The HPE Cray XD670 AI Server is a liquid-cooled GPU-accelerated platform built for large-scale AI and HPC workloads. Designed for dense deployments, it supports multiple NVIDIA GPUs and high-speed networking to enable efficient training, simulation, and data-intensive processing.
£65,000.00 Ex. VAT
Available with a Lead Time: Submit an enquiry for accurate lead time information.
Advanced AI and machine learning workloads require infrastructure capable of supporting extremely large GPU clusters and high-throughput data processing.
The HPE Cray XD670 AI Server supports high-density GPU configurations designed for large-scale AI training environments. The platform integrates modern NVIDIA GPU accelerators alongside high-performance server processors and high-speed interconnects. Support for modern memory, storage, and I/O keeps application, database, and virtualisation workloads responsive as utilisation rises and data movement increases across the platform.
The XD670 architecture supports PCIe Gen5 connectivity, high-capacity system memory, and fast network fabrics to enable efficient scaling across multi-node AI clusters. High-bandwidth PCIe lanes, accelerator support, and dense compute packaging allow large models and analytics pipelines to move data quickly between CPUs, GPUs, and storage. Organisations that do not require maximum GPU density in a single AI server may also consider the HPE Cray XD6500 AI Server, for supporting accelerator workloads with a lower-density platform step.
By combining GPU acceleration with scalable system architecture, the XD670 enables organisations to build AI clusters capable of supporting demanding machine learning and analytics workloads. GPU-ready design helps teams train, fine-tune, or serve AI workloads faster while keeping deployment under the same enterprise management and support model as the wider infrastructure estate.
Our specialists can design and deploy HPE Cray XD670 AI Server infrastructure aligned to enterprise AI and high-performance computing environments.
| Product Family | HPE Cray |
|---|---|
| Product Series | XD670 |
| Platform Generation | |
| Deployment | HPC / AI |
| Primary Use Case | AI / GPU Compute |
| Product Tier | Specialist |
| Chassis Type | |
| Form Factor | |
| CPU Vendor | |
| Processor Platform | |
| CPU Architecture Class | |
| Maximum CPUs Supported | |
| Memory Type | |
| Maximum Memory Capacity | |
| Memory Slots | |
| Maximum Memory Speed (MT/s) | |
| Drive Bay Configuration | |
| Maximum Storage Capacity | |
| NVMe Drive Support | |
| Maximum GPUs Supported | |
| PCI Express Slots | |
| PCIe Generation | |
| Embedded Network Ports | |
| Maximum Network Speed | |
| OCP / Network Expansion | |
| Maximum Power Supply (W) | |
| Power Redundancy | Hot-swappable PSUs |
| Remote Management |
The HPE Cray XD670 AI Server is designed for organisations that need a purpose-built GPU server for AI training, deep learning, and rack-scale accelerated compute. It is best suited to environments running large language models, multimodal training, and other GPU-intensive AI workloads at scale.
Teams building or tuning large language models need dense GPU infrastructure with strong interconnect and memory capability. The Cray XD670 fits these deployments where AI training scale is a central requirement.
AI projects that combine text, image, and other data types place heavy demands on compute and GPU throughput. The Cray XD670 is aimed at these environments where multimodal training and inference need purpose-built accelerated infrastructure.
Deep learning teams often need a platform that can move from training into tuning and optimisation without changing server architecture. The Cray XD670 works well where iterative AI development depends on dense GPU resources.
Many organisations want AI infrastructure that scales from a small deployment into larger clustered environments. The Cray XD670 is a strong fit for enterprise AI estates that need consistent rack-scale accelerated compute.
Some data centres need dense GPU compute while also paying close attention to power and cooling efficiency. The Cray XD670 suits these environments through optional direct liquid cooling and integrated rack-scale design.
Our team can help design and deploy HPE Cray XD670 infrastructure for LLM training, deep learning, and enterprise AI platforms at scale.
As an official HPE partner, we help enterprises unlock flexible financing options that align IT investment with business goals.
Through HPE Financial Services, we can help your business to:
Unlock smart HPE financing, built for enterprise scale.
Experience smart savings and enhanced sustainability with HPE Renew.
This circular initiative provides top-quality remanufactured HPE hardware, fully-certified and backed by warranty.
Industry-leading quality meets unbeatable value. Upgrade your infrastructure while minimising spend.
As a certified partner, we can unlock exclusive HPE promotions for your business, providing enhanced resilience and efficiency across your network at unbeatable discounts.
Below you’ll find alternative models, suitable software and services that pair with this solution – helping you to avoid compatibility issues, reduce support overhead and deploy with confidence.
Not all deployments fit standard configurations. If you’re weighing up options or want a second opinion on your setup, our team is here to help with honest, straightforward advice backed by decades of vendor knowledge.
Not sure which model is right for your environment? Our specialists can help you select the right platform for your infrastructure requirements.
