Skip to content
HomeSolutionsGPU ServersHPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled
Nexus ComputeNew

HPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled

Liquid-cooled 5U HGX H200 node engineered for dense, efficient training rows.

Full manufacturer warrantyAuthorized channel48-hour quote

We help you choose, configure, and deliver the right system — no obligation.

HPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled — Nexus Compute enterprise hardware
HPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled hardware detail 1
HPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled hardware detail 2
HPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled hardware detail 3

Configuration at a Glance

GPU / Accelerator8× NVIDIA H200 SXM5, 141GB HBM3e each
GPU InterconnectNVLink fully connected (HGX H200 baseboard)
CPU2× 5th Gen Intel Xeon Scalable, up to 64 cores each
MemoryUp to 32× DDR5-5600 DIMMs

Tailored per engagement. Full technical overview below.

Configuration Options

Core specifications for this system. Every component is configurable to your workload — request a quote for a tailored build.

GPU / Accelerator

8× NVIDIA H200 SXM5, 141GB HBM3e each

Processor

2× 5th Gen Intel Xeon Scalable, up to 64 cores each

Memory

Up to 32× DDR5-5600 DIMMs

Storage

Front-access NVMe bays, configurable capacity

Overview

The HPE Cray XD670 condenses eight NVLink-connected NVIDIA H200 SXM GPUs into a 5U single-node chassis with a direct liquid cooling option for higher rack density and power efficiency. Nexus Compute specifies, configures, and tests each system to your Slingshot, NDR InfiniBand, or Ethernet fabric, delivered warranty-backed through authorized HPE channels.

Who This Solution Is For

Operators maximizing GPU density per rack with liquid cooling
HPC and AI centers running HPE Slingshot fabrics
Teams pursuing lower PUE and power-efficient training
Organizations standardizing on HPE Cray supercomputing

Business Benefits

Higher density per rack

The compact 5U single-node design with direct liquid cooling packs more H200 compute into each rack than air-cooled equivalents.

Improved power efficiency

Direct liquid cooling lowers thermal overhead and PUE, reducing operating cost at scale.

Supercomputing fabric

Native Slingshot 11, NDR InfiniBand, and Ethernet options connect nodes into large, balanced training systems.

Typical Business Use Cases

1

Large-model AI training and deep learning

2

High-density liquid-cooled GPU rows

3

HPC simulation on Slingshot fabrics

4

Power- and space-constrained data center buildouts

Industry Applications

HPCAI & Machine LearningHigher Education & ResearchGovernment & DefenseManufacturing

Technical Overview

A 5U single-node chassis built on the NVIDIA HGX H200 baseboard with NVLink, driven by dual 5th Gen Intel Xeon Scalable CPUs across 32 DDR5-5600 DIMM slots and a direct liquid cooling loop. Eight PCIe Gen5 slots provide switchable connections to Slingshot 11, NDR InfiniBand, or Ethernet fabrics.

GPU / Accelerator8× NVIDIA H200 SXM5, 141GB HBM3e each
GPU InterconnectNVLink fully connected (HGX H200 baseboard)
CPU2× 5th Gen Intel Xeon Scalable, up to 64 cores each
MemoryUp to 32× DDR5-5600 DIMMs
StorageFront-access NVMe bays, configurable capacity
Networking / Fabric8× PCIe Gen5 slots — Slingshot 11 / NDR InfiniBand / Ethernet
Form Factor5U single node, direct liquid cooled
ManagementHPE iLO with HPE GreenLake options
WarrantyHPE support, configurable term

Specifications are indicative and configured to each engagement. Request a quote for a configuration tailored to your requirements.

Warranty, Support & Fulfillment

Every system ships from an authorized channel, configured and tested, with the documentation enterprise buyers need — backed by warranty and a dedicated account team.

Enterprise Warranty

Full manufacturer warranty with optional on-site, next-business-day support and extended coverage.

Authorized Channel

Sourced through Tier-1 distribution and OEM partners — never grey market. Asset & warranty records included.

Lead Time & Deployment

48-hour quotes, then configured, burn-in tested, and delivered on a committed schedule.

Nationwide Fulfillment

Coordinated logistics, rack-and-stack, and delivery wherever your infrastructure lives.

Frequently Asked Questions

What does my facility need for direct liquid cooling?

Direct liquid cooling requires facility coolant distribution (CDU and secondary loop) and appropriate plumbing. We confirm cooling, power, and rack requirements during the quote and advise where air-cooled alternatives may fit better.

How does liquid cooling change density and cost?

Liquid cooling lets the 5U node sustain full GPU power in tighter rows and lowers thermal overhead, improving PUE and operating cost at scale. The trade-off is the upfront cooling infrastructure, which we help you plan.

Is Slingshot required, or can I use InfiniBand?

Slingshot 11 is native to HPE Cray systems, but the XD670 also supports NDR InfiniBand and Ethernet. We specify the fabric that matches your existing estate and scaling plans.

Hardware Assistance

Configure the HPE Cray XD670 — 8x H200 SXM, Direct Liquid Cooled with Nexus Compute

Tell us your requirements and a hardware specialist will help you specify, configure, and quote the right system — typically within two business days. No obligation.