Can I use GPU Passthrough with ESXi from vSphere 7 Essentials?

[bogdan-k]

Hello Everyone,

I apologize if this questions was previously answered. I searched but could not find. 🙂

We have vSphere 7 Essentials with ESXi 7.0U3g (standard). We have been using it on a Dell server that has no GPUs so far with various Linux guest OS's. We would like to get another Dell server with a number of NVIDIA GPUs (e.g. RTX 4000/4500/5000 ADA, or L4). We would like to use those GPUs within the VMs by installing and configuring ESXi host.

We are interested in a passthrough approach, i.e. we'd like for each VM to have dedicated access to one of the GPUs. For example, with 4 GPUs we would have 4 VMs that would have access to one GPU each.

We have never virtualized the GPUs before. So, we would like to know if we have to get any additional software from vmware or NVIDIA in order to accomplish this, e.g., any special NVIDIA drivers for Linux guest OS or any NVIDIA drivers for the ESXi host?

We are not interested in NVIDIA AI Enterprise and splitting GPU into several vGPUs of smaller (memory) size. We just want to use the passthrough so each VM would have full access to a GPU.

Another question: If this is possible, do we have to use vCenter or we can configure it all directly through ESXi web portal?

Regards,
Bogdan

[benkeprashant · Accepted solution]

yes looks like that is possible. refer https://via.vmw.com/tchzcoreno3485  

[bogdan-k]

Thanks for a quick answer. 👍

I guess using vCenter or ESXi for configuration would not matter much. We can certainly install vCenter, but for just two ESXi hosts we did not see reason for it. Is there anything special we would gain from using vCenter as opposed to just ESXI directly?