Fails to start graphical mode after installing GeForce RTX 3050 Ti driver on OpenSUSE Linux

English Hardware
#9

Thanks @awerlang. that makes sense since there are two graphic cards installed, Intel integrated graphics and Nvidia GeForce RTX 3050 Ti.

The chapters you referred me to are very helpful. I need to read it more carefully to understand offloading graphics. So it’s possible to switch back and forth between graphic cards as the primary? if there are no entries in xorg.conf, then there is no description or configuration for the Nvidia card; how can this GPU be used if xorg.conf has no entries for it? If there is no explicit xorg.conf, it may be energy efficient but wouldn’t that be more demanding on other system resources (for example, the integrated Intel graphics card would use RAM whereas the Nvidia card would use its dedicated 4 GB of graphics memory)? Sorry for the questions and my confusion; this is my first time with a dedicated graphics card.

#10

Output:


$ xrandr --listproviders
Providers: number : 1
Provider 0: id: 0x46; cap: 0xf (Source Output, Sink Output, Source Offload, Sink Offload); crtcs: 4; outputs: 3; associated providers: 0; name: modesetting
    output eDP-1
    output DP-1
    output DP-2
#11

@dad4linux so not showing the Nvidia GPU, can you show the output from;

cat /proc/cmdline
#12

@dad4linux that’s not how offload works, not xorg.conf necessary, unless it is used as the primary graphics gpu (suse-prime, or if BIOS supports changing).

It could also be Leap 15.4 showing it’s age with respect to the release of your Ampere card…

#13

Output:

BOOT_IMAGE=/boot/vmlinuz-5.14.21-150400.24.46-default root=/dev/mapper/encSSD-encRoot splash=silent preempt=full mitigations=auto quiet security=apparmor

Question: both lspci and inxi show the Nvida card. Maybe xrandr doesn’t show Nvidia because xorg.conf is empty? How can you tell that xrandr --listproviders is not showing the Nvidia GPU? The output looks cryptic to me.

#14

@malcolmlewis
suse-prime SDB:NVIDIA SUSE Prime - openSUSE Wiki) looks like an awesome tool to easily switch between graphic cards. can I use this to activate my nvidia gpu when I like or is there still some config I’m missing? Is suse-prrime compatible with the official Nvidia drivers (no conflicts?

Thx for your help and patience.

Problemas con actualización del kernel en Leap 15.4
#15

@dad4linux I’ve run intel/nvidia before, my current setup is Nvidia/Nvidia (This is a Tumbleweed Workstation HP Z440) I only have an Xorg file for the primary GPU…

/sbin/lspci -nnk |grep -EA3 "VGA|Display|3D"

02:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU117GLM [Quadro T400 Mobile] [10de:1fb2] (rev a1)
	Subsystem: NVIDIA Corporation Device [10de:1489]
	Kernel driver in use: nvidia
	Kernel modules: nouveau, nvidia_drm, nvidia

03:00.0 3D controller [0302]: NVIDIA Corporation GP104GL [Tesla P4] [10de:1bb3] (rev a1)
	Subsystem: NVIDIA Corporation Device [10de:11d8]
	Kernel driver in use: nvidia
	Kernel modules: nouveau, nvidia_drm, nvidia

inxi -Gxxz

Graphics:
  Device-1: NVIDIA TU117GLM [Quadro T400 Mobile] driver: nvidia v: 525.89.02
    arch: Turing pcie: speed: 2.5 GT/s lanes: 16 ports: active: none
    off: DP-1,DP-2,DP-3 empty: none bus-ID: 02:00.0 chip-ID: 10de:1fb2
  Device-2: NVIDIA GP104GL [Tesla P4] driver: nvidia v: 525.89.02 pcie:
    speed: 2.5 GT/s lanes: 16 bus-ID: 03:00.0 chip-ID: 10de:1bb3
  Display: x11 server: X.Org v: 21.1.7 with: Xwayland v: 22.1.8
    compositor: gnome-shell v: 43.3 driver: X: loaded: nvidia
    gpu: nvidia,nvidia-nvswitch display-ID: :0 screens: 1
  Screen-1: 0 s-res: 3840x2160 s-dpi: 96
  Monitor-1: DP-1 pos: top-left res: 1920x1080 dpi: 94 diag: 598mm (23.53")
  Monitor-2: DP-3 pos: primary,bottom-c res: 1920x1080 dpi: 94
    diag: 598mm (23.53")
  Monitor-3: DP-5 pos: top-right res: 1920x1080 dpi: 94 diag: 598mm (23.53")
  Monitor-4: DVI-D-1-0 size-res: N/A
  API: OpenGL v: 4.6.0 NVIDIA 525.89.02 renderer: NVIDIA T400/PCIe/SSE2
    direct render: Yes

