Product Features

Outstanding Performance, the Top Choice for AI Workloads

Equipped with 2nd Gen/3rd Gen Intel® Xeon® Scalable Processors (4th/5th Gen in original), featuring 4 UPI links and a maximum TDP of 350W.

Supports up to 10 GPUs with 450W power rating, delivering extreme heterogeneous computing power.

Enables x32 PCIe bandwidth between GPU and CPU, doubling the mainstream x16 bandwidth in the industry to meet high-bandwidth interconnection requirements.

Accommodates 32 DDR5 memory modules with a peak frequency of 5600MHz, boosting memory bandwidth by 75%.

Advanced Architecture & Flexible Configuration

Two deployment modes available: 8-GPU direct connection and 10-GPU switch architecture. Diverse GPU topologies adapt to various application scenarios.

Supports GPU Direct, as well as RDMA for storage/network and GPU P2P data transmission, ideal for cluster deployment.

Excellent scalability with up to 15 standard PCIe slots. Configurable as 8 double-width GPUs + 7 standard PCIe cards + 1 OCP 3.0 network adapter.

Flexible storage options to cater to high-capacity and high-performance local storage needs, supporting a maximum of 24 U.2 NVMe drives.

High Reliability & Intelligent Management

Critical system components adopt redundant and hot-swap design with tool-less installation and removal, greatly improving maintenance efficiency and system availability.

Built-in intelligent management chip and open management platform, compatible with IPMI 2.0, Redfish, SNMP and other mainstream management protocols.

Integrated with remote KVM, virtual media, component status monitoring and alarm functions, enabling comprehensive remote intelligent system management.

This GPU configuration supports a maximum of 8 or 10 full-height, full-length double-width GPU cards. All 8/10 GPUs are assigned to a single CPU. See the figure below.

For the cable connection of this GPU configuration, refer to the figure below.

 General Mode：

Cascade Mode：

Balance Mode：