An 8-GPU Hopper server from Supermicro with 80 GB HBM3 per GPU. The HGX H100 has been succeeded by the H200 and Blackwell B300 platforms. Limited secondary market availability through Arc Compute.
Eight H100 SXM GPUs on the HGX H100 baseboard with 80 GB HBM3 per GPU and 640 GB total GPU memory. Fully interconnected via NVLink with NVSwitch for high-bandwidth GPU-to-GPU communication.
Dual Intel Xeon Platinum 8558P processors with 48 cores and 97 threads per socket. 32x DDR5 5600MT/s DIMM slots. A proven CPU platform with broad enterprise compatibility.
Supports air-cooling with 10 heavy-duty fans and optimal speed control in the air-cooled variant. Designed for sustained high-utilization operation.
The H100 is the most widely deployed GPU in production AI infrastructure. Broadest software ecosystem support, deepest driver maturity, and the largest base of real-world deployment references of any data center GPU.
The NVIDIA H100 was the GPU that launched the modern era of AI infrastructure. With 80 GB HBM3 per GPU, third-generation NVLink, and the Transformer Engine, the H100 set the performance standard for large-scale model training and inference. It remains the most widely deployed data center GPU with the deepest ecosystem of validated software, frameworks, and production references.
The H100 has since been succeeded by the H200 (141 GB HBM3e, higher memory bandwidth) and the Blackwell B300 (288 GB HBM3e, NVLink 5). For organizations expanding existing H100 clusters, the Supermicro SYS-821GE-TNHR H100 may be available through the secondary market. For new deployments, Arc Compute recommends evaluating H200 or B300 platforms.
Every cluster is designed around your workloads, your facility, and your operational model.
Systems are validated, burned in, and shipped with full deployment support.
Hands-on experience designing and deploying high-density GPU infrastructure.
From planning and procurement through deployment, operations, and scaling.
