NHR@SW
Universität Frankfurt am Main, Rheinland-Pfälzische Technische Universität, Kaiserslautern-Landau (RPTU), Universität Mainz, Universität des Saarlandes
Mainz, Frankfurt, Saarbrücken, Kaiserslautern-Landau

NHR South-West (NHR@SW) is a collaboration of Johannes Gutenberg University Mainz, Goethe University Frankfurt, RPTU Kaiserslautern-Landau, and Saarland University. As member of the NHR Alliance, NHR@SW's mission is to accelerate scientific discovery by combining cutting-edge high-performance computing with methodological innovation, expert support, and strong partnerships across disciplines. The HPC systems are operated at Mainz and Frankfurt.
Key Expertise and Profile
- Simulation and Data Labs for life sciences, molecular systems, and nuclear, particle, and astrophysics
- Method Labs for performance and programming, including optimization, parallel I/O, GPU computing, and parallel programming
- AI and quantum computing methods, including machine learning and hybrid quantum-HPC approaches
- Operating the NHR Computational Physics Center together with PC2 and NHR@KIT
- NHR@SW's infrastructure provides tens of thousands of CPU cores, GPU resources, multi-petabyte high-performance storage and prototype quantum computing access, enabling data-intensive and compute-heavy workloads.
Technical Specifications
| Center Name | NHR@SW |
|---|---|
| Total Resources: | |
| # CPU-Cores | 95 040 |
| # GPUs | 936 |
| Additional Information | |
| max. job run time [h] | 504 |
| jupyter notebook available [y/n] | y |
| CPU Resources: | |
| System 1 | MOGON NHR |
| CPU-Type | AMD EPYC 7713 |
| # nodes | 590 |
| # Cores per node | 128 |
| memory per node [GB] | 256 - 2048 |
| System 2 | Goethe NHR |
| CPU-Type | Xeon Skylake Gold 6148 |
| # nodes | 488 |
| # Cores per node | 40 |
| memory per node [GB] | 192 |
| GPU Resources | |
| System 1 | MOGON NHR |
| GPUs per node | 4 x NVIDIA A100 |
| # nodes | 10 |
| memory per GPU [GB] | 40 |
| System 2 | Goethe NHR |
| GPUs per node | 8 x AMD Instinct MI210 |
| # nodes | 112 |
| memory per GPU [GB] | 64 |