Cluster compute nodes

Also known as: comp nodes, batch nodes, or worker nodes.

The cluster consists of about 4000 CPU cores, spread across compute instances accessible via SLURM.

• comp-d-[001-044]
  • 32 CPU threads (2x Xeon E5-2660 @ 2.2GHz)
  • 128GB RAM
  • 1408 threads total
  • container instances on wn-d
• comp-[e,s]-[001-044]
  • 32 CPU threads (2x Xeon E5-2630 v3 @ 2.4GHz)
  • 128GB RAM
  • 1408 threads total
  • container instances shared on wn-e
• comp-[f,r]-[001-028]
  • 64 CPU threads (2x Xeon Silver 4216 @ 2.1GHz)
  • 256GB RAM
  • 1792 threads total
  • container instances shared on wn-f
• gpu1
  • 9x 8GB RTX2070S
  • 48 CPU threads (2x Xeon Silver 4214R @ 2.4GHz)
  • 384GB RAM
  • container instance on wn-x-01
• gpu2
  • 2x 80GB A100
  • 64 CPU threads (2x Xeon Silver 4314 CPU @ 2.4GHz)
  • 512GB RAM
  • container instance on wn-x-02
• gpu3-5
  • 2x 48GB L40s
  • 336 CPU threads (2x AMD EPYC 9634 84-Core Processor @ 3.7GHz)
  • 2.2TB RAM
  • container instance on wn-g-[01-03], shared with CMS production jobs

To view the current status of the cluster, run sinfo.