• Silver metallic supermicro as-8125gs-tnhr 8u gpu server with perforated front panel and dual epyc 9004 processors
1 1

Supermicro

Supermicro AS-8125GS-TNHR | 8U GPU Server, Dual EPYC 9004/9005, 8x HGX H100 SXM5

AS-8125GS-TNHR

POA
POA
POA
Sale Sold out
Taxes included. Shipping calculated at checkout.

Custom Built Servers

Need a different spec, quantity or configuration?

We build configured-to-order refurbished servers from Cisco, Dell, HPE, NVIDIA Lenovo and Supermicro. Tell us what you need and we'll come back with options within 24 hours.

Configure your server now →

Description

The Supermicro AS-8125GS-TNHR is an 8U rackmount GPU server engineered for large-scale AI training, high-performance computing, and deep learning workloads. It supports dual AMD EPYC 9004/9005 series processors, up to 6TB DDR5 memory across 24 DIMM slots, and 8x NVIDIA HGX H100 SXM5 GPUs interconnected with NVLink and NVSwitch for maximum GPU-to-GPU bandwidth. With 8 dedicated NICs for GPU-direct RDMA, PCIe 5.0 expansion, and up to 16 hot-swap NVMe bays, this system delivers enterprise-grade performance for demanding AI clusters and scientific computing environments.

Features

High-density 8U form factor optimized for NVIDIA HGX H100/H200 8-GPU configuration
- NVIDIA NVLink and NVSwitch interconnect for highest GPU-to-GPU communication bandwidth
- 8 dedicated NICs for GPU-direct RDMA with 1:1 GPU-to-NIC ratio, supporting 400G NVIDIA BlueField-3 or ConnectX-7
- Dual AMD EPYC 9004/9005 series processor support with up to 6TB DDR5 memory at 4800 MT/s
- PCIe 5.0 expansion with up to 12 slots (8x LP + 4x FHFL) for maximum I/O bandwidth
- Up to 18 hot-swap 2.5" drive bays (16x NVMe, 2x SATA) plus dedicated M.2 boot drive
- Advanced thermal design with 10 heavy-duty fans and optimized airflow for sustained GPU performance
- 6x 3000W redundant Titanium-level power supplies for high-efficiency operation
- IPMI 2.0 remote management with dedicated BMC port, KVM, and virtual media support
- Certified for Red Hat Enterprise Linux and compatible with major AI frameworks
- Scalable building block for AI clusters with support for NVIDIA Spectrum-X Ethernet or Quantum-2 InfiniBand

Warranty

All products sold by XS Network Tech include a 12-month warranty on both new and used items. Our in-house technical team thoroughly tests used hardware prior to sale to ensure enterprise-grade reliability.

All technical data should be verified on the manufacturer data sheets.

View full details

specs-tabs

Technical Specifications

FAQs

Technical Specifications

Form Factor: 8U rackmount GPU server
- Processors: Dual socket SP5 supporting AMD EPYC 9004/9005 series processors
- Memory: 24x DIMM slots, up to 6TB DDR5 RDIMM/LRDIMM, 4800 MT/s
- GPU: 8x NVIDIA HGX H100 SXM5 GPUs with 640GB HBM3 total, interconnected with NVLink and NVSwitch
- Storage: 12x 2.5" hot-swap NVMe drive bays (default), 2x 2.5" hot-swap SATA drive bays, 4x 2.5" optional NVMe bays, 1x M.2 NVMe for boot
- Network: 8x NICs for GPU-direct RDMA (1:1 GPU ratio), optional 2x 10GbE BaseT with Intel X550-AT2
- Expansion: 8x PCIe 5.0 x16 low-profile slots, up to 4x PCIe 5.0 x16 FHFL slots
- Management: IPMI 2.0 with dedicated RJ45 port, KVM, remote virtual media
- Cooling: 10 heavy-duty fans with optimal fan speed control
- Power Supply: 6x 3000W redundant power supplies (3+3 configuration), 80 PLUS Titanium efficiency
- Key Applications: AI/deep learning training, high-performance computing, industrial automation, climate modeling

FAQs

Q: What GPU configuration does the AS-8125GS-TNHR support?
A: The AS-8125GS-TNHR supports NVIDIA HGX H100 or H200 8-GPU SXM5 multi-GPU boards with full NVLink and NVSwitch interconnect for maximum GPU-to-GPU communication bandwidth. Each system includes 8 dedicated NICs in a 1:1 ratio with the GPUs to enable GPU-direct RDMA for scalable AI cluster deployments.

Q: What AMD processor generations are compatible with this server?
A: This server supports dual AMD EPYC 9004 series (Genoa) and 9005 series processors via Socket SP5. For optimal performance with the 8-GPU HGX configuration, Supermicro recommends processors with a minimum of 48 cores per socket, with 56 cores per socket recommended for demanding AI and HPC workloads.

Q: How much storage capacity does the AS-8125GS-TNHR provide?
A: The system includes 12 hot-swap 2.5" NVMe drive bays by default, 2 hot-swap 2.5" SATA drive bays, and an optional 4 additional 2.5" NVMe drive bays for a total of up to 18 drive bays. A dedicated M.2 NVMe slot is provided for the boot drive, keeping primary storage bays available for data.

Q: What networking options are available for this GPU server?
A: The AS-8125GS-TNHR includes 8 NIC ports specifically for GPU-direct RDMA in a 1:1 GPU-to-NIC ratio, supporting NVIDIA BlueField-3 or ConnectX-7 adapters with 400G capability for InfiniBand or Ethernet fabrics. Optional onboard networking includes dual 10GbE BaseT with Intel X550-AT2, plus flexible PCIe 5.0 expansion for additional network adapters.

Q: What are the typical use cases for this server?
A: The AS-8125GS-TNHR is designed for large-scale AI model training, deep learning with datasets requiring high GPU memory bandwidth, high-performance computing simulations, climate and weather modeling, and industrial automation applications that demand multi-GPU parallel processing with low-latency GPU interconnects.

  • Request a Quote

    Looking for competitive pricing? Submit a request, and our team will provide a tailored quote that fits your needs.

  • Contact Us Directly

    Have a question or need immediate assistance? Call us for expert advice and real-time support.

    Call us Now  

  • Contact Us Directly

    Have a question or need immediate assistance? Call us for expert advice and real-time support.

    Contact us