LenovoThinkSystem SR680a V3 with B200

The NVIDIA HGX B200 board with eight 180 GB GPUs and two 5th Gen Intel Xeon processors, in an air-cooled 8U. It's Lenovo's previous HGX generation, and it still makes sense when lead time or budget matter more than having the latest.

Ask for price and lead timeSee what it supports
Lenovo ThinkSystem SR680a V3 with B200Rack 8U

What it comes with

The HGX board is fixed; what you configure is the processors, memory, drives and network. These are its limits according to Lenovo's guide. Lenovo Press product guide

Processor2 × 5th Gen Intel XeonUp to 64 cores and 350 W
GPUNVIDIA HGX B2008 × 180 GB GPUs · NVLink
MemoryUp to 4 TB32 DDR5 slots · 5600 MHz with one DIMM per channel
StorageUp to 16 × 2.5" NVMePCIe 5.0 · boot from 2 M.2
Network10 front PCIe 5.0 slots8 with GPU Direct, for adapters of up to 400 Gb/s
Form factorRack 8UConnections at the front
Power
Eight 3,200 W Titanium power supplies, up to N+N
Cooling
Air: 2 front fans and 15 rear fans for the GPUs
Management
XClarity Controller 2
Operating systems
RHEL 9.5 and 9.6, Ubuntu 22.04 and 24.04 LTS
Warranty
3 years, with an on-site engineer next business day

An air-cooled HGX from the previous generation

The eight B200s sit on the same board and communicate over NVLink, so it does the same job as the SR680a V4: training, fine-tuning and serving large models split across all the GPUs. Each one has 180 GB, against 288 GB on the B300, and that sets the size of model that fits comfortably.

Don't confuse it with the SR680a V3 with H100 or H200, which Lenovo has already withdrawn: this is the B200 variant, which is still in the catalogue.

What we do

We install it with RHEL or Ubuntu, the systems Lenovo certifies, and the stack the job calls for: Slurm and NCCL for multi-node training, or Kubernetes with the GPU Operator and vLLM for serving models. The eight slots with GPU Direct take adapters of up to 400 Gb/s for linking nodes or reading from storage without going through CPU memory.

The GPU board

GPUMemoryHow many
NVIDIA B200 (HGX B200)180 GB8, at 1,000 W each

The HGX B200 board always carries all eight GPUs.

What your data centre needs

It takes 8U and has eight 3,200 W power supplies, with redundancy up to N+N, and each GPU can draw up to 1,000 W. It's air-cooled with fifteen rear fans for the GPUs, so the rack needs plenty of power and a cold aisle that works. We go through the room with you before the order is placed.

Compared with the rest of the family

ModelWhat it hasWhen to choose it
ThinkSystem SR650a V4Up to 4 RTX PRO 6000 · Intel Xeon 6The entry point: one to four GPUs in 2U, with Intel.
ThinkSystem SR675 V3Up to 8 PCIe GPUs · AMD EPYCUp to eight GPUs with AMD, for several models or teams at once.
ThinkSystem SR680a V4NVIDIA HGX B300 · 8 GPUsEight B300s linked by NVLink, for training and serving large models.
ThinkSystem SR780a V3NVIDIA HGX B200 · liquid-cooledThe same eight B200s in 5U, if your data centre has a water loop.
ThinkSystem SR665 V34 RTX PRO 4500 · AMD EPYCLenovo's entry point for inference, with the Hybrid AI 241 configuration.

See all Lenovo AI and GPU servers

What the delivery includes

From the factory

  • Server built to your configuration by Lenovo
  • 1- or 3-year warranty depending on the model, with an on-site engineer next business day

From us

  • GPU, memory and network sizing based on your model
  • Rack mounting, XClarity and up-to-date firmware
  • Operating system, NVIDIA drivers and the AI stack
  • Testing with your workload before it goes into production

If you need it

  • 24×7 on-call support with SLA
  • Lenovo warranty upgrades: 4- or 2-hour response, or one or two extra years

How we build it

What runs on top of these servers, explained in detail:

RAG over your own documents with DoclingAI inference with Ceph, OpenStack and KubernetesRocky Linux for AI, with NVIDIA drivers and CUDA readyAll our AI services

Support and training for your team

Independent support from the people who installed itLocal LLM and GPU infrastructure courseDeep learning course with PyTorch and TensorFlowRed Hat OpenShift training

Questions about the ThinkSystem SR680a V3 with B200

SR680a V3 with B200 or SR680a V4?

The V4 carries the B300, with 288 GB per GPU, Intel Xeon 6 and onboard 800 Gb/s networking. The V3 with B200 has 180 GB per GPU and 5th Gen Xeon. If the model you'll run fits comfortably in 180 GB per GPU, the V3 does the job; if memory is tight or you're going to grow, the V4.

Is it still in the catalogue?

Yes. Lenovo has withdrawn the SR680a V3 variants with H100 and H200, but the B200 version has its own product guide, updated in September 2026.

Do I need liquid cooling?

No, it's air-cooled. If your data centre has a water loop, the SR780a V3 carries the same HGX B200 in 5U with liquid cooling.

Is the SR680a V3 with B200 right for your project?

Tell us which model you want to train or serve and what data centre you have, and we'll come back with the configuration, what the room needs and a confirmed lead time.

Ask for price and lead timeWhatsApp+34 91 198 02 43