550.76 - EDID readout problem and nvidia-powerd error looping

The new 550.76 driver update brought problems that were not present with the previous 550.67 driver version.

The driver cannot read out the EDID information of the 4K monitor connected to a HDMI port, and now uses an emergency 1024x768 resolution.

Also, the nvidia-powerd continously logs several errors per second into the journal:

Apr 18 20:24:44 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:44 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:44 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:44 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:45 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Failed to get topology status 55
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: error setting power limit
Apr 18 20:24:46 DESKTOP-J9DPOC4 /usr/bin/nvidia-powerd[7081]: Error setting GPU limit: 115000.

All this on an Asus G713PV laptop with a RTX 4060 GPU which has an external 4K display connected via HDMI.
The system runs Fedora Kinoite 40 beta with kernel 6.8.5-301.fc40.x86_64 and this is the same kernel which also ran with the previous 550.67 nvidia driver version which did not show this (error) behavior.

nvidia-bug-report.log.gz (661.6 KB)

1 Like

I’m having this same problem, same video card, same fedora, same driver… different laptop (Lenovo Legion Slim 5)

I also have same powerd service problem. A workaround is downgrading to 550.67. Arch linux, Acer PHN16-71 RTX 4060 laptop.
nvidia-bug-report.log.gz (813.9 KB)

1 Like

I have the same issue as well and a (maybe related) side effect on my system (2023 ASUS Zephyrus G14) with an AMD 7940HS and RTX 4060 Laptop GPU. My gpu is stuck the P0 power state and will not idle below 13W, even on the TTY. With 550.67, it would idle at ~1W.

nvidia-bug-report.log.gz (545.2 KB)

Indeed, my GPU also seems stuck in P0 and consumes 20W at idle (no monitors attached).

Hi All,
Thanks for reporting the issue, I have filed a bug 4613646 internally for tracking purpose.

1 Like

I’m experiencing similar log messages and high idle power usage as well, using 550.76 on an RTX 4050 laptop. For me it’s stuck in P3, not P0 though, and I don’t have any resolution issues.

nvidia-bug-report.log.gz (614.5 KB)

I just realized my EDID readout problems on the HDMI port are monitor-specific and plug order specific.
I had the original EDID readout problems on a (more modern) Gigabyte Aorus FV43U (4K@144Hz, HDR).
An older Iiyama PL4071UH (4K@60Hz max) does not have such problems, there the readout works and 4K@60Hz is possible.

I then did some (wild) live-replugging session and at some point I had the Aorus connected via an USB-HDMI converter and the Iiyama via the native HDMI connector - both working at 4K.
Then I unplugged the Iiyama and changed the connection of the Aorus from the USB-HDMI converter to the native HDMI port - and in this situation the Aorus worked connected to HDMI and with 4K@120Hz.
Weird stuff.

Issue has been root caused and fix will be integrated in upcoming release drivers.
Thanks everyone for the help.

3 Likes

Same problems here with nvidia-powerd but i cannot set the refresh rate of external monitor.

The problem emerges when upgrading to 550.67 driver.

“… when upgrading to 550.67 driver” - can you confirm the driver version?
Because the rest of us saw no problems with 550.67 and the problems started for us with 550.76.

Hello @amrits ,
can you maybe disclose whether only a specific set of GPUs are affected by the root cause or whether this is affecting all GPUs?
I wonder whether I should make the upgrade to 550.76 on my workstation GPUs - Quadro RTX 4000 class and newer variants.
I would hold back for now, but if the problem is just limited to the Mobile RTX 4050/4060 range, then the driver might be ok.

sorry, scratch that, I just saw the release of the 550.78 driver

The 550.78 driver fixed the problem for me, thanks!