8 GPU H200 NVL PCIe AI server
Available on request- GPUs
- 8 x NVIDIA H200 NVL, 141 GB HBM3e each
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Chassis
- 4U, about 6.5 kW, air cooled
NVIDIA H200 NVL
Hopper in a PCIe card with 141 GB HBM3e at 4.8 TB/s, NVLink bridges for 2 or 4 GPUs and up to 600 W, for inference and fine tuning in air cooled servers.
- Family
- Hopper
- Edition
- PCIe card, dual slot, air cooled
- Memory
- 141 GB HBM3e per GPU
- Memory bandwidth
- 4.8 TB/s per GPU
- Interconnect
- 2 way or 4 way NVLink bridge at 900 GB/s per GPU, PCIe Gen5 x16 at 128 GB/s
- Form factor
- Dual slot PCIe card, full height full length, passive
- TDP
- Up to 600 W (configurable)
- Cooling
- Passive
- Best for
- Inference and fine tuning in PCIe servers from 2 to 8 GPUs, bridged in pairs or fours
- Availability
- Available on request
H200 NVL brings the Hopper H200 to a dual slot PCIe form factor so it fits in standard 2U and 4U air cooled servers. Each card has 141 GB of HBM3e at 4.8 TB/s, and NVLink bridges join two or four cards at 900 GB/s per GPU, which lets a 70B class model run across a bridged pair with memory to spare. NVIDIA publishes 3,341 TFLOPS of sparse FP8 and up to seven MIG instances of 16.5 GB each.
The card is passively cooled and depends on chassis airflow, so we quote it only in servers validated for 600 W PCIe GPUs. It is the right choice when a team wants large HBM capacity in a PCIe server, for example for inference serving, RAG pipelines or fine tuning, without moving to an HGX baseboard. It is supplied in 2, 4 and 8 GPU PCIe servers and as a replacement card, subject to end user registration.
Systems that carry this GPU
MGX 8 GPU AI server
- GPUs
- 8 x RTX PRO 6000 Blackwell Server Edition
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Chassis
- 4U, about 6.5 kW, air cooled
4 GPU PCIe AI server, 2U
- GPUs
- 4 x RTX PRO 6000 Blackwell Server Edition
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 1.5 TB DDR5 RDIMM
- Chassis
- 2U, about 3.5 kW, air cooled
4 GPU PCIe AI server, 4U
- GPUs
- 4 x RTX PRO 6000 Blackwell Server Edition
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Chassis
- 4U, about 3.5 kW, air cooled
2 GPU 2U inference server
- GPUs
- 2 x RTX PRO 6000 Blackwell Server Edition (96 GB)
- CPU
- 1 or 2 x AMD EPYC 9005, or 2 x Intel Xeon 6
- Memory
- Up to 1.5 TB DDR5 RDIMM
- Chassis
- 2U, about 2 kW, air cooled
NVIDIA H200 NVL GPU
Available on request- Memory
- 141 GB HBM3e, 4.8 TB/s
- Compute
- Hopper, 3,341 TFLOPS sparse FP8
- Interconnect
- 2 way or 4 way NVLink bridge at 900 GB/s per GPU
- Power
- Up to 600 W configurable
Related guides
Buying data center GPUs, export and registration
The compliance steps that come with every HGX, H200 NVL, Server Edition and NVL72 purchase, explained so your order ships on time.
In this guide
- Why a GPU purchase comes with paperwork
- Which products are covered
Choosing a GPU server for LLM inference
A practical method for sizing an inference server from the model, the quantization and the concurrency you expect, with worked examples.
In this guide
- Start with the model, not the server
- Step 1: how much GPU memory the weights need
HGX vs PCIe GPU servers
The real differences between an HGX node and a PCIe GPU server, and a simple rule for choosing between them.
In this guide
- Two kinds of 8 GPU server
- The interconnect is the point
Want this GPU in a different chassis?
We quote every class that carries it, from a single node to a pod. Tell us the workload and we propose the chassis.
- (858) 613-1818
- sales@computermodules.com
- Monday to Friday, 8am to 5pm Pacific