Overview
Supermicro 10U B200 Gold Series (SYS-A21GE-NBRT-G1) integrates NVIDIA HGX B200 — 8x Blackwell B200 GPUs (192GB HBM3e each) with NVLink/NVSwitch interconnect. At $792,000, this is cutting-edge AI training infrastructure for organizations working on frontier models requiring terabytes of GPU memory. B200 delivers ~2.5× the performance of H100.
Key Features
- 8x NVIDIA HGX B200 GPUs — 192GB HBM3e each (1.5TB total)
- NVLink/NVSwitch GPU-to-GPU fabric
- Blackwell architecture — 2.5× H100 compute
- 10U chassis with high-density GPU configuration
- Dual Intel Xeon host CPUs + redundant PSU
Ideal For
Frontier AI labs, cloud providers, and government research centers building next-generation LLM training clusters who require the maximum per-server GPU memory at Blackwell's flagship compute density.
Feature
8x NVIDIA B200 SXM GPUs, 1.4TB HBM3e GPU Memory Space
Feature
32x DIMM slots, DDR5 up to 5600MT/s(1DPC), 4800MT/s(2DPC)
Feature
2x 10GbE RJ-45, 1x Dedicated BMC/IPMI
Feature
6x 5250W Redundant (3+3) Power Supplies Titanium Level Efficiency
$792,000.00
Prices may vary. Verify on vendor site.
Quick Specs
- Feature
- 6x 5250W Redundant (3+3) Power Supplies Titanium Level Efficiency
Tags
model-trainingllm-inferencehpc