Black screen after installing proprietary NVIDIA driver 340.108

English Hardware
1

Hello,

I have HP EliteBook 8740w with NVIDIA Quadro FX 2800M. I used to install gfx driver either with nvidia repository(gfx-G03) or with run file(340.108). But Currently G03 is not available in the repository and G04 and G05 are not suitable for this card.

Also installing Hard way leads to a black screen after reboot with the following error in
/var/log/Xorg.0.log


NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
NVIDIA: system's kernel log for additional messages and
.
.

It is a couple of months that I could not find any solution for the problem, and I decided to migrate to another distro or to downgrade(to 15.1 for instance).

Is it any chance to stay with 15.2 and install Quadro FX 2800M?[FONT=times new roman]

Thanx in advance.
[/FONT]

2

What does your kernel log say?

3

The 340 driver will not work without patching in Leap 15.2…
https://forums.opensuse.org/showthread.php/539260-Nvidia-G03-for-Leap-15-2

And the driver from Leap 15.1 is building against an older kernel 4.14 and not for 5.3…

4

Thank u Sauerland](https://forums.opensuse.org/member.php/56271-Sauerland)

Patching was designed for fedora, So I changed some steps(2.6.2) and skipped others (2.6.3 and 2.6.4).

But luckily it worked and I am doing 3D modeling now.

5

Thanx for patching I installed 340 driver, but currently I want to install cuda (version 6.5) which is compatible with 340 driver via .run file.

This version of cuda is compiled with kernel 3.11.6 which is way far from installed kernel 5.3.18. Also It was compiled with older gcc which gives me following error.


Error: Unsupported compiler: 7.5.0. Use --override......
Missing recommended Library: libGLU.so
.
.
.

Is there any patch for cuda or any other options for installing cuda (from source for example) !?

6

Hi
AFAIK, you should just be able to install cuda from the run file and skip the driver install from the run file. This is what I do with later versions…

7

That is exactly what I have done, but the problem is run file is compiled with different version of compiler.

Also There is not suitable cuda-6.5 for opensuse 15.2. In fact, the latest release of cuda-6.5 is install-able for opensuse 13.1.

I want to know that is there any way to compile cuda from the source?
or installing correct compiler to terminate following error?

8

Hi
Can’t run cuda 10? The driver and cuda releases are too old for newer support… :frowning:

9

Unfortunately no. Because my laptop has old graphic card (NVIDIA Quadro FX 2800M), Highest supported version for this card is cuda-6.5.

Also because of bad economic in the country, we cannot buy up-to-date gfx or new laptop. So I it would be very useful if I can use cuda with the existing device.

10

Hi
Then you would need to head off to the Nvidia and Cuda forums to see if someone has patches…

11

Kepler GT 710 is still supported.

You can attach card for a desktop to a laptop with a PCI-E 1x - to - PCI-E 16x or PCI-E 1x - to - PCI-E 1x cable.

12

I am not sure that will help the OP very much since it will still cost money to get a GT710 and adapters+desktop PSU for eGPU. I did fashion myself an external GPU using NVidia GTX750Ti and GTX580Ti with an old desktop PSU with these (Mini PCI-E / Expresscard V8.0 EXP GDC for Laptop External Video Card Mini Dock | eBay).

I’ve never tried offload or prime but it worked with bumblebee for sure, as long as I use the right BUSID configured for bumblebee.