HPE Cray XD680 GPU Server

Brand: HPE | Category: GPUs

SKU: R9G80A | Part #: R9G80A | MPN: R9G80A

Contact for Pricing — Request a Quote

AI/ML training and large-scale inference workloads demand GPU-dense architectures that deliver sustained throughput without thermal compromise. The HPE Cray XD680 GPU Server (part number R9G80A) addresses this requirement with direct liquid cooling and support for up to 8 double-width GPUs in a compact 4U form factor, making it the platform of choice for AI infrastructure teams deploying next-generation training clusters and inference farms.

Built on the HPE Cray XD Series, this 4U rack server combines dual Intel Xeon Scalable processors with PCIe Gen 5 connectivity, up to 32 DDR5 RDIMM/LRDIMM slots, and internal NVMe SSD support to create a balanced, high-performance compute node. Direct Liquid Cooling ensures thermal efficiency across dense GPU deployments, while HPE iLO management and Redundant high-efficiency power supplies provide the reliability and observability enterprise data centers require. The platform integrates seamlessly into HPE Slingshot high-speed interconnect fabrics and supports Red Hat Enterprise Linux, SUSE Linux Enterprise Server, and Ubuntu operating systems, enabling flexible deployment across heterogeneous HPC and AI environments.

Typical Workloads & Deployment Scenarios

• Large-scale AI/ML model training with multi-GPU parallelism and high-bandwidth GPU communication
• HPC simulation and computational fluid dynamics (CFD) requiring sustained GPU compute and low-latency interconnect
• Inference serving platforms that consolidate multiple model instances across high-density GPU clusters
• Research computing environments combining deep learning frameworks with traditional HPC simulation codes
• Data center consolidation projects migrating legacy GPU clusters to next-generation PCIe Gen 5 and DDR5 architectures

For detailed specifications, configuration options, and integration support tailored to your infrastructure roadmap, contact the Omnixon Global enterprise sales team to request an RFQ.