Arch Linux

Ne0Wolf7

Member

Registered: 2024-06-06

Posts: 1

Asus GL552V laptop will not use nvidia 960M dGPU

Hi all, I am in the process of setting up arch on an Asus GL552V laptop, which has both Intel 530 integrated graphics and nvidia 960M dedicated graphics. I am following instructions from the wiki page NVIDIA Optimus (https://wiki.archlinux.org/title/NVIDIA_Optimus#) trying to get the computer to use the dedicated GPU only before moving on to switchable graphics.

I am using the proprietary nvidia driver, Wayland, KDE, and SDDM. Following the wiki instructions, I have enabled the DRM kernel mode setting, verified with

%> cat /sys/module/nvidia_drm/parameters/modeset
Y

I have also edited the files

/etc/X11/xorg.conf.d/10-nvidia-drm-outputclass.conf

and

/usr/share/sddm/scripts/Xsetup

to reflect the wiki. However, when I check which processor is being used as the renderer, glxinfo indicated that it is the Intel graphics:

%> glxinfo | grep NVIDIA
%> glxinfo | grep Intel
    Vendor: Intel (0x8086)
    Device: Mesa Intel(R) HD Graphics 530 (SKL GT2) (0x191b)
OpenGL vendor string: Intel
OpenGL renderer string: Mesa Intel(R) HD Graphics 530 (SKL GT2)

Usually the wiki instructions always do the trick for me, but something didn't work here. I've gone over the instructions some times now and I'm pretty sure I didn't miss anything, so I'm not sure why the computer will only use the Intel graphics. I would appreciate any help and guidance on why the instructions didn't work for me or where I went wrong.

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 25,076

Re: Asus GL552V laptop will not use nvidia 960M dGPU

Your xorg config and xsetup is irrelevant on wayland. Your computer will use the intel graphics by default because that's where the monitor is connected to -- and you should also not inherently want to change that (unless you have a switch in your UEFI to explicitly disable the integrated card, I doubt it though, that combination of HW normally didn't have that) -- performance of desktop based workloads is better and will consume less power (otherwise what has to happen is that the nvidia card renders the desktop and then needs to tell the intel GPU that it can now render something and copy the finished framebuffer over to the intel card for display... this is an immense overhead for tasks the integrated GPU is sufficiently fast for, it only makes sense to do this when the added power of the nvidia card is actually required).  Install nvidia-prime and use the prime-run command to move applications (games/blender and friends) to the dedicated card as the renderer on an as needed basis.

If despite that you really, really want your dGPU to be the renderer, there's a KWIN renderer selection environment variable .  KWIN_DRM_DEVICES you need to list both, the device corresponding to the nvidia card and the intel gpu, but list the nvidia card first. export hat somewhere that's sourced before you start Plasma/Kwin (e.g. /etc/profile for a global spot, or a drop-in service for plasma-kwin_wayland)

FWIW the Optimus article should probably either be extend by some known Wayland based approaches or probably warned about that most of the things mentioned there are not going to apply to wayland sessions and to check the corresponding compositors documentation/wiki pages.

Last edited by V1del (2024-06-06 10:14:22)