PLGrid Documentation

Lumi

LUMI

LUMI (Large Unified Modern Infrastructure) is a pre-exascale supercomputer located in Kajaani, Finland, managed by CSC – IT Center for Science and funded by the EuroHPC initiative. Serves as a key research platform for European scientists, supporting scientific computing, climate simulations, work on artificial intelligence, and digital twin models.

The LUMI architecture is based on the HPE Cray EX system and consists of GPU and CPU partitions. Its actual performance is ~379 petaflops, and its theoretical peak power is over 530 petaflops. The total memory and disk infrastructure capacity reaches ~117 PB, with I/O bandwidth reaching ~2 TB/s. LUMI regularly ranks very high in various ratings: in June 2025, it was ranked 9th in the TOP500 and 36th in the Green500 ( assessing the energy efficiency of supercomputers). It also stands out in terms of environmental concerns, running on 100% renewable energy, it is liquid-cooled, and the heat generated during computation is used to heat the neighboring residential infrastructure.

Access to LUMI for Polish scientists is provided via the PLGrid infrastructure, operated by the Cyfronet AGH Academic Computer Center, which plays the role of the EuroHPC National Competence Center in Poland. CPU and GPU resources on LUMI can be obtained through test grants (continuous call for applications), and competitions for the regular computing grants, announced twice a year on plgrid.pl and Cyfronet.pl websites. Please note! Only applicants affiliated with a scientific institution on the PLGrid Portal may apply for a test or regular LUMI grant.

Performance

  • Manufacturer: HPE
  • Cores: 2 752 704
  • Processor: AMD Optimized 3rd Generation EPYC 64C 2GHz
  • Interconnect: Slingshot-11
  • Linpack Performance (Rmax): 379.70 PFlop/s
  • Theoretical Peak (Rpeak): 531.51 PFlop/s

ln_lumi_5_cabinets

Other clusters

Athena (ACK Cyfronet AGH)

Ares (ACC Cyfronet AGH)

Helios (ACC Cyfronet AGH)

Lem (WCSS)

Tryton Plus (TASK)


Last update: October 22, 2025