gnome-remote-desktop-daemon prevents nvidia gpu from powering off even if there is no active rdp session
gnome-remote-desktop-43.2-1 and gnome-shell-43.3-1 on Fedora 37
On a dual-GPU laptop here and X11 session, with AMD iGPU being the primary GPU, as soon as the gnome-remote-desktop service starts, it seems to be using the Nvidia dGPU, even if an active remote desktop session which needs the hardware-accelerated encoding has not yet been started. /usr/libexec/gnome-remote-desktop-daemon
always shows up in the list of processes in the output of nvidia-smi
.
However, this prevents the dGPU from powering off, even when nothing else (monitors connected to it or other applications) is using the GPU, which increases power consumption and heating. This can be seen from watching /sys/class/drm/card1/device/power_state
staying at D0
, and nvidia-smi
showing the GPU consuming some power. After running systemctl --user stop gnome-remote-desktop.service
, the process is gone and /sys/class/drm/card1/device/power_state
changes to D3cold
in a few seconds.
Would it be possible for gnome-remote-desktop to start using the GPU encoding only when an actual remote desktop connection is made instead of when the daemon starts?