Bug 500955 - NVIDIA dGPU with RTD3 support could not be turned off due to plasmashell using the GPU in Wayland session
Summary: NVIDIA dGPU with RTD3 support could not be turned off due to plasmashell usin...
Status: RESOLVED WORKSFORME
Alias: None
Product: plasmashell
Classification: Plasma
Component: general (other bugs)
Version First Reported In: 6.3.2
Platform: openSUSE Linux
: NOR normal
Target Milestone: 1.0
Assignee: Plasma Bugs List
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2025-03-02 13:01 UTC by Sam Duan
Modified: 2025-03-18 08:25 UTC (History)
3 users (show)

See Also:
Latest Commit:
Version Fixed/Implemented In:
Sentry Crash Report:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Sam Duan 2025-03-02 13:01:35 UTC
SUMMARY
NVIDIA dGPU on a laptop with RTD3 power management cannot be powered off in Plasma 6.3+ (Wayland only session, sddm also runs on Wayland)

STEPS TO REPRODUCE
1.  Install the NVIDIA closed-source driver and set up RTD3 power management according to the official guide.
2.  Reboot and log in to Plasma.
3.  Wait. During the wait, check the dGPU suspend time using:
watch -n 2 cat /sys/bus/pci/devices/0000:01:00.0/power/runtime_suspended_time

OBSERVED RESULT
The suspend time do not accumulate. -> dGPU doesn't power off.

EXPECTED RESULT
The suspend time should be accumulating. -> dGPU is turned off.

SOFTWARE/OS VERSIONS
Linux version:  openSUSE Tumbleweed 20250228, kernel 6.13.4
KDE Plasma Version: 6.3.2
KDE Frameworks Version: 6.11.0
Qt Version: 6.8.2
NVIDIA driver version: 570.124.04 (seems not matter since version 550 also had this issue)

ADDITIONAL INFORMATION
* The kernel module parameter nvidia.NVreg_EnableGpuFirmware=0 is set.
* Powering off worked in Plasma 6.2 with kernel 6.12 and NVIDIA driver 550.
* Powering off works in Weston and Hyprland with the same Linux version and NVIDIA drivers.
* Checking ~~~ fuser -v /dev/nvidia* ~~~ gives:
/dev/nvidia0:        user    3873 F...m plasmashell
/dev/nvidiactl:      user    3873 F...m plasmashell
/dev/nvidia-uvm:     user    3873 F.... plasmashell
Comment 1 Nate Graham 2025-03-05 16:14:47 UTC
Is there an integrated GPU on the system that Plasmashell could use instead?
Comment 2 Sam Duan 2025-03-06 03:38:43 UTC
(In reply to Nate Graham from comment #1)
> Is there an integrated GPU on the system that Plasmashell could use instead?

Being a laptop, there is an Intel iGPU available. I have checked intel_gpu_top, kwin_wayland runs on the iGPU, but the dGPU is still occupied.
BTW, I have used supergfxctl to fully shutdown the dGPU, and when I switched from iGPU only to hybrid mode with it and loged off just now for some testing, the dGPU seemed not to be occupied in a new Plasma session without the panels loaded.
Comment 3 Sam Duan 2025-03-18 08:25:30 UTC
The bug seems to be fixed in Plasma 6.3.3 on my laptop.