The European partnership EuroHPC for high-performance supercomputers has now officially been launched. Over the next seven years, 7 billion euros are to flow into the project. Germany is also applying as a location and wants to put Europe's first exascale supercomputer into operation by 2024.
During the Council of the EU on Tuesday a regulation establishing the initiative EuroHPC (European High Performance Computing), the German Federal Ministry of Education and Research (BMBF) announced its plans for this undertaking (PDF).
Germany is applying for the exascale system
Accordingly, on behalf of Germany, the Gauss Center for Supercomputing (GCS) will apply for one of the first exascale supercomputers. The goal is “that by 2024 the first European supercomputer of the exascale class can be put into operation in Germany,” explains Wolf-Dieter Lukas, State Secretary in the Federal Ministry of Research. The application process should start this year.
The GCS is an association for the merger of the Jülich Supercomputing Center (JSC) in Jülich, the Leibniz Computing Center (LRZ) in Garching near Munich and the High Performance Computing Center Stuttgart (HLRS) to promote scientific high-performance computing. The facilities are already providing the fastest supercomputers in Germany. Led by JUWELS (JSC), which is currently the fastest supercomputer in Europe with 44.12 PetaFLOPS (71 PFLOPS Peak) and is currently ranked 8th in the world ranking “Top500”, the SuperMUC-NG (LRZ) and Hawk (HLRS) systems also belong ).
Exascale stands for computing power at least in the range of 1 ExaFLOPS, which corresponds to 1,000 PetaFLOPS. If Germany's application is successful, the first supercomputer in the exascale class in Europe is to be built by 2024 with EU funding. The goal is ambitious, because the reigning leader Fugaku from Japan “only” has around 442 PetaFLOPS (537 PFLOPS Peak), but is about ten times as fast as Germany's previous leader.
The US is targeting the mark much earlier. Two systems are planned for 2021 with Aurora (~ 1 ExaFLOPS) and Frontier (~ 1.5 ExaFLOPS). In 2023 El Capitan will then increase to 2 ExaFLOPS.
As part of the EuroHPC initiative, for example, the Leonardo (Italy) and Lumi (Finland) systems are also being created, which at least should achieve several hundred PetaFLOPS.