HPE Cray EX4000 (Hunter)

Image f. HPE Cray EX4000 (Hunter)

Hunter is HLRS’s newest flagship supercomputer. Based on the HPE Cray EX4000 platform from Hewlett Packard Enterprise, it is a world-class high-performance computing system that enables large-scale, state-of-the-art applications of simulation, artificial intelligence, and data analytics, including hybrid computing approaches that combine diverse methods into powerful workflows.

Hunter marks a milestone for HLRS, as it moves away from the center’s previous CPU-oriented approach to make greater use of GPUs. At its core is the AMD Instinct MI300A accelerated processing unit (APU), which combines CPUs, GPUs, and high-bandwidth memory in a single package.

The technology offers fast data transfer speeds, impressive performance, easy programmability, and great energy efficiency. Hunter is 50 % faster than its predecessor, Hawk, while slashing energy requirements at peak performance by approximately 80 %.

Hunter will go into production in early 2025.

HLRS conceived Hunter as a transitional system that will enable system users to prepare for its upcoming exascale supercomputer, Herder. Hewlett Packard Enterprise will also manufacture that future next-generation system, whose installation is scheduled to begin in 2027.

Funding

Funding for the Hunter supercomputer is provided by the German Federal Ministry of Education and Research and the Baden-Württemberg Ministry for Science, Research and Arts through the Gauss Centre for Supercomputing.

 Logo MWKLogo BMBF 

Technical details

Platform

HPE Cray EX4000

Performance benchmarks

HPL performance APU: 27 PFlop/s (FP64)
HPCG performance APU: 350 TFlop/s (FP64)

Processor types

APU node AMD Instinct MI300A
CPU node AMD EPYC 9374F

Number of processors

APUs 752 MI300A
CPUs 512 AMD EPYC 9004

Number of nodes

APU 188
CPU 256

Memory technology

APU 512 GB HBM3 (~5.3 TB/s)
CPU 768 GB DDR5-4800

Storage

Cray ClusterStor E2000 25 PB capacity on 2120 disks

Networking

HPE Slingshot 11 Dragonfly (APU: 4 x 200 Gbps per node)

Energy usage (average)

560 kW  

Start of operation

Early 2025  

Related news

All news

HLRS Wins Datacenter Strategy Award in the Category "Transformation"

HPC in Transition: Hunter and Herder Will Bring New Opportunities, New Challenges

Exascale Supercomputing Is Coming to Stuttgart