidth improvement. Flexible scalability with support for 16 PCIe4.0 slots and up to 50 hard disk slots. 96% power efficiency, as well as a standard operating temperature design of 5 ° C to 40 ° C, provides users with higher energy efficiency returns.

The new generation G7466 X6 server supports up to two cpus and provides up to 50% bandwidth improvement with 8-channel 5600MT/s DDR5 memory technology. Flexible scalability with support for 20 PCIe5.0 slots and up to 50 hard disk slots. 96% power efficiency, as well as a standard operating temperature design of 5 ° C to 40 ° C, provides users with higher energy efficiency returns.

AI work force

Parts and boards

Parts and boards

Parts and boards

The G7868 X6 server is compatible with the new HGX H800 8-GPU module or the standard OAM2.0 module, and the modular design fully improves the heat dissipation and maintainability of the whole machine, making the powerful performance of the GPU fully play. The main control unit supports a maximum of two cpus and integrates the 8-channel 5600MT/s DDR5 memory technology to provide users with up to 50% bandwidth improvement. With the support of 18 PCIe5.0 slots and up to 25 hard disk slots, it can meet the requirements of various computing architecture solutions and further improve the system computing performance. It can be used in large-scale training, high performance computing (HPC), large data centers and other large computing scenarios.

P6600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution. Supports RAID0, 1, 5, and 10 RAID levels.

P6600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution. Supports RAID0, 1, 5, and 10 RAID levels.

P6600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution. Supports RAID0, 1, 5, and 10 RAID levels.

Parts and boards

Parts and boards

Parts and boards

Parts and boards

P6600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution. Supports RAID0, 1, 5, and 10 RAID levels.

P6600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution. Supports RAID0, 1, 5, and 10 RAID levels.

P6600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution. Supports RAID0, 1, 5, and 10 RAID levels.

P7600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution.

Parts and boards

Parts and boards

Parts and boards

Parts and boards

P7600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution.

P7600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution.

P7600 series storage control card based on the latest chip, stable function, powerful performance, PCIe4.0 interface support, relative to the previous generation of product bandwidth doubled, equipped with dynamic channel multiplexing (DCM) technology, compared to the previous generation of Expander link efficiency increased by nearly 100%, to achieve unparalleled throughput; Internal PCIe Switch integration enables the industry's low-latency and high-bandwidth NVMe solution.

Parts and boards

Parts and boards

Parts and boards

Parts and boards

Parts and boards

Parts and boards

PCIE

PCIE

The R100 is equipped with the XPU-R architecture developed by the GPU customer, inheriting the mature software stack of the customer's second-generation chip, which greatly reduces the adaptation and use cost of users. As a low-power, small-size acceleration card, the R100 can be widely used in various edge reasoning scenarios such as smart retail, intelligent transportation, smart parks and intelligent manufacturing to achieve more efficient AI reasoning calculations.

The GPU customer 2 AI chip is suitable for high-performance inference scenarios in data centers, and fully supports various artificial intelligence tasks such as natural language processing, computer vision, voice, and traditional machine learning.

PCIE

PCIE

PCIE

PCIE

The GPU customer 2 AI chip is suitable for high-performance inference scenarios in data centers, and fully supports various artificial intelligence tasks such as natural language processing, computer vision, voice, and traditional machine learning.

The GPU customer 2 AI chip is suitable for high-performance inference scenarios in data centers, and fully supports various artificial intelligence tasks such as natural language processing, computer vision, voice, and traditional machine learning.

The R100 is equipped with the XPU-R architecture developed by the GPU customer, inheriting the mature software stack of the customer's second-generation chip, which greatly reduces the adaptation and use cost of users. As a low-power, small-size acceleration card, the R100 can be widely used in various edge reasoning scenarios such as smart retail, intelligent transportation, smart parks and intelligent manufacturing to achieve more efficient AI reasoning calculations.

