Memory is allocated on all GPUs

Graphics / Linux Linux
1

When an application is started, the memory is allocated on all GPUs. This affects both the X11 session and Cuda applications. I am not sure by which setting this is caused. It also does not occur in the multi-user target, so I suspect it has to do with the X11 session.

System
2x NVIDIA Quadro RTX 4000
Fedora 33 (5.12.12-200.fc33.x86_64)
CUDA Version: 11.3
Driver Version: 465.19.01

Tried settings
/etc/X11/xorg.conf

...
Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BusID          "PCI:101:0:0"
EndSection

Section "Screen"
    Identifier     "Screen0"
    Device         "Device0"
    Monitor        "Monitor0"
    DefaultDepth    24
    Option         "MultiGPU" "off"
    Option         "SLI" "off"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

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

Section "OutputClass"
    Identifier "nvidia"
    MatchDriver "nvidia-drm"
    Driver "nvidia"
    Option "AllowEmptyInitialConfiguration"
    Option "SLI" "off"
    Option "BaseMosaic" "on"
    Option "PrimaryGPU" "yes"
    Option "MultiGPU" "off"
EndSection

Section "ServerLayout"
    Identifier "layout"
    Option "AllowNVIDIAGPUScreens"
EndSection

nvidia-smi
Bildschirmfoto von 2021-07-08 13-04-21

2

You have basemosaic enabled so this is expected.

1 Like
3

this is the solution, I added:

Option         "BaseMosaic"   "FALSE"

to /etc/X11/xorg.conf

4

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.