SM Nodes

IBM3755.jpg

We have 11 IBM System X3755 M3 nodes for high memory jobs, this queue is restricted to members of Queen Mary University of London. This tends to be a busy queue, with longer wait times than the dn nodes.

SM IBM System X3755 M3
Count 11
Processor 4 x 12 Core AMD Opteron 6234 (Bulldozer)
Cores/Node 48
RAM 512GB
TMP Size 1.9TB
Interconnect 10Gb Ethernet

Notes

  • On a core-by-core comparison, a dn node core is faster than an sm node core. Avoid excluding your jobs from faster nodes due to restrictive configuration options (too much memory requested vs memory used, specifying a specific node_type). If your jobs regularly require large memory requirements, consider optimisation or splitting the job into smaller components.
  • sm nodes also have the largest local temporary folder space. Using tmpdir rather than central storage can result in considerable performance improvements.