PSA: bug in Nvidia driver causes dbus spam

Nvidia’s recent drivers, including 495.46 contains a potentially serious bug that may degrade performance. There is a workaround for now (in the form of an AUR package linked below), but no word yet from Nvidia on a fix.

systemctl daemon-reload
systemctl enable nvidia-fake-powerd.service
systemctl start nvidia-fake-powerd.service
systemctl status nvidia-fake-powerd.service
8 Likes

Amazing. This also fixes the issue with an laptop’s external panel on Xorg taking 100% CPU when the internal panel is disabled. This bug might also explain some weird performance issues I was having with lagging under heavy CPU use.

OK, need to try all of the things that were broken now, like high CPU for reverse PRIME…

Edit: Nope, Xorg and Wayland are still broken on resume, and reverse PRIME with an Xorg session still churns away.

3 Likes

I dont rly understand that.

How to check if somebody have this bug? At me everything seems to be fine.

Edit:

LOL. Forgot it. Im affected. And it seems im already a long time affected :joy:

I just never had any complications about that, because the Threadripper just rips through it… :

error time=1640009049.450861 sender=org.freedesktop.DBus -> destination=:1.44 error_name=org.freedesktop.DBus.Error.ServiceUnknown reply_serial=126987231
   string "The name nvidia.powerd.server was not provided by any .service files"
method call time=1640009049.482410 sender=:1.44 -> destination=nvidia.powerd.server serial=126987230 path=/nvidia/powerd/datapacket; interface=nvidia.powerd.datapacket; member=AutoflDatapacket
   struct {
      int32 4209
      int32 3839
      int64 27600860516
      double 3.76083e+18
      double 3.90044e+31
      double 1.23678e+31
      double 6.57634e-310
      int32 30991
      int32 30981
   }

126987231 counts unbeliveable fast. It seams to start at 0 at boot… So, since my PC booted today, it already managed to get so manys replies :joy:

BUT what is rly funny, this explained to me, why the last 3 days i saw my waterwheel in the cooling loop spinning. i NEVER saw this spinning. first thought it was, it is because of the weather outside and the heating in my home.

After “applying” this patch, it directly stoped. This cost me a few extra euros electricity. i sue nvidia! xD

1 Like

Addition:

It seems this didnt fixed it for me. After a reboot, dbus-monitor --system shows that still spams fast.

In fact it starts with 0. 2 Minutes later, am already over 10.000. It now only gets a “new name”:

method call time=1640010082.208548 sender=:1.45 -> destination=nvidia.powerd.server serial=12117 path=/nvidia/powerd/datapacket; interface=nvidia.powerd.datapacket; member=AutoflDatapacket
   struct {
      int32 4209
      int32 1386
      int64 165041968
      double 1.72938e+15
      double 8.19441e+30
      double 4.12088e+31
      double 6.46578e-310
      int32 30470
      int32 30471
   }

And my Waterwheel spins again. Dunno if this anyhow affect performance in games for me, because i play anyway only at 60fps locked…

Where do you find logs for it? I’ve hunted around /var/log/ but don’t see anything related to nvidia

sudo dbus-monitor --system

3 Likes

Just a forum hint, you can edit your post rather than posting an addition.

Did you enable the service?

I can recommend that you read the linked threads for more information and context. It also saves people re-posting information in multiple places.

If you’re affected you’ll see errors in the journal, e.g.

2 Likes

Oh man I hadn’t even noticed :sweat_smile:
There is definitely a TON of incessant spam… I will try the fake-powerd service.

Edit: Unfortunately doesn’t work for me. Systemctl status reports it’s up and running after a reboot, but there are still hundreds of dbus messages every second.

Didn’t work for me either.

Yes. Like you can see in my post, the messages are another ones. So i think this is the normal behaviour for this workaround?

Sadly, there is no documentation what should happen after this workaround.

For me, i let it be like it is now. Because i dont see any performance regressions.

What exactly happens?

Did the messages changed like for me?

That was before:

error time=1640009049.450861 sender=org.freedesktop.DBus -> destination=:1.44 error_name=org.freedesktop.DBus.Error.ServiceUnknown reply_serial=126987231
   string "The name nvidia.powerd.server was not provided by any .service files"
method call time=1640009049.482410 sender=:1.44 -> destination=nvidia.powerd.server serial=126987230 path=/nvidia/powerd/datapacket; interface=nvidia.powerd.datapacket; member=AutoflDatapacket
   struct {
      int32 4209
      int32 3839
      int64 27600860516
      double 3.76083e+18
      double 3.90044e+31
      double 1.23678e+31
      double 6.57634e-310
      int32 30991
      int32 30981
   }

And this after Workaround:

method call time=1640010082.208548 sender=:1.45 -> destination=nvidia.powerd.server serial=12117 path=/nvidia/powerd/datapacket; interface=nvidia.powerd.datapacket; member=AutoflDatapacket
   struct {
      int32 4209
      int32 1386
      int64 165041968
      double 1.72938e+15
      double 8.19441e+30
      double 4.12088e+31
      double 6.46578e-310
      int32 30470
      int32 30471
   }```
1 Like

All hail to Eye of Threadripper!

3 Likes

For me i try it …i uninstall it! Is bloat! :laughing:

1 Like

No, im serious. I dont have any problems. Son goku watches and protect everything!

2 Likes

I was linked here from Blender 3.0 viewport extremely laggy after days of working perfectly in Endeavouros
and am a bit lost. In terms of getting a certain feature to work, has a solution been found?

It seems that this depends.

Because i dont know what should happen if you apply that workaround from AUR.

But what rly fixes this, is to install the last stable 470.xx one.

So are there specific nvidia drivers that are causing this or will I have to uninstall all 495 drivers and install all 470.xx drivers.