HPE Cray XD675 (8x AMD Instinct MI300X OAM, Dual 4th Gen AMD EPYC)
1.5TB of HBM3 across eight MI300X for memory-bound AI and HPC.
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 AMD Instinct MI300X OAM, 192GB HBM3 each (1.5TB total)
Dual 4th Gen AMD EPYC (Genoa), up to 400W TDP
Up to 24x DDR5-5600 DIMMs
Up to 16x SFF NVMe SSDs + 2x M.2
Overview
The HPE Cray XD675 in this configuration hosts eight AMD Instinct MI300X OAM accelerators with dual 4th Gen AMD EPYC processors, delivering 1.5TB of GPU memory in a single node. Nexus Compute provides it specified, integrated, and tested on the ROCm stack with direct liquid cooling and HPE warranty through authorized channels.
Who This Solution Is For
Business Benefits
Massive GPU memory
1.5TB of HBM3 across eight MI300X lets very large models load and serve with less sharding overhead.
Open ROCm software
The AMD ROCm ecosystem offers an open, portable path for AI and HPC codebases.
Supply diversification
An AMD Instinct platform broadens accelerator sourcing options beyond a single vendor.
Typical Business Use Cases
Large-model LLM inference and serving
Memory-bound AI training
FP64-intensive HPC simulation
Mixed AI and scientific computing workloads
Industry Applications
Technical Overview
The XD675 carries an eight-way AMD Instinct MI300X OAM baseboard with Infinity Fabric GPU links, dual 4th Gen AMD EPYC Genoa CPUs, and up to 24 DDR5-5600 DIMMs. The chassis supports direct liquid cooling and eight PCIe Gen5 fabric slots for InfiniBand NDR or Ethernet, delivering roughly 10.4 PFLOPs peak FP16 per server.
| GPU / Accelerator | 8x AMD Instinct MI300X OAM, 192GB HBM3 each (1.5TB total) |
| GPU Interconnect | AMD Infinity Fabric (OAM baseboard) |
| CPU | Dual 4th Gen AMD EPYC (Genoa), up to 400W TDP |
| Memory | Up to 24x DDR5-5600 DIMMs |
| Storage | Up to 16x SFF NVMe SSDs + 2x M.2 |
| Networking / Fabric | 8x PCIe Gen5: InfiniBand NDR / Ethernet |
| Form Factor | 5U GPU node |
| Cooling | Direct liquid cooling |
| Warranty | HPE support, configurable terms |
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
When does MI300X beat an NVIDIA node for our workload?
MI300X shines on memory-bound jobs: 192GB per GPU and 1.5TB per node let large models fit with minimal sharding, which can simplify serving and improve throughput. We assess your models and software to confirm fit.
Will our existing code run on ROCm?
Major frameworks like PyTorch and many HPC libraries support ROCm, and AMD provides HIP tooling to port CUDA code. We help validate your stack before deployment.
Is the XD675 suited to double-precision HPC?
Yes. MI300X delivers strong FP64 performance, making this node effective for scientific simulation as well as AI. We tune the configuration to your balance of AI and HPC work.
Hardware Assistance
Configure the HPE Cray XD675 (8x AMD Instinct MI300X OAM, Dual 4th Gen AMD EPYC) 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.