Loading Events

« All Events

Toward energy-efficient exascale CFD: Lessons learnt from the CEEC project

December 8 @ 2:00 pm – 3:00 pm
Webinar announcement poster for CEEC titled “Toward energy-efficient exascale CFD: Lessons learnt from the CEEC project.” Top left reads “Webinar” with a small laptop icon, and top right has the CEEC logo. A dark blue box on the left gives details: “December 8, 2026, 14:00–15:00 CET, Online.” Below it: “Discover … how CEEC analysed the energy use of CFD simulations on European supercomputers. Learn about … the tools, metrics and strategies that can help make future HPC and exascale simulations more energy-efficient!” On the right, two side‑by‑side bar charts compare time‑to‑solution and energy‑to‑solution for three systems (MN5‑GPP Intel Sapphire Rapids, MN5‑ACC NVIDIA H100, and LUMI‑G AMD MI250X), showing lower bars for the accelerator and LUMI systems, indicating better performance and energy efficiency. Curved blue lines decorate the bottom of the poster.

Energy efficiency has become a major challenge in modern high-performance computing (HPC), as growing computational demands and increasingly complex architectures lead to substantial energy consumption. This webinar presents the collective experience of the EuroHPC JU Centre of Excellence in Exascale CFD (CEEC) in measuring, analyzing, and optimizing the energy consumption of computational fluid dynamics (CFD) simulations across major European HPC systems. We first introduce methodologies and tools for energy measurement and discuss key metrics for evaluating and reporting energy efficiency. Through case studies involving representative CFD applications (waLBerla, FLEXI/GALÆXI, Alya, Neko, and NekRS), we examine energy-to-solution and time-to-solution across diverse CPU- and GPU-based architectures, including LUMI, MareNostrum 5, MeluXina, and JUWELS Booster. Our findings demonstrate the potential of hardware accelerators and mixed-precision computing to reduce energy consumption while preserving numerical accuracy. Finally, we discuss the importance of making energy measurement more accessible and systematic on HPC systems, both to raise awareness within the scientific computing community and to support informed decisions toward more energy-efficient exascale computing and CFD simulations.

Details

Venue

  • Online