NVIDIA
NVIDIA HGX H200 (Supermicro 6U 8-GPU)
NVIDIA HGX H200 8-GPU Hopper server — 1,128 GB HBM3e per node on the air-cooled Supermicro 6U SXM platform, offered in networking-optimized (1:1 GPU:NIC) and compute-optimized (1:2) builds.
8-GPU NVIDIA HGX H200 (Hopper) server, built on the air-cooled Supermicro 6U HGX H200 SXM platform. Eight H200 SXM GPUs deliver 141 GB of HBM3e each — 1,128 GB per node at roughly 4.8 TB/s per GPU — over a 4th-generation NVLink baseboard at 900 GB/s per GPU for large-model training and inference. Two profiles are offered: a networking-optimized 1:1 GPU:NIC build with dual Intel Xeon Platinum 8580 and eight NVIDIA ConnectX-7 NDR400 InfiniBand adapters (3.2 Tb/s per node), and a compute-optimized 1:2 build with dual Intel Xeon Platinum 8568Y+. Configuration and availability confirmed at quote.
Specifications
| GPU architecture | NVIDIA H200 (Hopper) |
|---|---|
| GPUs per node | 8× NVIDIA HGX H200 SXM |
| GPU memory | 141 GB HBM3e per GPU · 1,128 GB per node · ~4.8 TB/s per GPU |
| CPU | 2× Intel Xeon Platinum 8580 (1:1 networking-optimized) or 2× Intel Xeon Platinum 8568Y+ (1:2 compute-optimized) |
| NVLink fabric | 4th-gen NVLink baseboard · 900 GB/s per GPU |
| Scale-out networking | 1:1 GPU:NIC build — 8× NVIDIA ConnectX-7 NDR400 InfiniBand (400 Gb/s · 3.2 Tb/s per node); 1:2 build reduces NIC count for compute density |
| Form factor | 6U rackmount, air-cooled · unit of sale: 8-GPU node |
| Cooling | Air-cooled |
| GPU board power | 8× 700 W H200 SXM (5.6 kW GPU baseboard) |
| Condition | New |
| Warranty | 3-yr Supermicro on-site hardware warranty |
Specifications shown are headline figures. Confirm full configuration details when you request a quote.
Last updated
Request a quote for NVIDIA HGX H200 (Supermicro 6U 8-GPU)
Tell us your requirements and we'll get back to you with availability and pricing.
or reach us directly at