Question about virtualization for Machine Learning student and GPU challenges

[ellamella]

First of all, I apologize in advance if I sound stupid because I have little experience in virtualization.

I need to run machine learning/deep learning algorithms, and I would like to make use of a GPU. Virtualbox seems to have a 128MB GPU limitation and no PCI(e) passthrough. VMWare has a 2GB GPU limitation.

I have a flexible budget and was planning to get two RTX 3080 GPUs, one for Windows 11 host and one for the VM. Is it possible to get the full performance out of a GPU on a VM? Should I discard the idea of a VM if I want to make full use of a powerful system and just have a separate Linux (most likely Ubuntu) installation? If a VM is feasible, which one should I use if I don't want to spend too much on a subscription to Virtual Learning Environment(s)? I do get VMware Workstation Pro for free through school.