Converting a Fedora Server install into a Workstation

One of out machines was originally configured without a GUI (since it was intended to purely run at GPU card in GPGPU mode, accessed via ssh).

However, now it’s going to be slightly repurposed, so that a new developer can also use the machine for editing, etc. However, since the GPU is still to be used for GPGPU activities, the GUI X-session will be connected to the onboard intel graphics, with the GPU card left free of graphics duties.

Adding required Groups and RPMs

dnf group install xfce
dnf install xorg-x11-drv-evdev lightdm

Tell systemd about the new target

systemctl set-default graphical.target
systemctl enable lightdm

Remove ‘Nvidia cruft’ from X11 install

The proprietary Nvidia driver makes strong assumptions that it is going to be the principal graphics card in a system, and so adds a lot of Nvidia-specific details to /etc/X11/* without asking. These can be safely commented out.

ls -l /etc/X11/xorg.conf
ls -l /etc/X11/xorg.conf.d/

Make X11 only want the intel card

The following idea should be represented in /etc/X11/xorg.conf :

Section "ServerLayout"
    Identifier "layout"
    Screen 0 "intel"
    #Inactive "nvidia"
EndSection

#Section "Screen"
#    Identifier "nvidia"
#    Device "nvidia"
#    Option "AllowEmptyInitialConfiguration" "Yes"
#EndSection

Section "Screen"
    Identifier "intel"
    Device "intel"
EndSection

#Section "Device"
#    Identifier "nvidia"
#    Driver "nvidia"
#    BusID "PCI:1:0:0"
#EndSection

Section "Device"
    Identifier "intel"
    Driver "modesetting"
    #Option "AccelMethod"  "none"
    #BusID "PCI:00:02.0"
EndSection

Finally…

reboot

and if it’s not working yet…

more /var/log/Xorg.0.log


Martin Andrews

{Finance, Software, AI} entrepreneur, living in Singapore with my family.



blog comments powered by Disqus