Lem

Lem (WCSS)

In 2024, the Lem cluster was launched, the Center's newest computing machine, which in June 2024 ranked 80th on the TOP500 list of the world's fastest computers.

The Lem supercomputer infrastructure combines CPU and GPU computing machines. The CPU partition consists of 188 nodes, 23,808 cores, and 279 TB of RAM. The GPU partition consists of 76 nodes equipped with 304 NVIDIA H100 accelerators connected via NVLink technology with 96 GB of cache per accelerator.

In addition, the cluster is equipped with 64 service nodes, including: 40 GPU nodes (NVIDIA A30), 24 CPU nodes (48 cores per node). The CPU and GPU computing nodes are equipped with 1.5 TB and 1 TB of RAM, respectively. The network is based, among other things, on InfiniBand 4x NDR 200 Gb/s interfaces for GPU nodes; 2x NDR 200 Gb/s for CPU nodes and 2x HDR 100 Gb/s for service nodes. The total computing power of the system is 21 Pflops.

Cluster storage:

Disks shared using the NFS, HTTP(s), and CIFS network protocols – 2 PB of space, 6 GB/s bandwidth,
Disks shared using the LUSTRE network protocol – a total of 2 PB of space (permanent and temporary),100 Gb/s bandwidth,
S3 object storage – 6 PB, bandwidth 200 Gb/s,
Local temporary storage space of supercomputer nodes – a total of 1 PB NVME.

Lem

Lem (WCSS)

Other clusters

Ares (ACC Cyfronet AGH)

Athena (ACC Cyfronet AGH)

Helios (ACC Cyfronet AGH)

Tryton Plus (CI TASK)