Hi guys, I have a problem

I have several VMs under Hyper-V and my system crashes when I use the GPU within a VM. It doesn't matter if i use GPU-PV OR pass the GPU completely through. with GPU-PV the crash happens as soon as i start a benchmark for example furmark or heaven. with pass through it happens as soon as the driver is loaded (immediately after/while booting).

the entire system crashes, not just the vm

host:

threadripper 3970x

8x32gb samsung ecc 2666

2x rtx 3070

1x gtx 970

some SSDs and hard disks

SR-IOV and IOMMU are of course activated, otherwise it wouldn't have worked before.

It was first tested on Windows 11 (since you can't select the GPU for GPU-PV under Windows 10) and then Server 2019 since non-server version can't pass through PCI-e.

Drivers used: the latest drivers freshly downloaded from nvidia.

it doesn't matter if only one or all 3 graphics cards are connected. it also doesn't matter which one, a 970 or one of the 3070. it still crashes as soon as there is a load in a VM with a GPU.

the graphics cards work. there is never an issue when i run e.g. furmark on all 3 graphics cards at the same time as long as it is on the host. i have also recreated the VM several times and reinstalled the windows in it several times to rule out problems with drivers.

there are absolutely no clues in the event log. only that the system was shut down unexpectedly beforehand, which happens after a crash.

does anyone have any idea why it crashes ONLY INSIDE A VM? i cant provide any logs or similar since there are non. it does not bluescreen and thherefore no memory dump