Is this new? Last time I tried, Nvidia on Wayland had performance issues and did not support Optimus at all (Edit: after rereading what you said I am guessing that is still an issue.)

If it is working though I retract what I said, though it’s absurd how long it took to happen.

Even without wayland, in recent Fedora releases (I forget if it started with 28 or 29), I've found optimus switchable graphics to be impossible or at least impractical to configure. What's become easier is having the entire X display use the nvidia gpu, with the intel gpu transparently slaved to relay pixels to its connected laptop display.

But, it does return the annoying, periodic driver breakage, where a dnf update replaces the xorg-x11-drv-nvidia RPMs and suddenly the userspace is incompatible with the still-running kernel module's slightly older API, so new processes cannot use opengl until I reboot. This was one benefit of optimus via the optirun infrastructure. The main desktop was on the intel gpu and I could actually unload and upgrade the nvidia module if desired, without forcing a disruptive reboot.

Do you have any documentation, or even random notes, on how you set that up? I have an optimus laptop and couldn't make the proprietary drivers work ( tried both rpmfusion and negativo17). I m happily using nouveau for now but at some point I ll need to use CUDA again.

I have an older Thinkpad T440p with Geforce GT 730M running Fedora 29. I use the MATE desktop rather than GNOME.

I just installed the akmod-nvidia packages from rpmfusion which also pulls in xorg-x11-drv-nvidia and xorg-x11-drv-nvidia-cuda. Search for "PRIME" discussions in the README.txt under /usr/share/doc/xorg-x11-drv-nvidia. I cannot be sure every bit of my config is still necessary, as I left it alone once I got it working.

First, I have /etc/X11/xorg.conf.d/nvidia-prime.conf with a small amount of config:

  Section "Module"
    Load "modesetting"
  EndSection

  Section "Device"
    Identifier "nvidia"
    Driver "nvidia"
    Option "AllowEmptyInitialConfiguration"
  EndSection

Then, I created a custom script in /etc/lightdm/display_setup.sh:

  #!/bin/sh
  
  if rpm -q xorg-x11-drv-nvidia \
     && lsmod | grep '^nvidia' '
  then
    xrandr --setprovideroutputsource modesetting NVIDIA-0
    xrandr --auto
    xrandr --output eDP-1-1 --set "PRIME Synchronization" 1
  fi

And in my /etc/lightdm/lightdm.conf I added one line to call that script:

  ...
  [Seat:*]
  ...
  display-setup-script=/etc/lightdm/display_setup.sh

To be honest, my old GPU is too slow and with too small RAM to be worthwhile for OpenCL. I find it more practical to just use the Intel OpenCL runtime for multi-core CPUs on this old quad-core laptop or on newer workstations. I do get some use out of xorg-x11-drv-nvidia-cuda on a Titan X desktop GPU though.

Thanks a lot!

I 've already created a thread about my problems in askfedora, made a last post there now linking to your post here in case they help somebody else. https://ask.fedoraproject.org/t/black-screen-after-installin...

I 'll try your notes when I have enough free time to re-install if things go bad.

Why is your standard for a Linux desktop renderer whether a pretty Linux-hostile company's niche laptop energy saving hybrid GPU driver works well on it?