LenovoThinkSystem SR780a V3

Eight NVIDIA B200 GPUs on an HGX board and two 5th Gen Intel Xeon processors in 5U, cooled with warm water by the Lenovo Neptune system. It's how you fit an HGX into less rack space when the data centre has a water loop.

Ask for price and lead timeSee what it supports
Lenovo ThinkSystem SR780a V3Rack 5U

What it comes with

The HGX board can be B200 or H100; everything else is configurable. 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 B200 or HGX H1008 GPUs: 180 GB per GPU on the B200
MemoryUp to 4 TB32 DDR5 slots · 5600 MHz with one DIMM per channel
StorageUp to 12 × 2.5" NVMe8 front and 4 rear · boot from 2 M.2
NetworkUp to 11 PCIe 5.0 slots8 with GPU Direct, for adapters of up to 400 Gb/s
Form factorRack 5ULenovo Neptune liquid cooling
Power
Eight 2,600 W Titanium power supplies, up to N+N
Cooling
Warm water for CPUs, GPUs and NVLink: removes up to 75% of the heat
Management
XClarity Controller 2
Operating systems
RHEL 9.4 to 9.6, Ubuntu 22.04 and 24.04 LTS and VMware ESXi 8.0 U3
Warranty
3 years, with an on-site engineer next business day

Eight GPUs in 5U, water-cooled

It does the same work as an air-cooled HGX, training and serving large models with the eight GPUs linked by NVLink, but in 5U instead of 8U. It manages that by taking warm water to the processors, the GPUs and the NVLink switches, which removes up to 75% of the heat; the fans handle the rest.

It makes sense when rack space is expensive or the room can't take any more air-cooled load, provided the data centre has, or will have, a liquid cooling loop. If it doesn't, the alternative is the SR680a V3 with B200 or the SR680a V4.

What we do

We install it with RHEL, Ubuntu or VMware ESXi 8.0, the systems Lenovo certifies, and set up the training or inference stack the job calls for: Slurm and NCCL, or Kubernetes with the GPU Operator and vLLM. Before you order it we go through the water loop it will connect to, with you and with whoever maintains the room.

The GPU board

GPUMemoryHow many
NVIDIA B200 (HGX B200)180 GB HBM3e8, at 1,000 W each
NVIDIA H100 (HGX H100)80 GB HBM38, at 700 W each

The HGX H200 version of this model has been withdrawn.

What your data centre needs

It takes 5U and has eight 2,600 W power supplies, with redundancy up to N+N. It needs a warm-water loop in the room, because up to 75% of the heat leaves that way, and there's no air-cooled version. Flow rate, temperature and connections are checked before anything is agreed.

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 SR680a V3 with B200NVIDIA HGX B200 · 8 GPUsThe previous HGX, also air-cooled, when lead time or budget comes first.
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 SR780a V3

What does my data centre need for the SR780a V3?

A liquid cooling loop with warm water to connect the server to, plus the power for eight 2,600 W supplies. If your room doesn't have one, the SR680a V3 with B200 and the SR680a V4 carry an air-cooled HGX in 8U.

B200 or H100?

The B200 has 180 GB per GPU against 80 GB on the H100 and belongs to the Blackwell generation. The H100 is still available on this model for those who need that platform; the H200 version has been withdrawn.

What warranty does it come with?

Three years, with an on-site engineer next business day. On top of that we can provide our 24×7 on-call support with SLA.

Is your data centre ready for a water-cooled HGX?

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

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