xrandr --listproviders 

Providers: number : 2
Provider 0: id: 0x217; cap: 0x1 (Source Output); crtcs: 4; outputs: 6; associated providers: 1; name: NVIDIA-0
    output DP-0
    output DP-1
    output DP-2
    output DP-3
    output DP-4
    output DP-5
Provider 1: id: 0x3d1; cap: 0x2 (Sink Output); crtcs: 4; outputs: 4; associated providers: 1; name: NVIDIA-G0
    output DVI-D-1-0
    output DVI-D-1-1
    output DVI-D-1-2
    output DVI-D-1-3

glxinfo | grep "OpenGL renderer"
OpenGL renderer string: NVIDIA T400/PCIe/SSE2

switcherooctl glxinfo | grep "OpenGL renderer"
OpenGL renderer string: Tesla P4/PCIe/SSE2
#16

@dad4linux it should work, what suse-prime does is create an xorg.conf file when you switch to nvidia, then removes it when you switch back to intel.

You need suse-prime and the bbswitch files as well… So you can have either intel, nvidia, or AFAIK it should work offload, when you try that you need to check xrandr outpout again, if you donot see both devices then it will not work, so can only switch between the two.

I suspect if you tried booting a Tumbleweed Live USB the results will be different…

#17

I installed suse-prime package from YaST and its dependencies and ran the command from the Terminal as root:

# prime-select boot nvidia

After rebooting, it appeared that my NVIDIA GPU was activated and in use:

$ glxinfo | grep 'OpenGL renderer string'
OpenGL renderer string: NVIDIA GeForce RTX 3050 Ti Laptop GPU/PCIe/SSE2

$ xrandr --listproviders
Providers: number : 2
Provider 0: id: 0x1f1; cap: 0x1 (Source Output); crtcs: 4; outputs: 1; associated providers: 1; name: NVIDIA-0
    output HDMI-0
Provider 1: id: 0x46; cap: 0xf (Source Output, Sink Output, Source Offload, Sink Offload); crtcs: 4; outputs: 3; associated providers: 1; name: modesetting
    output eDP-1-1
    output DP-1-1
    output DP-1-2

However, nvidia-smi and nvidia-settings seemed to be uninstalled. I also installed from YaST the package nvdock to see info on the GPU as it is being used. The program failed because it couldn’t find nvidia-settings. I reinstalled the downloaded drivers from NVIDIA’s website:

# ./NVIDIA-Linux-x86_64-525.89.02.run

The installer asked me if I wanted to install over the existing drivers, I clicked yes. The installation was successful and after rebooting, success! nvidia-smi and nvidia-settings were reinstalled. nvdock works and I see the metrics and stats on the GPU.

I tried reverting back to Intel integrated GPU by running:
# prime-select boot intel

However, X server didn’t start at all upon reboot and I was at a text-only prompt. Will try again and report but happy the NVIDIA GPU works! Thanks so much to @malcolmlewis and @awerlang for your help and suggestions.

#18

@dad4linux Hi, check after you switch back to intel if there is a xorg.conf file in /etc/X11/ or /etc/X11/xorg.conf.d and in that file if it refers to the nvidia device. If it is, copy it out of that location and reboot and it should be running the intel device. I would try setting suse-prime to use offload mode and see if that works.

I suspect if you switch back to nvidia, it may start working as the run file may have created it’s own file, AFAIK suse-prime creates/deletes as required.

