Is OpenPBS able to manage GPUs when MIG is enabled?

[pablo_h]

I’m trying to implement Multi-Instance GPU in a HPC cluster in order to increase its throughput. Let’s say I use MIG to create 7 GPU instances (1g.10gb, NVIDIA A100 80GB) and try to run 8 jobs using GPUs. Does OpenPBS automatically manage which job is going in which GPU instance (without the need of especifying the UUID) and places the 8th job in queue?
Also, is there the need of any additional configuration to be able to use MIG?

[adarsh]

Please refer:

MIG - open PBS support

Hi, Does MIG - Multi instances GPU supported in open-pbs? This link suggests it is supported without any change to configuration files etc, but I succeed to work on GPU and fail to work on MIG (I get an allocation of only GPU - not MIG although they are defined and works statically). Thanks at advance

https://openpbs.atlassian.net/wiki/spaces/PD/pages/2313453569/Nvidia+MIG+Support

@pablo_h

Does OpenPBS automatically manage which job is going in which GPU instance (without the need of especifying the UUID) and places the 8th job in queue?

Yes it would do the scheduling as stated in your query.

@pablo_h

Also, is there the need of any additional configuration to be able to use MIG?

References above should help
Refer the guide: https://help.altair.com/2024.1.0/PBS%20Professional/PBS2024.1.pdf and
Nvidia MIG Support