HPC Queue Prioritisation

The University provides centralized high performance computing (HPC) resources to support its academic and research communities.  The existing HPC facility consists of CPU clusters that provides up to 1650 x86-64 Intel Xeon cores and 14336 2nd Gen AMD EPYC cores, as well as GPU clusters with 64 double-precision GPU cards and 8 single-precision GPU cards.

Users can submit their jobs under the Linux environment through the Slurm scheduler, or through the LiCo Web portal. All jobs will be run when resource is available on a first-come-first-served basis according to the fair-sharing mechanism setup in the scheduler.  Users are grouped by their registration status at University, and currently:

• Academic staff are allowed to use up to 180 CPU cores on 5 computing nodes for 7 days, or 4 double-precision GPU cards on 2 nodes for 3 days.
• Students are allowed to use 4 CPU cores in a single node for 4 hours, or 2 double-precision GPU cards for 1 day.

Job queue configurations, scheduling policies, and fair sharing mechanism are reviewed regularly based on the demand and availability of resources, governed by the HPC Steering Committee.