HPE Cray Supercomputing EX4000 System

The HPE Cray Supercomputing EX4000 System is a purpose-built, liquid-cooled blade chassis infrastructure designed to power the most demanding high-performance computing and large-scale AI training environments. Built on the proven HPE Cray EX architecture and launched in 2025, the EX4000 introduces a higher-density blade form factor with enhanced thermal management via direct liquid cooling, enabling sustained operation of the latest GPU accelerators including NVIDIA H100 and H200 SXM series alongside next-generation AMD Instinct accelerators. The chassis integrates native HPE Slingshot 200Gb/s high-bandwidth, low-latency interconnect fabric, eliminating the need for external top-of-rack switching at the blade level and reducing congestion at scale across tens of thousands of compute nodes.

The EX4000 chassis supports a flexible mix of compute blades, including CPU-only nodes based on AMD EPYC Genoa and Intel Xeon Scalable processors, as well as fully accelerated GPU blades optimized for tensor operations and large language model training. The system's centralized management fabric leverages HPE's Cray System Management (CSM) software stack integrated with HPE Performance Cluster Manager (HPCM), providing unified monitoring, firmware orchestration, and workload scheduling across the entire supercomputing infrastructure. The liquid-cooled design achieves Power Usage Effectiveness (PUE) values approaching 1.03 in optimized data center environments, significantly reducing operational energy overhead compared to air-cooled alternatives.

The EX4000 is engineered for deployment in national laboratories, hyperscale AI research facilities, defense simulation environments, and enterprise HPC clusters requiring exascale-class throughput. Its modular blade and chassis architecture allows incremental scaling from a single cabinet to multi-thousand-node configurations without architectural redesign. Tight integration with HPE's Cray Programming Environment (CPE) and support for industry-standard MPI, OpenMP, CUDA, and ROCm programming models ensures compatibility with existing HPC application portfolios while enabling optimized performance on the latest accelerated compute blades.

Ideal for

• Large Language Model (LLM) pre-training and fine-tuning at scale using GPU blade clusters with NVLink and Slingshot fabric for all-reduce collective operations
• Climate and weather simulation workloads requiring tight MPI coupling across thousands of CPU and GPU nodes with sub-microsecond interconnect latency
• Computational fluid dynamics (CFD) and structural mechanics simulations for aerospace and automotive engineering at national laboratory scale
• Genomics and drug discovery pipelines leveraging GPU-accelerated molecular dynamics with high-throughput storage integration via HPE Cray ClusterStor
• Defense and intelligence community mission-critical simulation workloads requiring certified secure, air-gap-capable supercomputing infrastructure
• Seismic imaging and reservoir modeling for energy sector HPC applications demanding extreme memory bandwidth and low-latency node interconnect

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.