Based on GPU-customer-developed XPU-R architecture, RG800 is an AI acceleration card for data center application scenarios, which can be used for routine model training, but also for multi-service concurrent high-performance AI reasoning applications, helping various industries to achieve cost reduction and efficiency, and promoting industrial intelligent upgrading.

PCIE

PCIE

OAM

OAM

PT200-X1 is a GPU customer can launch a 2-generation chip-based AI accelerator group, which can support PCIE5.0; AI acceleration cards located in data center application scenarios can be used for training conventional models.

PT200-M1 is a 2-generation chip-based AI accelerator group launched by a customer, which uses a common substrate, equipped with 8-core AI acceleration module, PT200-M1 can realize inter-chip interconnection, and can support PCIE 5.0.

R300 is an AI accelerator group based on the second-generation chip launched by a customer, which uses a common substrate R480, equipped with 8 AI acceleration modules, R300 can realize the interconnection between chips.

The 106B is a PCIe board with a peak power consumption of 300W. It provides flexible deployment of powerful universal computing power for PCIe GPU servers widely used in data centers.

OAM

UBB

UBB

UBB

106M product form OAM module, with strong power supply and heat dissipation capabilities, can fully liberate the surging computing power, driving the rapid development of general computing including artificial intelligence deep learning.

UBB

DPU

DPU

DPU

106M product form OAM module, with strong power supply and heat dissipation capabilities, can fully liberate the surging computing power, driving the rapid development of general computing including artificial intelligence deep learning.

Taihang DPU2.0 is a DPU product using intelIcelake-D+AgilexFM8 FPGA architecture, which provides 2 100GQSFP28/56 service interfaces, 2 gigabit management interfaces, USB interfaces, and adopts AST2600 BMC for out-of-band management. The CPU board supports two DDR4 SODIMMs and 10 DDR4 chips onboard the FPGA. Goldfinger provides PCIe4.0x16 or PCIe4.0x8*2 ports.

Silver Fir 2.0 is a DPU using intel Eddylake-D CPU+ Agilex FM41 FPGA architecture. It provides two 200G QSFP56 service ports and USB ports, and adopts AST2600 BMC for out-of-band management. The CPU board supports two DDR4 SODIMMs, the FPGA board supports 13 DDR4 particles, and the Goldfinger provides PCIe5.0x16 or PCIe5.0x8*2 ports.

The JM800 is a DPU product developed based on a customer's self-developed ASIC chip. It provides six 56G SFP56 service ports, two gigabit management ports, and USB ports. It adopts AST2600 BMC for out-of-band management and 20pcs on-board DDR5 chips. Goldfinger provides PCIe5.0x16 or PCIe5.0x8*2 ports.

DPU

DPU

DPU

The hybrid DPU card is a DPU based on the domestic Haiguang 3 CPU+ Agilex FM8 FPGA architecture. It provides two 100G QSFP112/56 and two SFP+ service interfaces and USB interfaces. The CPU board supports two DDR4 SODIMMs. FPGA board onboard 14pcs DDR4 particles, Goldfinger provides PCIe4.0x16 or PCIe4.0x8*2 interface.

Wugongshan PCIe card DPU is a DPU product using intel Icelake-D+ Agilex FM8 FPGA architecture. It provides 2 100G DSFP service interfaces, 2 gigabit management interfaces, USB interfaces, etc. The AST2600 BMC is used for out-of-band management, the CPU board supports two DDR4 SODIMMs, the FPGA board supports 13 DDR4 particles, and the Goldfinger provides a PCIe4.0x16 port.

Wugongshan OCP card DPU is a DPU using intel Icelake D+ Agilex FM8 FPGA architecture. It provides 2 100G DSFP service interfaces, 2 gigabit management interfaces, USB interfaces, and adopts AST2600 BMC for out-of-band management. The CPU supports two DDR4 RDIMMs and 13 DDR4 chips attached to the FPGA. Goldfinger provides PCIe4.0x16 ports.