Supercomputers/Systems/LUMI
Finland · upgraded · mixed
LUMI
Quantum: linked to VTT Q50 (50 superconducting qubits), Helmi (5 superconducting qubits), Aalto Q20 (20 superconducting qubits)
Operated by CSC at LUMI Data Center, Kajaani .
Researched with AI assistance, checked by scripts. These rows were researched by AI agents reading public operator, laboratory and vendor sources, appended by scripts and checked by automated validation; no subject-matter expert has reviewed them (how this site is made). A mechanical second-reader pass has since fetched every cited source: 1350 of 1581 systems have a readable citation that names them, and 64 are genuinely weakly sourced. Every claim carries its source and a confidence tier. Treat anything below verified as a lead, not a citation.
Measured performance
- Rmax
- 379.7 PFlop/s
- Rpeak
- 531.5 PFlop/s
- Rmax ÷ Rpeak
- 71.4%
- Cores
- 2,752,704
- Power
- 7.11 MW
- Per watt
- 53.4 GF/W
Figures are as last publicly reported for the configuration described below, not a live measurement. Where a system was upgraded in place, the post-upgrade configuration is shown and the earlier one appears in the timeline.
How it compares
Ranked against the other machines in this dataset that report the same figure: those that first came into service within two years of LUMI, and those in Finland. Computed from our tables; a metric is ranked only when at least eight peers report it.
| Metric | Value | Against its era | Against its country |
|---|---|---|---|
| Rmax | 379.7 PFlop/s | #8 of 246 · top 3% | - |
| Energy efficiency | 53.4 GF/W | #23 of 130 · top 17% | - |
| Rmax ÷ Rpeak | 71.4 % | #123 of 243 · top 50% | - |
| Power draw | 7,107 kW | #8 of 133 · top 5% | - |
Every metric is ranked largest-first, so a high power-draw rank means a big power draw, not a good one. Rmax ÷ Rpeak uses HPL-measured results only. Rankings move as machines are added; they describe this dataset, not the world.
What this machine is made of
One row per supplier relationship. “Supplier at build” is the company that shipped the part at the time; where that company has since been acquired, the parent it rolls up to today is shown beside it. That distinction is what makes ticker-level aggregation possible across a thirty-year dataset.
| Role | Part | Supplier at build | Quantity | Confidence | Source |
|---|---|---|---|---|---|
| Integrator | - HPE Cray EX235a | Hewlett Packard Enterprise | - | Reported | lumi-supercomputer.euamd.com |
| Accelerator | AMD Instinct MI250X Accelerators ROCm (CDNA 2) · 1.7 GHz · TSMC N6 · 560 W AMD Instinct MI250X, 4 per LUMI-G node | AMD | 11,912 reported | Reported | lumi-supercomputer.euamd.com |
| CPU | AMD EPYC 7A53 (Trento) CPUs x86-64 · 64C · 2 GHz · 7 nm · 280 W AMD Optimized 3rd Generation EPYC 64C 2GHz | AMD | - | Reported | lumi-supercomputer.euamd.com |
| Interconnect | HPE Slingshot-11 Interconnect Dragonfly · 200 Gb/s per port · Around 2 microseconds Slingshot-11 | Hewlett Packard Enterprise | - | Reported | lumi-supercomputer.euamd.com |
| Storage | HPE Cray ClusterStor E1000 Storage Lustre appliance, flash and disk tiers · 700 PB on Frontier (Orion) · About 10 TB/s peak LUMI-P, Lustre on ClusterStor E1000 | Hewlett Packard Enterprise | - | Reported | lumi-supercomputer.euamd.com |
| Cooling | Direct liquid cooling, warm water Cooling Direct-to-chip cold plates, facility water loop · Removes 90+ percent of node heat to liquid direct liquid cooling with district heat recovery | Hewlett Packard Enterprise | - | Reported | lumi-supercomputer.euamd.com |
| Operating system | HPE Cray Operating System Operating systems SUSE-derived Linux, container-based management · Proprietary layer over open-source Linux HPE Cray Operating System | Hewlett Packard Enterprise | - | Reported | lumi-supercomputer.euamd.com |
Engineering, funding and access
What LUMI runs on, who paid for it and how people get time on it. Each line carries who said it and when.
