HPE Cray XD670 GPU-Optimized Server

The HPE Cray XD670 is an 8-GPU dense server architecture engineered specifically for large-scale AI and machine learning training operations. Each system accommodates eight NVIDIA H100 SXM5 GPUs with full NVLink support, enabling high-bandwidth GPU-to-GPU communication essential for distributed training across multiple nodes. The XD670 integrates dual-socket AMD EPYC processors, high-speed interconnect options, and optimized thermal management to sustain peak compute performance in production data center environments.

Designed for enterprises deploying transformer models, deep learning frameworks, and scientific computing workloads, the XD670 delivers approximately 5.3 petaflops of theoretical FP8 compute per system. The server supports seamless integration into existing HPE and third-party cluster management ecosystems, with flexible networking configurations including HDR InfiniBand and Ethernet for both storage and compute plane connectivity.

Launched in Q2 2023, the XD670 represents HPE Cray's commitment to purpose-built AI infrastructure, addressing the computational intensity requirements of modern generative AI model training and fine-tuning operations at enterprise scale.

Ideal for

• Training large language models (LLMs) and foundational transformer architectures at scale
• Distributed deep learning across multiple XD670 nodes for computer vision and NLP applications
• Scientific computing and computational fluid dynamics simulations requiring GPU acceleration
• Generative AI model development and experimentation with reduced time-to-insight
• Multi-tenant AI model training platforms and MLOps infrastructure
• Financial modeling and risk simulation leveraging GPU-accelerated analytics

Technical specifications

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