HPE Cray XD670 AI Server (8x GPU)

The HPE Cray XD670 AI Server is a dense, high-performance computing platform engineered specifically for GPU-accelerated artificial intelligence applications. The system accommodates up to 8 NVIDIA H100 or H200 SXM GPUs in a single 4U form factor, connected via NVIDIA NVLink for ultra-low-latency inter-GPU communication essential for distributed training of large language models and foundation models. The server integrates direct liquid cooling for GPUs and CPU to manage thermal density while maintaining operational reliability at full compute utilization.

Architecturally, the XD670 combines dual-socket AMD EPYC processors with high-bandwidth PCIe 5.0 infrastructure, delivering the CPU-to-GPU and GPU-to-GPU bandwidth required for modern AI workflows. The system supports up to 2TB of DDR5 memory and multiple high-speed network options including InfiniBand and Ethernet, enabling seamless integration into HPC clusters and cloud data centers. Launched in Q3 2023, the XD670 addresses critical demand from cloud service providers, telecommunications companies, and research institutions developing and deploying next-generation large language models.

The XD670's compact, liquid-cooled design reduces data center footprint and power delivery requirements compared to air-cooled alternatives while supporting sustained peak GPU performance. Sophisticated power distribution, redundant cooling loops, and integrated systems management enable reliable operation in production AI training clusters where continuous uptime and thermal stability are paramount.

Ideal for

• Large language model training and fine-tuning for enterprise AI applications and cloud service providers
• Generative AI model development and inference serving for research institutions and AI companies
• Multi-GPU distributed deep learning training across clusters for computer vision and natural language processing tasks
• High-performance scientific simulation and molecular dynamics modeling requiring sustained GPU acceleration
• Inference acceleration for real-time conversational AI, recommendation systems, and content generation services
• Telecommunications network optimization and signal processing workloads requiring dense GPU parallelism

Technical specifications

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