If you switch back to the Nvidia rpms, you may need to ensure the nvidia-utils package is installed, this should have nvidia-smi etc present.

#19

There was no xorg.conf file in either /etc/X11/ or /etc/X11/xorg.conf.d/. I noticed that after running prime-select, a file (either 90-nvidia.conf or 90-intel.conf was generated in /etc/X11/xorg.conf.d/. Running prime-select boot intel or prime-select boot offload didn’t work; X didn’t start and was dropped to text login screen. In fact, noticed that network didn’t even start - there was no internet after logging in.

After running prime-select boot intel, this is what /etc/X11/xorg.conf.d/90-intel.conf contained (there was no 90-nvidia.conf):

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

Section "Device"
    Identifier "intel"
    Driver "modesetting"
    BusID "PCI:0:0:2"
EndSection

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

After running prime-select boot offload, this is what /etc/X11/xorg.conf.d/90-intel.conf contained (there was no 90-nvidia.conf):

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

Section "Device"
    Identifier "intel"
    Driver "modesetting"
    BusID "PCI:0:0:2"
EndSection

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

# needed for NVIDIA PRIME Render Offload
Section "Device"
  Identifier "nvidia"
  Driver "nvidia"
  BusID "PCI:1:0:0"
EndSection

After running prime-select boot offload, this is what /etc/X11/xorg.conf.d/90-nvidia.conf contained (there was no 90-intel.conf):

Section "ServerLayout"
    Identifier "layout"
    Screen "nvidia"
    Inactive "intel"
    Inactive "modesetting"
EndSection

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

Section "Device"
    Identifier "intel"
    Driver "intel"
EndSection

Section "Screen"
    Identifier "modesetting"
    Device "modesetting"
EndSection

Section "Device"
    Identifier "modesetting"
    Driver "modesetting"
EndSection

Section "Screen"
    Identifier "nvidia"
    Device "nvidia"
EndSection

Section "Device"
    Identifier "nvidia"
    Driver "nvidia"
    BusID "PCI:1:0:0"
    Option "DPI" "96 x 96"
    Option "AllowEmptyInitialConfiguration"
    Option "AllowExternalGpus"
EndSection

Now it appears only the NVIDIA GPU works.

#20

@dad4linux I would suggest uninstalling the run file and install the Nvidia rpms, suse-prime and the bbswitch stuff with the -f (force) option and see how it goes. I would also install the nvidia-utils package too.

#21

I think you had a typo in the last command, I guess you meant prime-select boot nvidia. The output and outcome would make more sense.

One thing I’d try is removing the PCI: from the BusID so it becomes, for Intel:

    BusID "0:0:2"

Also, let’s double check the PCI identifiers again with different options:

sudo /sbin/lspci -nnD | grep -E "VGA|Display|3D"

If booting leaves you at text mode, inspect the content of /var/log/Xorg.0.log, in particular lines with (EE). You might paste a link to it with:

cat /var/log/Xorg.0.log | susepaste
#22

I don’t understand why you use the Nvidia driver from Nvidia.
Add the NVIDIA repo with Yast software manager.
Restart the laptop
Go to yast software manager. You see that the required Nvidia packages and prime packages are automatically selected. Just accept the selection to install the driver and restart.

That’s all.

For your RTX card i assume G06 packages are installed.

In case you need to do a manual installation of the driver and Prime there is a new meta package for a full installation “nvidia-drivers-G06”. I did not test this.

#23

After an kernel update, the Nvidia dock (nvdock) and nvidia-settings (which is needed to run nvdock) no longer start. I didn’t know how to uninstall the drivers I downloaded from Nvidia and manually installed. I think some update went awry because some strange behaviour for certain packages started happening after the update. Reinstalled and did as you suggested - only using the provided Nvidia software and drivers from the openSUSE repos. I also followed the directions: SDB:NVIDIA drivers - openSUSE Wiki.

Found this method the easiest and straight forward and tailored to openSUSE.

