Geforce RTX 3060

English Install/Boot/Login
1

Hi!

Got an RTX 3060 for my TW and now I wonder which repos/drivers I need.

I currently have an NVS 310, which I would like to keep (kernel-firmware-nvidia and libdrm-nouveau2 installed, according to YasT) for video output. The Geforce is meant for computing stuff.

Is it possible to have these two cards at the same time?

Would that work:

# zypper install-new-recommends --repo NVIDIA

as recommended here

https://en.opensuse.org/SDB:NVIDIA_drivers

?

Or is “the hard way” more adequate?

Many thanks in advance…

2

@suse_rasputin to run the RTX you need to blacklist nouveau, maybe can get it working with the other driver @mrmazda may offer advise on this.

I would suggest working on getting that functional first, then look at the repo driver or the open driver.

3

Thanks for reply, so which work flow do you recommend? I installed the NVIDIA repo, should I undo that? Where to start?

4

@suse_rasputin Get the current card working without nouveau.

I see you have a post from a few years ago… https://forums.opensuse.org/t/nvidia-quadro-k2200-which-graphics-driver/147203/14

See this https://forums.opensuse.org/t/amd-intel-nvidia-x-graphics-driver-primer-third-edition/148576

5

The NVS 310 needs the G04 driver. The RTX 3060 will use G06. Can that even work at the same time?

The SDB page mentiones:

“If you are going to install CUDA later, then you must use nvidiaG05 version for compatibility (as of Jan 18, 2022).”

Can that even work with G04 installed?

Should I uninstall nouveau? Install G06? Then shutdown, remove the NVS 310 and install the RTX 3060?

6

@suse_rasputin that’s all up to you… for simplicity I would suggest just the one gpu. Set the default to multi-user, shutdown, install new gpu, boot, install the rpms, set default to graphical target and reboot…

7

I doubt you can install two different driver at the same time. I used to use two nvidias but both using G04. When I upgraded to RTX I had to remove the old gforce. Your new card is 300% more powerful why need a secondary old graphic card. Just my opinion.

8

@conram or use the old card as passthrough (vfio-pci) to a vm…

pinxi -Gxz

Graphics:
  Device-1: NVIDIA GP104GL [Tesla P4] driver: nvidia v: 550.54.14 arch: Pascal
    bus-ID: 02:00.0
  Device-2: NVIDIA TU117GLM [Quadro T400 Mobile] driver: nvidia v: 550.54.14
    arch: Turing bus-ID: 03:00.0
  Device-3: NVIDIA GK208B [GeForce GT 710] vendor: ZOTAC driver: vfio-pci
    v: N/A arch: Fermi 2 bus-ID: 05:00.0
  Display: x11 server: X.Org v: 21.1.11 with: Xwayland v: 23.2.4 driver: X:
    loaded: nvidia gpu: nvidia,nvidia-nvswitch resolution: 1: 1920x1080~60Hz
    2: 1920x1080~60Hz 3: 1920x1080~60Hz
  API: EGL v: 1.5 drivers: nvidia,swrast platforms:
    active: gbm,x11,surfaceless,device inactive: wayland,device-2,device-3
  API: OpenGL v: 4.6.0 compat-v: 4.5 vendor: nvidia mesa v: 550.54.14
    glx-v: 1.4 direct-render: yes renderer: NVIDIA T400/PCIe/SSE2
  API: Vulkan v: 1.3.275 drivers: N/A surfaces: xcb,xlib devices: 2
9

YaY! I will need a “tuts” for that. That’s kind of nice I still have those old one lurking in my drawer. :grinning:

10

@conram it’s a simple process, when your motivated, start a new thread…

1 Like
11

OK, I remove the NVS 310. What is the CLI procedure for installing the g06 driver?

I guess this here: SDB:NVIDIA drivers - openSUSE Wiki

There is no alternative to the NVIDIA repo?

There will be trouble with the driver from NVIDIA each time the kernel in TW will get an update, correct?

12

@suse_rasputin not in my testing, new kernel, nvidia module rebuilds on zypper dup…

Well you can always use the open driver…
https://sndirsch.github.io/nvidia/2022/06/07/nvidia-opengpu.html

13

This is FUD which is unfortunately spread by non Nvidia users and ppl which can‘t follow simple instructions.

With modern cards like yours which use the G06 and G05 drivers, at each kernel/driver update, everything is rebuild on the users machine automatically.

Issues may arise when you use the legacy drivers G04, G03 and hardware older than 10 years. These drivers need heavy patching to be compatible with actual kernels.

