The Stuttgart Center for High Performance Computing (HLRS) has taken a significant step towards exascale computing with the introduction of its new flagship supercomputer, Hunter. This advancement marks an important milestone in the evolution of high-performance computing capabilities in Germany and Europe.
Founded in 1996, the HLRS has established itself as the first national center for high-performance computing in Germany. Affiliated with the University of Stuttgart, the HLRS is also a founding member of the Gauss Center for Supercomputing, an alliance of Germany’s leading supercomputing centers. Over the years, the HLRS has been a pillar in the research and development of advanced computing technologies, providing crucial resources for a wide range of scientific and engineering applications.
Hunter: The New Jewel in the Crown
Hunter, based on the HPE Cray EX4000 platform from Hewlett Packard Enterprise, represents the latest addition to the HLRS infrastructure. This high-performance computing system is designed to support large-scale data analytics, artificial intelligence, and simulation applications. One of the most notable features of Hunter is its ability to integrate hybrid computing approaches, combining diverse methods into powerful and efficient workflows.
At the heart of Hunter is the AMD Instinct™ MI300A accelerated processing unit (APU), which combines high-speed CPUs, GPUs and memory in a single package. This technology not only offers fast data transfer speeds and impressive performance, but is also highly efficient in terms of power consumption. Compared to its predecessor, Hawk, Hunter is 50% faster and reduces power requirements by approximately 80% at peak performance.
Preparing for the Future
Hunter is not only a technological advancement, but also a transitional platform that prepares HLRS users for the next generation of supercomputers. The HLRS is already planning the installation of its future exascale system, Herder, which will also be manufactured by Hewlett Packard Enterprise and whose installation is scheduled to begin in 2027.
Simulation using supercomputers is an essential tool for advanced research in numerous fields of science and engineering. Hunter users address complex problems in areas such as computational fluid dynamics, climate and weather modeling, biomedical research, and materials science. Additionally, Hunter is also used by industry, including startups and small and medium-sized businesses, as well as public sector agencies, providing secure access to powerful resources for high-performance computing and artificial intelligence.
Technical Details
Hunter is based on the HPE Cray EX4000 platform and offers performance of 27 PFlop/s (FP64) on its APU and 350 TFlop/s (FP64) on its HPCG. The system has 752 AMD Instinct MI300A APUs and 512 AMD EPYC 9374F CPUs, distributed across 188 APU nodes and 256 CPU nodes. Memory technology includes 512 GB of HBM3 for the APUs and 768 GB of DDR5-4800 for the CPUs. In terms of storage, Hunter has a capacity of 25 PB on 2120 disks, and uses the HPE Slingshot 11 Dragonfly network with 4×200 Gbps per APU node. The average energy consumption of the system is 560 kW.
With operations scheduled to begin in early 2025, Hunter is poised to drive research and innovation across multiple disciplines. The combination of its superior performance, energy efficiency and ability to handle complex applications makes it an invaluable tool for scientists and engineers. As HLRS moves into the era of exascale computing, Hunter represents a crucial step on this exciting journey, positioning Stuttgart and Germany at the forefront of high-performance computing globally.
