Is there a way how to utilize NVIDIA MIG feature (Multi-instance GPU) with Abaqus? We have a (dual socket) server with only one NVIDIA A30 GPGPU card and we would like to run an Abaqus/Standard job with DMP enabled (mp_host_split=2). But we were not successful so far.

Remark. For NVIDIA MIG feature, see 

https://docs.nvidia.com/datacenter/tesla/mig-user-guide/