Due to ongoing cooling issues, most COSMOS GPU nodes are currently unavailable. Currently there are currently only two nodes with NVIDIA A100 cards and Intel processors available for GPU computing. Check the LUNARC user guide (Read the docs website) for instructions on how to use these nodes using the batch system. Unfortunately long queueing times are expected.
LUNARC's on-demand applications are also affected, if they request GPU resources in the menu. GPU resources are named "AMD/NVIDIA A40 48c 24h", "Intel/NVIDIA A40 32c 48h", "AMD/NVIDIA A100 48 cores" and "Intel/Nvidia A100 32 cores". For many applications the on-demand system also offers CPU versions, running on CPU nodes. Utilising CPU applications on CPU nodes might be a work around for the current issues.
We currently expect to offer more GPU nodes during week 34. Please contact LUNARC support if you have questions or need assistance.
GPU nodes in the COSMOS SENS system are not affected.
LUNARC apologises for the issues and any inconvenience they may cause.