Skip to main content
Out of stock
  • $264,000.00
  • Free 72-day Returns
  • $16.14 Shipping ()
Sold by Newegg

Shipped by Newegg

Price alert
Shop it at NeweggBusiness.

GIGABYTE G593-ZD2-AAX1 HPC/AI Server - AMD EPYC™ 9004 - 5U DP NVIDIA HGX™ H100 8-GPU 4-Root Port

  • NVIDIA-Certified Systems™ - Data Center Servers
  • NVIDIA HGX™ H100 with 8 x SXM5 GPUs
  • 900GB/s GPU-to-GPU bandwidth with NVIDIA® NVLink® and NVSwitch™
  • Dual AMD EPYC™ 9004 Series Processors (with AMD 3D V-Cache™ Technology)
  • 12-Channel DDR5 RDIMM, 24 x DIMMs
  • Dual ROM Architecture
  • 2 x 10Gb/s LAN ports via Intel® X710-AT2
  • 2 x M.2 slots with PCIe Gen3 x4 and x1 interface
  • 8 x 2.5" Gen5 NVMe/SATA/SAS-4 hot-swappable bays
  • 4 x FHHL PCIe Gen5 x16 slots
  • 8 x LP PCIe Gen5 x16 slots
  • 4+2 3000W 80 PLUS Titanium redundant power supplies
+
+
Overview
Reviews
See all Q & A

Any questions? Ask CelesteSM will help you find out quickly.

Warranty & Returns

Warranty, Returns, And Additional Information

Warranty

  • Limited Warranty period (parts): 3 years
  • Limited Warranty period (labor): 3 years
  • Read full details

Return Policies

    We want to ensure your holiday shopping this year is hassle-free. That’s why we’re automatically extending our return policies on almost every product! For purchases placed between November 1st, 2026 and December 31, 2026 you will have until January 31, 2027 to return your product.

  • Return for refund until: January 31, 2027
  • Return for replacement until: January 31, 2027
  • This item is covered by Newegg.com's Extended Holiday Return Policy.
  • Read full return policy for details.
LOADING...

The GIGABYTE G593-ZD2-AAX1 is a 5U HPC/AI server with NVIDIA HGX H100 and eight SXM5 GPUs, built for data-center teams running GPU-intensive AI and high-performance computing workloads. Its NVLink and NVSwitch interconnect provides 900GB/s GPU-to-GPU bandwidth.

Recommended For

  • AI research and infrastructure teams training or fine-tuning large models on one server, where eight H100 GPUs and the NVLink/NVSwitch fabric support tightly coupled multi-GPU workloads.
  • Enterprise AI teams serving large inference workloads on-premises, especially when they need an eight-GPU node for shared, multi-user compute rather than a single-GPU workstation.
  • HPC teams running GPU-accelerated scientific or engineering workloads that can use multiple GPUs in one server and are equipped to rack and operate data-center hardware.
  • Data-center teams adding a node to an enterprise AI cluster, with staff and infrastructure to integrate the server into a separate high-speed network and deployment environment.

Spec Summary

AttributeValue
Best ForHPC and AI workloads requiring an NVIDIA HGX H100 8-GPU server platform