Radeon rx 6950 xt reset bug question

[softwaretools]

Hi,

I was wondering if the 6950 xt series has the reset bug? I know the 6800 xt series is supposed to have been fixed but does that mean future generations are fixed too? Thanks!

[Sensai]

I have no idea what this reset bug is you speak of, but I just upgraded my 2080 super to an ASRock 6950xt in an ASRock x370 Tai chi with a 5950x CPU, and have had zero problems. And It is really powerful / fast.

[wayland]

The reset bug is that when used in a VM you have to reboot the physical PC to get the GPU to reset. It means a VM with a passed through GPU cannot be rebooted. The HD5970 did not have this problem but the VEGA 56 did. The problem was said to have been fixed in the RX 6000 series but I’ve not been able to try it.

It’s not a driver problem but a hardware or VBIOS problem with the GPU.

[Sensai]

Thanks for the clarification. I just learned something today. It is a good day.

[H-i-v-e]

6000 Series AMD still has the reset bug as this issue has not been addressed from AMDs side. You might be lucky and your card resets, you might be unlucky and your card does not. For PCI passthrough I can only repeat what @gnif has told people repeatedly: “For now your best bet is an NVIDIA card, unfortunately”.

[Janos]

A friend of mine has the 5700XT with the reset bug.
With this script and vendor-reset, the reset works without problems.
This is for Promox, but you can also customize it for Manjaro or Arch by using libvirt hooks.

libvirt-hooks

apt install pve-headers;apt install git dkms build-essential

git clone GitHub - gnif/vendor-reset: Linux kernel vendor specific hardware reset module for sequences that are too complex/complicated to land in pci_quirks.c

cd vendor-reset

dkms install .

echo “vendor-reset” >> /etc/modules

update-initramfs -u

reboot

–create script

mkdir /var/lib/vz/snippets

nano /var/lib/vz/snippets/gpu-hookscript.sh

------Hook Script

#!/usr/bin/perl -w

use strict;
use warnings;

print "GUEST HOOK: " . join(' ', @ARGV). "\n";

my $vmid = shift;
my $phase = shift;

# Reuse our command
my $cmd ="echo 'device_specific' > /sys/bus/pci/devices/0000:03:00.0/reset_method";

if ($phase eq 'pre-start') {

    print "$vmid is starting, doing preparations.\n";
    system($cmd);

} else {
    die "got unknown phase '$phase'\n";
}

exit(0);

---

chmod +x /var/lib/vz/snippets/gpu-hookscript.sh

##Add script to VM config
qm set 101 --hookscript local:snippets/gpu-hookscript.sh

[Janos]

@H-i-v-e

6000 Series AMD still has the reset bug as this issue has not been addressed from AMDs side

I don’t know, all I can say is my Sapphire NITRO+ 6800XT works

[wayland]

Thank you. It would be nice if Proxmox added this as a GUI option but I think I can handle adding it manually.

[louhy]

@H-i-v-e

6000 Series AMD still has the reset bug as this issue has not been addressed from AMDs side. You might be lucky and your card resets, you might be unlucky and your card does not. For PCI passthrough I can only repeat what @gnif has told people repeatedly: “For now your best bet is an NVIDIA card, unfortunately”.

Ugh, so we really can’t do any better than “if you are lucky”? I mean is it dependent on the manufacturer? The motherboard or CPU you’re using with it? How much RGB is on the card? (just kidding)

I’ve got this XFX 6800 XT “Merc” (still in box) which I was hoping would be good to go, now I’m not so sure.

[Susanna]

Sometimes my 6900XT will reset, sometimes it won’t. Mostly it won’t.

[louhy]

Maybe that video where Wendell said it was working (for 6800 XT anyway) was just luck. Still I might as well try it once I get everything else I need to do so.