After upgrading from NVIDIA 610.57.04 to 615.71.09, my DisplayPort monitor intermittently fails to wake after KDE turns it off due to inactivity.
This is display power management only. The computer is not entering suspend or hibernation.
System information
Distribution: CachyOS
Desktop: KDE Plasma 6.7.5
Session: Wayland
KWin: 6.7.5
GPU: NVIDIA GeForce RTX 5090
Driver package: nvidia-open-dkms
Affected driver: 615.71.09
Kernel: 7.2.4-1-cachyos
Monitor: Samsung Odyssey G95NC
Connection: DisplayPort
Mode: 2560x2160 at 119.96 Hz
Scale: 125%
Symptoms
- I leave the computer idle.
- KDE turns off the display according to its normal energy-saving timeout.
- When I return, the display remains black and does not respond to keyboard or mouse input.
- Plasma reports that no display outputs are available.
I reproduced this twice with 615.71.09.
During the first occurrence, the desktop appeared completely unresponsive and SSH was unavailable. I eventually performed a forced reboot. However, the journal shows that the kernel and other services continued logging for approximately 100 seconds after the display disappeared, so it was not an immediate kernel lockup.
During the second occurrence, I repeatedly tried switching virtual terminals with combinations such as:
Ctrl+Alt+F3
Ctrl+Alt+F7
The display eventually recovered, although recovery was delayed and I cannot confirm which specific key combination triggered it.
Relevant journal messages
At the time of both failures:
plasmashell: requesting unexisting screen available rect -1
plasmashell: There are no outputs - creating placeholder screen
The second occurrence happened at approximately:
20:51:36 - Plasma reported no outputs and created a placeholder screen
After recovery, the connector was available again:
Output: DP-1
Status: enabled and connected
Mode: 2560x2160@119.96
Scale: 1.25
I found no corresponding NVIDIA Xid, NVRM fault, PCIe/AER error, kernel panic, GPU timeout, or lockup message during either occurrence.
Troubleshooting performed
I initially suspected LACT because it was running when the first failure occurred. I disabled LACT completely, but the issue reproduced, so LACT does not appear to be the cause.
The upgrade that introduced the problem changed both the driver and kernel:
Before:
NVIDIA 610.57.04
Kernel 7.2.3-1-cachyos
After:
NVIDIA 615.71.09
Kernel 7.2.4-1-cachyos
I restored NVIDIA 610.57.04 and then installed kernel 7.2.4 again. The current test configuration is:
Kernel: 7.2.4-1-cachyos
NVIDIA-SMI: 610.57.04
NVIDIA KMD: 610.57.04
Driver package: nvidia-open-dkms
NVIDIA 610 compiled successfully against kernel 7.2.4, and the kernel module and userspace versions match. I am currently testing this combination to separate the NVIDIA driver change from the kernel change.
Before the upgrade, NVIDIA 610.57.04 successfully recovered from multiple normal display-off events.
This appears to be a DisplayPort DPMS/hotplug regression introduced with 615.71.09. Is there any additional logging or a specific module parameter that would help confirm this?
Update: I tested the same kernel, 7.2.4-1-cachyos, with NVIDIA 610.57.04. After leaving the system idle for over an hour, the monitor woke normally and the desktop was immediately responsive. The failure occurred twice with 615.71.09 but has not occurred with 610.57.04 on the same kernel.