How do I know if I'm using the nvidia open griver or the proprietary ones?

I just did a zypper dup and during the install the installed the open drivers. I thought that I was using the proprietaries ones.

How do I tell?

$ nvidia-smi 
Mon Jul  7 07:01:21 2025       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.169                Driver Version: 570.169        CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 4060 ...    On  |   00000000:01:00.0  On |                  N/A |
| N/A   47C    P8              4W /   55W |     314MiB /   8188MiB |      6%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

and

$ inxi -Gxxx
Graphics:
  Device-1: Intel Meteor Lake-P [Intel Arc Graphics] vendor: Dell driver: N/A
    arch: Xe-LPG bus-ID: 00:02.0 chip-ID: 8086:7d55 class-ID: 0300
  Device-2: NVIDIA AD107M [GeForce RTX 4060 Max-Q / Mobile] vendor: Dell
    driver: nvidia v: 570.169 arch: Lovelace pcie: speed: 2.5 GT/s lanes: 8
    ports: active: none off: HDMI-A-1,eDP-1 empty: DP-1 bus-ID: 01:00.0
    chip-ID: 10de:28e0 class-ID: 0300
  Device-3: Realtek Integrated_Webcam_FHD driver: uvcvideo type: USB
    rev: 2.0 speed: 480 Mb/s lanes: 1 bus-ID: 3-9:7 chip-ID: 0bda:557c
    class-ID: fe01 serial: 200901010001
  Display: x11 server: X.org v: 1.21.1.15 compositor: xfwm4 v: 4.20.0
    driver: X: loaded: nvidia unloaded: modesetting,vesa
    alternate: fbdev,nouveau,nv gpu: nvidia,nvidia-nvswitch display-ID: :0.0
    screens: 1
  Screen-1: 0 s-res: 2560x1440 s-size: <missing: xdpyinfo>
  Monitor-1: not-matched mapped: DP-2 size-res: N/A modes: N/A
  Monitor-2: not-matched mapped: HDMI-0 pos: primary res: mode: 2560x1440
    hz: 60 scale: 100% (1) dpi: 109 size: 597x336mm (23.5x13.23")
    diag: 685mm (26.97") modes: N/A
  API: OpenGL v: 4.6.0 vendor: nvidia v: 570.169 glx-v: 1.4
    direct-render: yes renderer: NVIDIA GeForce RTX 4060 Laptop GPU/PCIe/SSE2
  Info: Tools: api: glxinfo de: xfce4-display-settings
    gpu: nvidia-settings,nvidia-smi x11: xprop,xrandr

with

 $ uname -a
Linux dodoite 6.12.35-1-longterm #1 SMP PREEMPT_DYNAMIC Mon Jun 30 08:45:40 UTC 2025 (393c607) x86_64 x86_64 x86_64 GNU/Linux

@elfroggio modinfo will show the following for the open driver;

/sbin/modinfo nvidia | grep license
license:        Dual MIT/GPL
1 Like

Is it possible to check what module is loaded currently?

Another way is to check for the kernel kmp package. You can only install the open driver or the closed one.
The output of
zypper se -si kmp-default and zypper se -si kmp-longterm
will tell.

nvidia-driver-G06-kmp-default → closed
nvidia-open-driver-G06-signed-kmp-default → open
nvidia-open-driver-G06-signed-kmp-longterm → open

I don’t have anything nvidia equipped sitting in front of me, or what the name of the “open” kernel module is, but lsmod | grep nvidia will tell you if the proprietary module is loaded or not, IIRC

zypper info --provides nvidia-open-driver-G06-signed-kmp-longterm
...
    kmod(nvidia.ko)
    kmod(nvidia_drm.ko)
    kmod(nvidia_modeset.ko)
    kmod(nvidia_uvm.ko)

yes, I’m on the open source:

 $ sudo /sbin/modinfo nvidia | grep license
license:        Dual MIT/GPL

and

$ zypper se -si kmp-default
Loading repository data...
Reading installed packages...

S  | Name                                      | Type    | Version                  | Arch   | Repository
---+-------------------------------------------+---------+--------------------------+--------+------------------
i  | nvidia-open-driver-G06-signed-kmp-default | package | 570.169_k6.15.3_2-1.4    | x86_64 | (System Packages)
i  | nvidia-open-driver-G06-signed-kmp-default | package | 570.153.02_k6.15.1_1-3.1 | x86_64 | (System Packages)
i  | nvidia-open-driver-G06-signed-kmp-default | package | 570.169_k6.15.4_1-2.1    | x86_64 | repo-oss

Which leads me to next question:

  • Is the open source or the proprietary faster than the other?

Thanks

Offhand, the packaging looks to me, like if a user has nvidia-open-driver-G06-signed-kmp-longterm installed, that the Conflicts: would prevent having the proprietary rpm installed at the same time, without doing some manual overrides:

[sfalken@mustang ~]$ zypper info --conflicts nvidia-open-driver-G06-signed-kmp-longterm

   
Information for package nvidia-open-driver-G06-signed-kmp-longterm:
...

Conflicts      : [5]
    nvidia-driver-G06-kmp
    nvidia-gfxG05-kmp
    nvidia-gfxG06-kmp
    nvidia-open-signed-kmp(cuda)
    nvidia-open-driver-G06-signed-kmp-longterm-k6.12.35_1

so checking what RPM is installed should be reasonable enough.

Now what happens with the “open” driver, if you install the proprietary driver through the .run I’m not sure, but I suspect you’d run into issues with it overwriting the modules as they’re named the same. But that’s completely a semi-educated guess on my part.

@elfroggio the open driver and GSP gets the most development AFAIK, no issues seen here, works fine for me with cuda?

@sfalken the run file by default installs the “open” driver on supported hardware. You have the set the option --kernel-module-type=proprietary with the run file to install the “closed” driver on open driver supported hardware. If it’s older than Turing, then it will install the closed driver without user intervention.

1 Like
 $ nvidia-smi
Mon Jul  7 11:18:22 2025       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.169                Driver Version: 570.169        CUDA Version: 12.8     |

and the cudas work, I use them “all the time” for both darktable and ollama.

Thanks

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.