Hardware
| Item | Detail | As of | Basis | Source |
|---|---|---|---|---|
| Compute nodes | 2,978 nodesLUMI-G GPU partition; each node has one 64-core AMD EPYC Trento and four MI250X modules | 2025 | Operator stated | docs.lumi-supercomputer.eu |
| Storage capacity | 117 PB7 PB flash, 80 PB Lustre and 30 PB Ceph; EuroHPC page quotes up to 2 TB/s maximum I/O | 2025 | Operator stated | eurohpc-ju.europa.eu |
| Interconnect and topology | HPE Slingshot-11 on all partitions except LUMI-K; each MI250X module connects directly at 25+25 GB/s peak | 2025 | Operator stated | docs.lumi-supercomputer.eueurohpc-ju.europa.eu |
Software
| Item | Detail | As of | Basis | Source |
|---|---|---|---|---|
| Parallel filesystem | Lustre (80 PB LUMI-P), 7 PB flash and 30 PB Ceph object storageCapacities from the EuroHPC page; LUMI docs list LUMI-P Lustre, LUMI-F flash and LUMI-O object tiers | 2025 | Operator stated | eurohpc-ju.europa.eudocs.lumi-supercomputer.eu |
Funding
| Item | Detail | As of | Basis | Source |
|---|---|---|---|---|
| Funded by | EuroHPC Joint Undertaking, half of a EUR 144.5 million total budget | 2020-10 | Operator stated | eurohpc-ju.europa.eu |
| Funded by | LUMI Consortium of participating countries (Finland, Belgium, Czech Republic, Denmark, Estonia, Iceland, Norway, Poland, Sweden, Switzerland), the other half of EUR 144.5 millionLUMI about page lists 11 countries including the Netherlands | 2020-10 | Operator stated | eurohpc-ju.europa.eulumi-supercomputer.eu |
| Funding programme | EU Connecting Europe Facility and Horizon 2020 through EuroHPC JU, plus national contributions of the consortium states | 2025 | Operator stated | lumi-supercomputer.eu |
| Procurement | HPE selected after a EuroHPC call for tender launched in November 2019; procurement contract signed 21 October 2020 | 2020-10 | Operator stated | eurohpc-ju.europa.eu |
Results
| Item | Detail | As of | Basis | Source |
|---|---|---|---|---|
| Notable result | 2022: third on the November 2022 TOP500 at 309 PFlop/s HPL, then Europe's fastest, with 7th on the Green500 at 51.38 GFlops/W | 2022-11 | Operator stated | csc.fi |
| Notable result | 2022: second place on HPL-AI at 2.168 exaflops (November 2022 list) | 2022-11 | Operator stated | csc.fi |
AI datacenter metrics
Figures beyond an accelerator count, each with who stated it. What can and cannot be compared across AI datacenters.
| Metric | Value | As of | Basis | Note | Source |
|---|---|---|---|---|---|
| Accelerators | 11,912 accelerators | 2026-10 | Operator stated | LUMI-G: 2,978 nodes x 4 AMD MI250X modules (product computed from the docs). Slurm counts each module as two GPUs (23,824 GCDs). | docs.lumi-supercomputer.eu |
The read
Architecturally a smaller Frontier (same AMD compute, same Slingshot fabric, same HPE integration) which makes it the site's best example of why component share and machine count diverge. Its interest here is the facility layer: hydroelectric supply, a reported PUE near 1.03, and waste heat sold into the Kajaani district heating network. Owned by the EuroHPC Joint Undertaking and hosted by CSC, so the operator and the funder are different entities, which the schema keeps separate.
Timeline
- 2020 Announced source
- 2020-10-21 EuroHPC JU signed the LUMI procurement contract with HPE source
- 2022-06 List debut at #3 (June 2022) source
- 2022-06 First operational source
- 2022-11 LUMI ranked third on the November 2022 TOP500 at 309 PFlop/s, then the fastest system in Europe source
- 2022-11 LUMI began its second pilot phase ahead of general availability to European users source
- 2023-06 LUMI-G partition expansion source
Public list appearances
Pointers only: rank and edition, linking to the canonical entry. We do not reproduce list tables. See sourcing policy.
Site & facility
LUMI Data Center, Kajaani
- Location
- Kajaani, Finland
- Cooling
- Direct liquid cooling with waste-heat recovery
- PUE
- 1.03
- Commissioned
- 2022
Nearby in scale
The machines closest to LUMI by Rmax, smaller to larger.
Related systems
Same site
Same operator
- Mahti · Finland, decommissioned, 2020
- Puhti · Finland, decommissioned, 2019
- Sisu · Finland, decommissioned, 2012
- Taito (CSC) · Finland, operational, 2012
- Louhi · Finland, operational, 2007
Same accelerator
Source check
We fetched this system's own citations and recorded whether each page actually mentions it. This is a corroboration signal, not a fact check, and it is published so you can see how well the sourcing holds up rather than take it on trust.
| Cited source | Result | Found on the page |
|---|---|---|
| eurohpc-ju.europa.eu | does not name this system | - |
| lumi-supercomputer.eu | names system only | - |
Checked 2026-10-18 by pnpm verify. Re-run it and the table changes with the web.
Related news
- Bull Beats Out HPE For Next-Gen Lumi AI Supercomputer · The Next PlatformEuroHPC and CSC Finland picked Bull over incumbent integrator HPE to build the LUMI-AI upgrade partition, with delivery slated for the second half of 2027. Worth tracking on the graph because it's a rare case of an existing HPE/Slingshot machine losing its re-up bid to a European vendor running its own BXI interconnect, which fits EuroHPC's stated preference for buying European where it can.
- IQM Sets Three-Stage Upgrade Path for LUMI-IQ Quantum System · HPCwireThis is a procurement commitment, not a delivered system: CSC's LUMI site in Kajaani gains an IQM quantum computer on a three-stage path, from a 2027 Halocene H4 install to a 2029 Halocene H5 upgrade, sourced entirely to the vendor's own press release. It matters to the graph because it ties a quantum system to a site LUMI already occupies here, well before there is any hardware to add as a component row.
- LUMI enters service in Kajaani with a reported PUE near 1.03 · LUMIHydroelectric supply and waste heat sold into the municipal district heating network. The clearest facility-layer story in the European fleet.
Further reading
Inside a node
Why accelerators exist, what the memory hierarchy costs, and how the CPU and accelerator merged onto one package.
Power and cooling
Why air ran out at 20 kW a rack, how warm-water cooling removes the chillers, and what PUE conceals.
Operating a supercomputer
Acceptance testing, why "installed" and "in production" are six months apart, and the machine lifecycle.