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.
| LUMI | |
|---|---|
| Name | LUMI |
| Country | Finland |
| Location | Kajaani |
| Operator | EuroHPC Joint Undertaking |
| Architecture | HPE Cray EX |
| Cpu | AMD EPYC |
| Gpu | AMD Instinct |
| Memory | 2.5 PB (aggregate) |
| Storage | 10 PB (parallel) |
| Peak | 550 PFLOPS (mixed precision) |
| Year | 2020s |
LUMI
LUMI is a high-performance supercomputer installed in Kajaani, Finland, operated within a European consortium and deployed to support scientific research, engineering, and industrial workloads. It is part of a network of European supercomputers funded and coordinated by the EuroHPC Joint Undertaking, hosting advanced hardware from HPE Cray, AMD, and European partners to deliver exascale-class capabilities for climate modeling, materials science, life sciences, and artificial intelligence. The system emphasizes energy-efficient design, pan-European access policies, and collaboration among national research organizations, national laboratories, and private sector partners.
LUMI provides compute and data services to researchers from institutions such as the European Commission, Academy of Finland, CERN, EMBL, and national research centres across France, Germany, Italy, Spain, Sweden, Norway, and Poland. The project aligns with strategic initiatives like the Digital Europe Programme, Horizon Europe, and the European High Performance Computing Strategy to strengthen capacity in computational science, machine learning, and large-scale simulation. Managed under frameworks similar to those used by PRACE, EuroHPC JU, and national supercomputing centres, LUMI integrates user support, training, and data stewardship services provided by partners including the CSC — IT Center for Science and regional universities.
Planning for LUMI was coordinated among stakeholders such as the Finnish Government, the European Investment Bank, and suppliers including Hewlett Packard Enterprise and AMD. Procurement and deployment drew on precedents from systems at Jülich Research Centre, the Barcelona Supercomputing Center, and the Pascal/Top500 community. Installation in Kajaani involved collaboration with municipal authorities like Kainuu Region and industrial partners including Nokia and energy companies connected to the Nordic electricity market. The timeline intersected with projects funded through European Regional Development Fund initiatives and partnerships with national research infrastructures such as MET Norway and the National Institute for Health and Welfare (Finland).
The system architecture was designed by HPE Cray using an EX architecture chassis populated with AMD EPYC CPUs and AMD Instinct GPUs. The interconnect employs technologies with lineage from designs used at Oak Ridge National Laboratory and Lawrence Livermore National Laboratory, while the parallel file system and storage architecture draw on solutions similar to those adopted at NERSC and CSCS. Cooling and site engineering referenced standards and vendors that have serviced installations at CEA, Argonne National Laboratory, and Rutherford Appleton Laboratory. Software stack choices mirror ecosystems used by OpenMPI, Slurm Workload Manager, Singularity, and ROCm-based toolchains, enabling portability for codebases originating from labs such as Max Planck Society, CNRS, ETH Zurich, and Imperial College London.
LUMI's reported peak mixed-precision performance reaches hundreds of petaflops, comparable to top-tier systems at Oak Ridge National Laboratory's Summit and Argonne's Aurora roadmap, while double-precision capability supports workloads typical of facilities like Jülich's JUWELS. Standard benchmarks and community codes—such as those developed by teams at University of Cambridge, Massachusetts Institute of Technology, Stanford University, and Princeton University—have been used to validate sustained performance on molecular dynamics, computational fluid dynamics, and deep learning workloads. Performance tuning and scaling studies have involved collaborations with software teams from NVIDIA ecosystem projects and open-source initiatives at The Linux Foundation and OpenHPC.
LUMI serves multidisciplinary applications including climate and weather modeling led by groups at ECMWF and Met Office, materials discovery programs associated with Max Planck Institute for Chemical Physics of Solids and Lawrence Berkeley National Laboratory, genomics and bioinformatics pipelines used by Wellcome Sanger Institute and EMBL-EBI, and AI research from consortia involving DeepMind-adjacent academic partners. Industrial users include energy and automotive research from firms and labs linked to Siemens, Volvo, Airbus, and pharmaceutical modeling inspired by groups at Roche and GlaxoSmithKline. Use cases also encompass cosmology simulations undertaken by teams at Space Telescope Science Institute and CEA Saclay, and particle physics analysis workflows Cooperating with ATLAS and ALICE collaborations.
LUMI emphasizes energy-efficient design, utilizing advanced liquid cooling and heat-reuse options inspired by implementations at FRACTAL projects and systems at PDC Center for High Performance Computing. The site leverages Finland's renewable electricity mix and grid connections consistent with practices at Statnett and the Nord Pool electricity market to minimize carbon footprint. Sustainability measures draw on methodologies promoted by European Environment Agency and align with targets emerging from European Green Deal and corporate sustainability strategies of suppliers like Hewlett Packard Enterprise and AMD.
Governance is structured through partnerships among the EuroHPC Joint Undertaking, national ministries such as the Ministry of Education and Culture (Finland), research organizations including CSC — IT Center for Science, and industrial vendors. Funding combined contributions from European Union instruments, national co-investments from consortium countries, and procurement contracts with HPE and AMD, following procurement precedents set by projects funded through European Structural and Investment Funds and co-funding models used by PRACE and ESFRI infrastructures. Access policies and allocation committees involve representatives from member states, academic stakeholders, and research infrastructures like EATRIS and ELIXIR.