14

@suse_rasputin what @hui said… :wink:

On my primary desktop, I use the hardway because of cuda, cudnn, nccl etc and wanting the latest drivers…

pinxi -Gxz

Graphics:
  Device-1: NVIDIA GP104GL [Tesla P4] driver: nvidia v: 550.54.14 arch: Pascal
    bus-ID: 02:00.0
  Device-2: NVIDIA TU117GLM [Quadro T400 Mobile] driver: nvidia v: 550.54.14
    arch: Turing bus-ID: 03:00.0
  Device-3: NVIDIA GK208B [GeForce GT 710] vendor: ZOTAC driver: vfio-pci
    v: N/A arch: Fermi 2 bus-ID: 05:00.0
  Display: x11 server: X.Org v: 21.1.11 with: Xwayland v: 23.2.4 driver: X:
    loaded: nvidia gpu: nvidia,nvidia-nvswitch resolution: 1: 1920x1080~60Hz
    2: 1920x1080~60Hz 3: 1920x1080~60Hz
  API: EGL v: 1.5 drivers: nvidia,swrast platforms:
    active: gbm,x11,surfaceless,device inactive: wayland,device-2,device-3
  API: OpenGL v: 4.6.0 compat-v: 4.5 vendor: nvidia mesa v: 550.54.14
    glx-v: 1.4 direct-render: yes renderer: NVIDIA T400/PCIe/SSE2
  API: Vulkan v: 1.3.275 drivers: N/A surfaces: xcb,xlib devices: 2

On my test system, I used the K620, but it runs offload now as have an Intel Arc board installed for oneapi. I have not had issues on that system with the G06 driver.

pinxi -Gxz

Graphics:
  Device-1: Intel Xeon E3-1200 v2/3rd Gen Core processor Graphics driver: i915
    v: kernel arch: Gen-7 bus-ID: 00:02.0
  Device-2: Intel DG2 [Arc A380] vendor: ASRock driver: i915 v: kernel
    arch: Gen-12.7 bus-ID: 03:00.0
  Device-3: NVIDIA GM107GL [Quadro K620] vendor: Hewlett-Packard
    driver: nvidia v: 550.54.14 arch: Maxwell bus-ID: 05:00.0
  Display: x11 server: X.Org v: 21.1.11 with: Xwayland v: 23.2.4 driver: X:
    loaded: modesetting,nvidia unloaded: fbdev,vesa dri: crocus gpu: i915
    resolution: 1920x1080~60Hz
  API: EGL v: 1.5 drivers: crocus,iris,nvidia,swrast platforms:
    active: gbm,x11,surfaceless,device inactive: wayland,device-3
  API: OpenGL v: 4.6.0 compat-v: 4.2 vendor: intel mesa v: 23.3.6 glx-v: 1.4
    direct-render: yes renderer: Mesa Intel HD Graphics P4000 (IVB GT2)
15

Hi malcolm, what is the advantage of “hardway” regarding CUDA (which I will have to install here, too…)?

What is the procedure of updating g06 after “hardway” install after kernel update?

16

@suse_rasputin Just some finer grained control, keep ahead of the rpm version and just because you can… It is a few extra minutes and an extra reboot… :wink:

If using cuda, then you want that download as can also install the driver as part of that download.

17

So with hardway you see “new kernel available” and interrupt zypper dup and upgrade the driver, before trying zypper dup again?

This here

does not list Tumbleweed as compatible, so no CUDA on TW?

And one moe question regarding

https://en.opensuse.org/SDB:NVIDIA_drivers

What are the pros and cons of using the opensuse NVIDIA repos vs. using NVIDIA repos on Tumbleweed install?

18

There is no “opensuse NVIDIA repos”. NVIDIA drivers are built on internal SUSE servers and submitted to NVIDIA which publishes them. There is only one repository which is provided by NVIDIA.

19

Here

https://en.opensuse.org/SDB:NVIDIA_drivers

I read:

openSUSE-repos

We've recently added NVIDIA repository under the new RIS management which is already used by default on MicroOS as well as openSUSE Leap Micro.

You'll be asked whether you agree to import the 3rd party NVIDIA gpg key on the first refresh. We currently don't have a good way how to "trust" the thirdparty key and avoid this step. So please confirm the import of the key with yes.

Leap and Tumbleweed

As a root, enter the following in a terminal:

# zypper install openSUSE-repos-NVIDIA
20

You need to read completely:

Add the Nvidia Repository
The NVIDIA drivers can not be included with openSUSE because of their license. Conveniently, NVIDIA has an openSUSE repository that can be added and downloaded from.