Nvidia AI enterprise - vGPU / MIG confusion

Hi!

Looking at ways to deploy AI workloads on vSphere.
I’m very familiar with how vGPUs work. MIG - not so much…

Have I understood this correct:
You either have to choose between deploying vGPUs or MIG GPUs to your VMs.
But you can also deploy a VM with a vGPU and split this vGPU with MIG from the guestOS?

I really want to keep the flexibility up my having the ability to vmotion VMs.

1 Like