NVIDIA HGX H100 8-GPU Baseboard
Brand: NVIDIA | Category: GPUs
SKU: NVIDIA-HGX-H100
Contact for Pricing — Request a Quote
About the NVIDIA HGX H100 8-GPU Baseboard
NVIDIA HGX H100 8-GPU Baseboard
The NVIDIA HGX H100 baseboard is the NVLink-interconnected multi-GPU compute substrate that powers DGX H100 systems and OEM AI servers from HPE, Dell, Lenovo, and Supermicro. Mounting 8× H100 SXM5 GPUs interconnected by NVLink 4.0 at 900 GB/s per GPU and NVSwitch 3.0 fabric, the HGX H100 baseboard enables all-to-all GPU communication at memory bandwidth that makes 8-GPU training 8× faster than 8 isolated PCIe GPUs — eliminating the PCIe bottleneck that cripples large language model training on non-NVLink platforms.
HGX vs DGX: What You're Buying
The HGX H100 baseboard is the compute subsystem — without CPU host, storage, or system chassis. OEM integrators (HPE ProLiant Compute, Dell PowerEdge XE9680, Supermicro SYS-4U) mount the HGX baseboard alongside their own processors, memory, networking, and enclosures. For UAE AI data-centre operators who have already selected a preferred server vendor, the HGX H100 provides the full NVLink GPU performance of a DGX H100 in their vendor's chassis and support ecosystem.
Key Specifications
| Parameter | Value |
|---|---|
| GPUs | 8× NVIDIA H100 SXM5 80GB |
| Total GPU Memory | 640GB HBM3 |
| GPU-to-GPU Bandwidth | 900 GB/s (NVLink 4.0) |
| NVSwitch | 4× NVSwitch 3.0 |
| Total FP8 Performance | 16 PFLOPS |
| PCIe Host Interface | 8× PCIe 5.0 x16 |
| Total TDP | ~10.2 kW (8× 700W GPU + switch) |
Technical Specifications
| Brand | NVIDIA |
| Category | GPUs |
| SKU | NVIDIA-HGX-H100 |
| Condition | New |
| Capacity | 80GB |
| Form Factor | OEM Baseboard |
| RAID Support | H100 |
| GPU Support | NVIDIA H100 |
| GPU Count | 8x |
| AI Optimized | Yes |
| GPUs | 8x H100 80GB SXM5 |
| Total GPU Memory | 640GB HBM3 |
| Interconnect | NVSwitch 3.0 |
| Form Factor | OEM Baseboard |
| Use Case | Custom AI Server Building Block |
Frequently Asked Questions about NVIDIA HGX H100 8-GPU Baseboard
What does the NVIDIA HGX H100 8-GPU Baseboard do?
The NVIDIA HGX H100 8-GPU Baseboard accelerates AI/ML training, inference, scientific HPC, and virtualization (vGPU) workloads. Typical deployments include LLM training clusters, computer-vision pipelines, financial risk modeling, and rendering farms.
What are the headline specs of the NVIDIA HGX H100 8-GPU Baseboard?
Key specifications for the NVIDIA HGX H100 8-GPU Baseboard: new condition; brand NVIDIA; series HGX; model HGX H100; gpus 8x H100 80GB SXM5; total gpu memory 640GB HBM3; interconnect NVSwitch 3.0. Manufacturer part number NVIDIA-HGX-H100. For the full datasheet with electrical, environmental, and compliance details, contact our pre-sales engineering team.
Is the NVIDIA HGX H100 8-GPU Baseboard compatible with my infrastructure?
The NVIDIA HGX H100 8-GPU Baseboard requires a PCIe Gen4 or Gen5 x16 slot, server power adequate for the card's TDP, and CUDA/ROCm driver support in your hypervisor or bare-metal OS. Sales engineering will confirm chassis fit (1U/2U/4U), PCIe lane count, and PSU headroom before quoting.