This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| SuperMUC-NG | |
|---|---|
| Name | SuperMUC-NG |
| Location | Leibniz Supercomputing Centre, Garching, Germany |
| Manufacturer | Lenovo |
| Operators | Bavarian Academy of Sciences and Humanities |
| Architecture | Lenovo Neptune, Intel Xeon Scalable |
| Peak performance | 26.9 PFLOPS (theoretical) |
| Memory | 6 PB (file system) |
| Storage | 15 PB Lustre |
| Power | 3.9 MW |
| Network | Omni-Path |
| Deployment | 2018 |
SuperMUC-NG SuperMUC-NG is a petascale cluster housed at the Leibniz Supercomputing Centre in Garching, Germany, providing high-performance computing resources to European researchers. The system supports large-scale simulations, data analysis, and machine learning workloads across scientific domains and collaborates with institutions such as the Max Planck Society and Ludwig Maximilian University of Munich. It integrates cutting-edge processors, high-bandwidth interconnects, and advanced cooling to achieve multi-petaflop performance for projects spanning physics, climate science, and bioinformatics.
SuperMUC-NG operates as a national research infrastructure node connected to European initiatives like PRACE and hosts projects funded by the Deutsche Forschungsgemeinschaft and the European Commission. The system serves communities including researchers from the Max Planck Institute, Technical University of Munich, Helmholtz Association, and the European Space Agency, facilitating workloads from computational fluid dynamics to genomics. Resource allocation and scheduling leverage policies from institutions such as the Bavarian Academy of Sciences and Humanities and collaborations with industrial partners like Lenovo and Intel.
The system is based on a Lenovo-built architecture integrating Intel Xeon Scalable processors across thousands of compute cores and employs Lenovo Neptune liquid cooling technology. The cluster comprises hundreds of compute nodes interconnected via Intel Omni-Path fabric and features a Lustre parallel file system provided by vendors used by centers like the Swiss National Supercomputing Centre and the Barcelona Supercomputing Center. Storage and memory resources are provisioned to support workflows similar to those run on systems at CERN, NASA, and the European Centre for Medium-Range Weather Forecasts. Hardware vendors and technology partners include Intel, NVIDIA, Lenovo, Mellanox (now part of NVIDIA), and Cray (part of HPE).
SuperMUC-NG achieved multi-petaflop performance measured by benchmarks used in the field such as HPL and HPCG, comparable to installations like Summit, Sierra, and MareNostrum. Performance tuning draws on methodologies developed at institutions like Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, and the Jülich Supercomputing Centre, optimizing MPI implementations from OpenMPI and Intel MPI. Benchmarking and application scaling involve communities working on codes such as GROMACS, VASP, WRF, and OpenFOAM, with optimizations for Intel architectures guided by teams from the University of Edinburgh, ETH Zurich, and the Barcelona Supercomputing Center.
The software stack includes Linux distributions used at centers like the National Center for Supercomputing Applications and programming models such as MPI, OpenMP, and CUDA for GPU-accelerated systems, alongside compilers from Intel and GNU toolchains. Scientific packages and frameworks commonly deployed mirror those at Lawrence Berkeley National Laboratory and Argonne National Laboratory, including PETSc, Trilinos, TensorFlow, and PyTorch for machine learning workloads. Workflow and job management utilize schedulers and resource managers like SLURM and PBS Pro, with performance analysis aided by tools from Arm Forge, Intel VTune, and Score-P developed in conjunction with partners like the High Performance Computing Center Stuttgart and Forschungszentrum Jülich.
The installation employs liquid cooling solutions inspired by designs used at the Piz Daint and SuperMUC installations, enabling high power density operation while maintaining energy efficiency metrics considered by the Green500 list. Cooling engineering draws on experience from the Leibniz Supercomputing Centre and collaborations with companies such as Asetek and CoolIT, and uses measurement practices similar to those at the National Renewable Energy Laboratory when reporting Power Usage Effectiveness. Energy-aware scheduling and dynamic power management techniques are informed by research from the European Centre for Medium-Range Weather Forecasts and the Barcelona Supercomputing Center.
SuperMUC-NG supports diverse scientific applications, including climate modeling by groups at the Max Planck Institute for Meteorology, astrophysics simulations by researchers affiliated with the Max Planck Institute for Astrophysics, materials science calculations by teams at the Fritz Haber Institute, and bioinformatics pipelines used by the European Molecular Biology Laboratory. Industrial collaborations involve automotive simulations with partners such as BMW and Volkswagen, and energy sector modeling in cooperation with Siemens and E.ON. The platform contributes to international projects coordinated by PRACE, EuroHPC, and the European Space Agency, enabling cross-institutional research with partners including CERN, ESA, and EMBL.
Development of the system built on experience from predecessor systems at the Leibniz Supercomputing Centre, drawing lessons from the original SuperMUC and other European centers like the Jülich Supercomputing Centre and the Barcelona Supercomputing Center. Procurement and commissioning involved consortia of suppliers, research institutions, and funding bodies such as the German Federal Ministry of Education and Research and the Bavarian state government. Ongoing upgrades and operational support are coordinated with academic partners including Ludwig Maximilian University of Munich, Technical University of Munich, and the Bavarian Academy of Sciences, reflecting a continuous evolution aligned with European HPC roadmaps and initiatives such as EuroHPC and PRACE.