Resource types can have different versions, and each version can have different performance characteristics. For example, a CPU job might run on a Sandy Bridge, Haswell, or Skylake CPU. These processors belong to different generations and have different processing speeds. SLURM tracks these differences and charges service units (SUs) accordingly. The table below shows the current SU charges by resource type and version.

Service units (SUs) are the billing units used to account for compute usage on the system.

The SU charge for H200 GPUs is still being evaluated and will be updated when finalized.

Sandy Bridge and Haswell CPUs are only used in the DLX 2 nodes. DLX 2 nodes are available at no cost to users. As a result, the SU charge per core per hour on DLX 2 nodes is 0 SUs.

The following table shows the relative performance of the processors that are part of the LCC system. Sandy Bridge CPUs and P100 GPUs are used as the baseline with a relative performance of 1.0.