Lenovo ThinkSystem SR680a V3 Rack Server

Delivering Massive Computational Performance for AI and HPC

8x NVIDIA H100 GPUs (640GB of total GPU memory at 3.3TB/s) or 8x NVIDIA H200 GPUs (1,128GB of total GPU memory at 4.8TB/s), utilizing 4x NVIDIA NVSwitches for a total of 7.2TB/s of bi-directional GPU-to-GPU bandwidth
8x AMD Instinct™ MI300X (1,536GB of total GPU memory at 5.2TB/s) for a total of 6.4TB/s of bi-directional GPU-to-GPU bandwidth
Two 5^th Gen Intel^® Xeon^® Scalable processors with 32 DDR5 DIMM sockets
Up to 10x PCIe Gen5 x16 slots with 8x directly connected to the GPU complex
Up to 16x 2.5″ NVMe drives for maximum storage capacity and performance

Massive Computational Performance

The ThinkSystem SR680a V3 offers groundbreaking performance with two 5th Gen Intel® Xeon® Scalable processors. With support for 32x DDR5 memory at 5600MHz, up to 4TB.

It is designed to maximize GPU performance with support for eight of the latest high-speed interconnected GPUs to drive innovations in AI like LLM, machine learning, model training, and the most demanding HPC workloads.

Experience the fastest transfer rates with the ThinkSystem SR680a V3 utilizing NVIDIA NVLink (up to 900GB/Sec) or AMD’s Infinity Fabric (up to 896GB/Sec).

Flexible Acceleration

The ThinkSystem SR680a V3 supports the industry’s leading GPU platforms for AI: NVIDIA’s HGX™ H100/H200/Blackwell or AMD’s Instinct™ MI300X GPUs in an 8U frame that fits into an industry-standard 19″ rack.

NVIDIA H100/H200/Blackwell Tensor Core GPU platforms deliver unprecedented acceleration — at every scale — to power the world’s highest-performing elastic data centers for AI and HPC applications.

The AMD Instinct™ MI300X discrete GPU is based on next-generation AMD CDNA™ 3 architecture and delivers leadership efficiency and performance for compute-intensive AI and HPC workloads.

Resilient, Simple & Future-Forward

Lenovo designed the ThinkSystem SR680a V3 with their customer and the future in mind. It offers resilient power supplies, thermals for next-generation GPUs, and easy systems management.

The 8x CFFV4 power supplies in the ThinkSystem SR680a support full N+N redundancy without throttling, maximizing resiliency.

Looking to support the next generation of higher-power GPUs, the ThinkSystem SR680a is Air-cooled with N+1 hot-swap fans, and has thermal headroom for future GPUs.

The ThinkSystem SR680a V3 includes efficient deployment and simplified management with Lenovo XClarity software.

See Standard specifications

Lenovo ThinkSystem SR680a V3 Rack Server

Specifications

Form Factor/Height

8U rack

Processors

2x 5th Gen Intel Xeon Scalable processors, up to 350W

Memory

Up to 4TB using 32x DDR5 DIMMs with maximum frequency at 5600MHz

GPU

Supports 8x high-performance GPUs:

8x NVIDIA H100/H200/Blackwell GPUs with NVLink interconnects at 900 GB/s
8x AMD Instinct™ MI300X GPUs with Infinity Fabric interconnects at 896 GB/s

I/O Expansion

Up to 10x PCIe Gen5 x16 FHHL adapters

8x in front connected to PCIe switch for GPU connectivity
2x in rear connected to CPU for CPU connectivity

Storage

Up to 16x 2.5″ hot-swap NVMe SSDs
Up to 2x non-hot swap M.2 for boot (RAID via VROC)

Power

Up to 8x hot-swap power supplies, allowing full N+N redundancy

Cooling

Air-cooled with N+1 hot-swap fan solution

Management

XClarity Controller2 (XCC2), which provides advanced service-processor control, monitoring, and alerting functions. The XCC2 consolidates the service processor functionality, super I/O, video controller, and remote presence capabilities into a single chip on the server system board

OS Support

RHEL, Ubuntu, Alma Linux, Rocky Linux, ESXi

Datasheet

Lenovopress.com/ds0180

Ask About Other Model and Parts number