HPCW 3.0 Released: Open-Source Benchmarking Suite for Climate and Weather Simulations
The High-Performance Climate and Weather (HPCW) benchmarking suite has officially gone open-source with the release of version 3.0.
HPCW is designed to make it easier to benchmark new supercomputing systems, focusing on applications related to weather and climate research. The suite automates much of the complex setup required to build and run large-scale scientific models, helping users test performance with minimal manual intervention.
For over 30 years, the global Top500 list has ranked supercomputers using the LINPACK benchmark — a test of pure mathematical performance. However, as today’s systems and applications become more diverse, a single benchmark no longer tells the whole story. Complementing benchmarks like Green500 (energy efficiency) and Graph500 (data-intensive workloads), HPCW brings a new domain-specific approach, assessing performance on realistic weather and climate workloads.
The HPCW suite includes both full models and model components, such as:
- ICON – the German weather and climate model framework
- NEMO – the Nucleus for European Modelling of the Ocean
- ecTrans – a key numerical library used by ECMWF and Météo-France
Version 3.0 introduces support for the Japanese NICAM model (Nonhydrostatic Icosahedral Atmospheric Model), aligning with the goals of the HANAMI project. HPCW now includes the open-source dynamical core of NICAM and infrastructure for an idealised test case.
HPCW currently runs on major European and Japanese HPC systems — including DKRZ/Levante, ECMWF/HPC2020, R-CCS/Fugaku, and BSC/MareNostrum5 — and the developers welcome contributions and extensions to new platforms.
Explore HPCW on GitLab: https://gitlab.dkrz.de/hpcw/hpcw
Documentation: https://hpcw.gitlab-pages.dkrz.de/hpcw/
