Hi,
This question: I can passthrough a PCIe device to a VM. Also a graphics card should work I understood. But what if this is the ontly card in the system? I don't need a display output for the PM install: the web interface and CLI over ssh will do just fine. But I need it for booting I assume...?
The idea:
- install proxmox with the GPU available to the PM installation. Works.
- pass the GPU through to a VM (iommu). But this means I'll take away 'live' the GPU of the PM install, since the PCIe device is now isolated, and only availabe to the VM. How will PM react? crash? hang? unexpected?
I have 2 systems in mind;
1) a intel nuc with integrated graphics. but that's not different from any other PCIe device to pass to a host I assume?
2) a system with an AMD without integrated graphics, but with only 1 discrete graphics ward, that I would like to make availabe to a single VM.
In both cases, the boot process would of course with the GPU available to the system. But then iy would be taken away in favor og being available to the host.
Is this possible? and if so, what would be the best approach?
Thanks!
This question: I can passthrough a PCIe device to a VM. Also a graphics card should work I understood. But what if this is the ontly card in the system? I don't need a display output for the PM install: the web interface and CLI over ssh will do just fine. But I need it for booting I assume...?
The idea:
- install proxmox with the GPU available to the PM installation. Works.
- pass the GPU through to a VM (iommu). But this means I'll take away 'live' the GPU of the PM install, since the PCIe device is now isolated, and only availabe to the VM. How will PM react? crash? hang? unexpected?
I have 2 systems in mind;
1) a intel nuc with integrated graphics. but that's not different from any other PCIe device to pass to a host I assume?
2) a system with an AMD without integrated graphics, but with only 1 discrete graphics ward, that I would like to make availabe to a single VM.
In both cases, the boot process would of course with the GPU available to the system. But then iy would be taken away in favor og being available to the host.
Is this possible? and if so, what would be the best approach?
Thanks!