I have two cards, a primary one, used for the desktop, and a secondary one, used for other jobs; both use the Nvidia proprietary driver (“nvidia” module), from the Nvidia PPA.
I need the second card not to be controlled by Xorg, however, I can’t find a way to accomplish this.
This is my Xorg configuration:
Section "ServerFlags"
Option "AutoAddGPU" "false"
EndSection
Section "Device"
Identifier "Device0"
Driver "nvidia"
VendorName "NVIDIA Corporation"
BusId [...] # bus id of primary GPU
EndSection
Section "Screen"
Identifier "Screen1"
Device "Device0"
Monitor "Monitor1"
DefaultDepth 24
Option "Stereo" "0"
SubSection "Display"
Depth 24
EndSubSection
EndSection
In theory, the AutoAddGPU
option should accomplish this; in practice though, Xorg still controls the card: I can’t unbind the nvidia
module until X is shutdown, but also, I can see from the Xorg log:
# X boot (extract):
[ 15.516] (II) NVIDIA(GPU-1): NVIDIA GPU NVIDIA GeForce RTX [...]
[ 15.516] (II) NVIDIA(GPU-1): PCI:[...] (GPU-1)
[ 15.516] (--) NVIDIA(GPU-1): Memory: [...]
[ 15.516] (--) NVIDIA(GPU-1): VideoBIOS: [...]
[ 15.516] (II) NVIDIA(GPU-1): Detected PCI Express Link width: 16X
# X shutdown (extract):
[ 29.088] (II) NVIDIA(GPU-0): Deleting GPU-0
[ 29.088] (II) NVIDIA(GPU-1): Renaming GPU-1 to GPU-0
[ 29.088] (II) NVIDIA(GPU-0): Deleting GPU-0
If I use a stub driver for the card, Xorg does not take control (there is no GPU-1 mention in the log), however, I wanted to know if it’s possible to still retain the nvidia
driver, but have Xorg not interact with it at all.