If I am running Proxmox VE and I passthrough a PCIe GPU, does that mean that the host system won’t be able to use the device that’s been passed through at all anymore?
I was watching a few YouTube videos about this topic earlier today and some of them were talking about how they needed a GPU for the host system itself and the second GPU that is going to be passed through.
If I have a GPU (like one of the older Nvidia cards that actually has two GPU cards on a single PCB), would that still be counted as two GPUs or would that be counted as one?
Your help in educating me in regards to this is greatly appreciated.
correct, the host will not be able to use the device that is passed to a VM. the good news though, once proxmox is set up, the host does not usually need a display.
most of them do count as 2 GPUs, getting them to split and work correctly across host and vm with passthrough is a known migraine trigger.
Yeah, a part of the thought is that I was looking at getting a dual Xeon server, and was thinking of setting that up. But you are right in that once the host has been set up, the rest of the management can be done either over ssh and/or the web interface. So this is true and that is a helpful reminder about that the host doesn’t really need a GPU for itself post-install.
I was thinking about putting in a 3090 (as an example) for gaming, and then a Quadro K6000 for simulations work. (Sidenote: I’m not set on the 3090 yet because I am trying to consolidate my systems down to a single system, in order to cut my total power consumption, so I am going to be optimising between Watts per frame vs. total absolute power consumption of the GPU and to try and find a balance between the two. I just haven’t started testing the GPUs that I have yet for that.