Nvidia Quadro partially working with performance spikes

[aronmal]

Hello,

 

I have set up a Windows VM with a Nvidia Quadro P2000 and it kind of works, but VNC, Anydesk and Parsec all have a laggy transmission. I noticed the performance graph in the Task Manager of the GPU being 0% and spiking rarely to 100% (seem to match, when I do receive frames). I feels like getting worse over time and eventually ends in crashes (at least not getting any more frames at some point).

 

I used a Geforce RTX Card some time ago with my main PC and it worked all really well with a Windows VM with dumped vbios and everything. I tried to repeat it with the Quadro (dumping the vbios as well as editing the GPU part of the XML to have multifunction and adding to the same slot) but sadly no luck in improvements.

 

And I did not notice any problems running the Quadro in a PC bare metal, so the card should not be broken. I am using the i440fx-6.2 machine and OVMF Bios. I tried the other but with so far without better results.

 

Any Ideas or suggestions what I could try?
 

Thanks for participation.

[aronmal]

I also noticed, the graph in Task Manager moving very slow. Like if the VM would run in slow motion. This all did not happen when using the VNC graphics.

 

But neither the VM nor Unraid do seem to be exhausted:

 

 

 

 

[aronmal]

And here you can see the up time 5 mins later and the 100% spike of GPU 0 just went left out of the frame (I was to slow) which is really odd.

 

[aronmal]

Ok, further test. The Nvidia logo in the preview is spinning smoothly, but the task manager is still only updating every 7-8 sec. And when logging as well as starting up programms, the picture freezes and takes quite long to open/load.

 

 

Here I opened VS Code. The logo stopped spinning, next frame was the blank vs code window in maximize and after being able to hit the minimize button, I have this window arrangement where the Nvidia logo is visible again and also spins smoothly again, but the vs code window is still not loaded, but loaded eventually after some time.

[aronmal]

I also noticed:

 

I had 8 cores (16 threads) of my 16 core (32 threads) assigned as well as 12 GB of 64 GB RAM. I noticed with the prometheus exporter in Grafana that the RAM usage was oddly high. So the hard changes is turning the VM on and off and I reduced RAM from 12 to 8 GB, according to this diagram at around 13:45. I also reduced to 4 cores (8 threads), but did not seem to make a difference, same as isolating the cores (the restart caused the "crash" of Memory usage at the end of the diagram).

 

 

But what wonders me is, the Unraid Dashboard only seems to show the yellow "RAM Used", but not letting the green "RAM Free" run out and being overtaken by the blue "RAM Cache + Buffer" left me at least with some frames after waiting some time, and not getting less and less frames like before.