HPE Cray XD670 digital data sheet.PDF
HPE cray supercomputing xd670.pdf
What’s New
- Support for eight NVIDIA H200 Tensor Core GPUs with 5th Generation Intel® Xeon® Scalable processors.
- Support for up to 32x DDR5 DIMMs, up to 5600 MT/s.
- PCIe Gen5 increased data transfer speed over PCIe Gen4, giving faster data transfer speeds for SSDs, GPUs, and other peripheral cards.1
Key Features
Purpose-built for AI Workloads
The HPE Cray XD670 system is powered by eight NVIDIA H200 or H100 Tensor Core GPUs and Transformer Engine support.
#1 result for Natural Language Processing (NLP) with BERT 99.0 Offline scenario in MLPerf ™ Inference v4.0 benchmark.2
Complete solution for AI that can scale out from one to thousands of nodes to support growing AI environments.
Efficient Flexibility in a Dense Architecture
The HPE Cray XD670 is a 5U chassis system that contains a single 2x CPU node.
Choice of air cooling or plug-and-play direct liquid cooling (DLC) that increases power efficiency and enables energy reuse, helping you advance your sustainability goals.
Purpose-built integrated storage delivers strong read and write performance cost-effectively.
Supports a wide range of interconnect options including InfiniBand, high-speed Ethernet, and HPE Slingshot.
Simplify Management and Deployment
The HPE Cray XD670 system includes industry-standard, DMTF-compliant Redfish for simple and secure system management.
HPE Performance Cluster Manager (HPCM) provides automated provisioning, monitoring, remote support, and firmware updates for HPE Cray XD670.
HPE Services brings together technology and expertise to accelerate your transformation and prepare for what comes next.
1 Visit URL https://www.intel.com/content/www/us/en/gaming/resources/what-is-pcie-4-and-why-does-it-matter.html
2 “New MLPerf Inference Benchmark Results Highlight The Rapid Growth of Generative AI Models,” MLCommons(tm), March 2024
Intel Xeon is a trademark of Intel Corporation or its subsidiaries in the U.S. and/or other countries. NVIDIA is a trademark and/or registered trademark of NVIDIA Corporation in the U.S. and other countries. MLCOMMONS™ and MLPERF™ are trademarks and service marks of MLCommons Association in the United States and other countries. All third-party marks are property of their respective owners.
-
Chassis
- 5U chassis system
- Single 2x CPU node
-
GPU acceleration – Option A
- 8x NVIDIA H200 SXM 700W TDP GPUs with 141 GB HBM each
- 2x 5th Gen Intel Xeon Scalable processors CPUs, up to 400W TDP
- 32x DIMMs DDR5, 8-channel memory per socket, up to 5600 MHz
-
GPU acceleration – Option B
- 8x NVIDIA H100 SXM 700W TDP GPUs with 80 GB HBM each
- 2x 4th Gen Intel Xeon Scalable processors CPUs, up to 300W TDP
- 32x DIMMs DDR5, 8-channel memory per socket, up to 4800 MHz
-
Power and cooling
- Air cooled and also available with a Direct Liquid Cooling option for improved power efficiency
Reviews
There are no reviews yet.