I did not install the drivers directly from Nvidia but enabled the Nvidia repositories in YaST’s Software Management: Configuration > Repositories > Add button > Community Repositories > check / enable the Nvidia respositories. Nvidia packages appeared already selected (*-G06). After installing the selected packages and disabling the nouveau driver, I rebooted but the Nvidia dock didn’t work; nvidia-settings still wasn’t installed. Went back to YaST Software Management and installed nvidia-utils-G06. Rebooted and nvidia-settings and nvdock worked and was able to see the resources used by the GPU. Thank you everyone for your help.

Earlier in this thread, it was mentioned that I could switch between the Intel integrated GPU and the Nvidia GPU. How does one do this? And how can I tell which GPU is being currently used? When I run:

# /sbin/lspci -nnk | grep -EA3 "VGA|Display"
0000:00:02.0 VGA compatible controller [0300]: Intel Corporation TigerLake-H GT1 [UHD Graphics] [8086:9a60] (rev 01)
        Subsystem: Acer Incorporated [ALI] Device [1025:1539]
        Kernel driver in use: i915
        Kernel modules: i915
--
0000:01:00.0 VGA compatible controller [0300]: NVIDIA Corporation GA107M [GeForce RTX 3050 Ti Mobile] [10de:25a0] (rev a1)
        Subsystem: Acer Incorporated [ALI] Device [1025:153a]
        Kernel driver in use: nvidia
        Kernel modules: nouveau, nvidia_drm, nvidia

So from this output, it is reported that both Kernel drivers are in use but does this mean that both GPUs are active? How is that possible? Is it possible to select one GPU at a time; if so how?
Thanks again.

#24

You don’t need to disable nouveau driver. When installing NVIDIA driver then automatically NVIDIA driver replaces nouveau driver.

If you have a laptop, there is a technology based on hardware and software to save energy. In case of apps don’t need graphical power, the laptop uses the Intel iGPU. In case of apps need graphical power, the laptop uses the Nvidia dGPU.
This tech does not exist for desktop.
For openSUSE, the software part is named “Prime”. If you get a laptop when installing the Nvidia driver then the following Prime packages are installed ;

plasma5-applet-suse-prime - Plasma 5 applet for controlling SUSE Prime

suse-prime - GPU (nvidia/intel) selection for NVIDIA optimus laptops with bbswitch support

#25

Best information for graphics status I find comes from inxi -Ga. inxi -Gaz does a little filtering of possibly sensitive host information. Inxi in Leap is antique and broken, particularly -G*. Run inxi -U to update it to current version from upstream, which includes many changes to -G.

#26

@promeneur with some newer ‘optimus’ type laptops and powered with the RTX series cards suse-prime is deprecated in favor of prime render offload depending on the hardware setup.

It does exist for desktops, I use prime render offload here with a secondary Nvidia card, my primary GPU (T400) running at full power max is only 30W, offload gpu (Tesla P4, around 50-75w at full power).

I ran this setup with the primary card which was intel as well…

I use switcherooctl (dbus service) to offload, it is integrated with GNOME to add a right-click option to application desktop files to launch with the ‘discrete’ gpu this also works on my HP Dual AMD Polaris GPU powered laptop :wink:

#27

If I change graphics card from “nvidia” to “intel” (or even “intel2”) and reboot, X doesn’t start and I’m dropped down to a text login prompt.

# prime-select boot intel

HDMI (video) works only with "nvidia"selected and graphical boots normally. I’d like to switch between intel and nvdia graphics but I’m ok if only can use nvidia that except that the hdmi sound doesn’t work. Not sure how to troubleshoot or the solutions. I’m hesitant to try Tumbleweed because in my experience with it I found sooner or later an update causes some packages to break. I chose Leap because it’s a bit more stable and updates aren’t so big and would wear less on my nvme ssd drive.

Does anyone have any such trouble with dual graphic cards? Anyone with dual graphic cards have any success?

#28

Hi,
I have no dual graphic card but I am using nvidia rtx 3050 also.
I am using the .run installer with no problem.

Edit:
With the .run driver installer sometimes during a kernel update and you reinstall the said driver and it failed to reach the graphical target, issuing mkinitrd will clear the problem and reach the gui again.
Note: mkinitrd is gone in tumbleweed and you need to use dracut.