LUMI

EuroHPC/CSC

Node type

CPU Accelerator Interconnect

Reading this map

LUMI node architecture

LUMI is operated by EuroHPC/CSC in Finland. This profile covers Ranked system / partition. Specifications are listed below; anything the operator has not published is itemised in the documented-fields list.

Published benchmark (TOP500 · June 2026)

HPL rank: #11
HPL Rmax: 379.7 PFlop/s
Theoretical peak Rpeak: 531.51 PFlop/s
Ranked core count: 2,752,704
Measured power: 7107 kW
Energy efficiency: 53.43 GFlop/s per watt

Cluster · LUMI-G

Operator / site: EuroHPC/CSC
Country: Finland
Profile scope: Ranked system / partition
Data checked: 14 September 2026
Manufacturer: HPE
Ranked core count: 2752704
Theoretical peak (PFlop/s): 531.51
Measured power (kW): 7107
Energy efficiency (GFlop/s per watt): 53.42619952159843
Operating system: HPE Cray OS
Network topology: Dragonfly

LUMI-G

System / node model: HPE Cray EX235a
Compute nodes of this type: 2978
CPU model: AMD EPYC 7A53 "Trento" · 64 cores
CPU sockets per node: 1
Cores per socket: 64
Physical CPU cores: 64
Accelerator model: AMD MI250X
Physical accelerators per node: 4
Host / unified memory: 512 GiB DDR4
Accelerator memory: 128 GB per physical accelerator
Accelerator ↔ accelerator link: Infinity Fabric
On-node fabric: Infinity Fabric — up to 400 GB/s bidirectional in-package between the two GCDs of one MI250X; 100 GB/s (single link) or 200 GB/s (double link) between GCDs on different modules
Network interfaces per node: 4
Interface model: HPE Cray Cassini (Slingshot 11)
Interface speed: 200 Gb/s per endpoint · 25+25 GB/s peak
Interface-to-accelerator topology: Four Slingshot endpoints, a dedicated adapter per MI250X accelerator module
Local storage: None — CSC states there is no local storage on LUMI-G compute nodes

Scope and uncertainty

LUMI-G partition only. Each physical MI250X contains two GPU dies; diagram counts physical accelerators. LUMI-C and LUMI-D are outside this profile.

Documented fields: 24 / 29

Every field this catalog checks. Fields listed above have a value; the remainder are shown here. A field marked absent is not publicly available in the sources we could verify, which does not mean the hardware is missing. Fields that cannot apply, such as benchmark results for an unranked system, are excluded from the total.

Cluster — 7 / 7

Manufacturer
Ranked core count
Theoretical peak (PFlop/s)
Measured power (kW)
Energy efficiency (GFlop/s per watt)
Operating system
Network topology

LUMI-G — 17 / 22

System / node model
Compute nodes of this type
CPU model
CPU sockets per node
Cores per socket
Physical CPU cores
CPU note
Accelerator model
Physical accelerators per node
Host / unified memory
Accelerator memory
Accelerator ↔ accelerator link
On-node fabric
CPU ↔ accelerator link
Network interfaces per node
Interface model
Interface speed
Interface-to-accelerator topology
Aggregate scale-out per node
GPUDirect RDMA
Management / storage networking
Local storage
NVMe data disks
Max remote data disks
Uncached disk IOPS
Uncached disk throughput (MBps)
Form factor

Sources

Browse all cluster architecture pages →