Supermicro AS-8125GS-TNHR 8x NVIDIA H100 SXM5 (AMD EPYC 9004)
AMD EPYC flagship for full-stack large-model training
We help you choose, configure, and deliver the right system — no obligation.
Configuration at a Glance
Tailored per engagement. Full technical overview below.
Configuration Options
Core specifications for this system. Every component is configurable to your workload — request a quote for a tailored build.
8x NVIDIA H100 SXM5 (80GB HBM3 each, 640GB total)
2x AMD EPYC 9004 Series (up to 128 cores total)
Up to 6TB DDR5-4800 ECC RDIMM (24 DIMM slots)
12x 2.5" hot-swap NVMe + 2x 2.5" SATA bays
Overview
The AS-8125GS-TNHR is an 8U dual-socket AMD EPYC platform configured with eight NVIDIA H100 SXM5 GPUs fully interconnected through NVLink and NVSwitch for non-blocking multi-GPU training. Each unit is built to a validated specification, burn-in tested, and delivered warranty-backed through authorized Supermicro channels.
Who This Solution Is For
Business Benefits
Non-blocking GPU fabric
Four NVSwitches deliver 900GB/s all-to-all NVLink so eight H100s scale as a single accelerator.
High PCIe lane density
Dual EPYC 9004 CPUs expose abundant PCIe 5.0 lanes for 1:1 GPU-to-NIC RDMA networking.
Validated and warranty-backed
Every system arrives pre-configured, thermally tested, and supported through authorized Supermicro service.
Typical Business Use Cases
Foundation model pre-training
Distributed deep-learning training
Mixed HPC and AI simulation
Multi-node GPU cluster building blocks
Industry Applications
Technical Overview
Built on the Supermicro 8U Universal GPU architecture, the system pairs the NVIDIA HGX H100 8-GPU baseboard with dual SP5 AMD EPYC 9004 processors across 24 DDR5 DIMM slots. NVLink and four NVSwitch chips provide 900GB/s of GPU-to-GPU bandwidth, while PCIe 5.0 connectivity supports eight dedicated RDMA network adapters.
| GPU/Accelerator | 8x NVIDIA H100 SXM5 (80GB HBM3 each, 640GB total) |
| GPU Interconnect | NVLink + 4x NVSwitch, 900GB/s GPU-to-GPU |
| CPU | 2x AMD EPYC 9004 Series (up to 128 cores total) |
| Memory | Up to 6TB DDR5-4800 ECC RDIMM (24 DIMM slots) |
| Storage | 12x 2.5" hot-swap NVMe + 2x 2.5" SATA bays |
| Networking/Fabric | 8x 400Gb NICs for 1:1 GPU RDMA + onboard management LAN |
| Form Factor | 8U rackmount, air-cooled |
| Power | 6x 3000W (3+3) redundant, 80 PLUS Titanium |
| Warranty | Manufacturer warranty via authorized channel; extended options available |
Specifications are indicative and configured to each engagement. Request a quote for a configuration tailored to your requirements.
Warranty, Support & Fulfillment
Every system ships from an authorized channel, configured and tested, with the documentation enterprise buyers need — backed by warranty and a dedicated account team.
Enterprise Warranty
Full manufacturer warranty with optional on-site, next-business-day support and extended coverage.
Authorized Channel
Sourced through Tier-1 distribution and OEM partners — never grey market. Asset & warranty records included.
Lead Time & Deployment
48-hour quotes, then configured, burn-in tested, and delivered on a committed schedule.
Nationwide Fulfillment
Coordinated logistics, rack-and-stack, and delivery wherever your infrastructure lives.
Frequently Asked Questions
How many H100 GPUs can this server hold?
It is configured with the full NVIDIA HGX H100 8-GPU baseboard, providing eight SXM5 GPUs and 640GB of aggregate HBM3 in a single 8U node.
Is this suited to large language model training?
Yes; the 900GB/s NVSwitch fabric and 1:1 GPU-to-NIC RDMA networking make it an ideal building block for single-node and multi-node LLM pre-training.
Why choose the AMD EPYC version over the Intel platform?
The EPYC 9004 configuration offers high core counts and generous PCIe 5.0 lane availability, which benefits data-loading pipelines and dense RDMA fabric layouts.
Hardware Assistance
Configure the Supermicro AS-8125GS-TNHR 8x NVIDIA H100 SXM5 (AMD EPYC 9004) with Nexus Compute
Tell us your requirements and a hardware specialist will help you specify, configure, and quote the right system — typically within two business days. No obligation.