-1

Problem

I have seen many similar questions in this forum, and I have tried many of the solutions proposed, however nothing seems to work :c

I was trying to install CUDA from the CUDA installation guide. This process asks to install the CUDA toolkit and during this process I had to update the nvidia drivers.

I tried updating my nvidia drivers from the website for my configuration. I have an NVIDIA RTX4000 ADA, and I'm using a lunix computer with Ubuntu 22.04. On the website it says that the latest driver is the nvidia driver-550. I dowloaded the file and tried running the installation from terminal by running: ./NVIDIA-Linux-x86_64-550.67.run. However there was a warning prompt saying that I should do the update from the ubuntu application "Software & updates". So I cancelled the updates and I went to this app and selected the drivers that said recommended. I reboot my pc and now whenever I try running nvidia-smi I get the following error:

NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

I tried reinstalling the drivers and I broke the setup even further or something because now when I open "Software & updates" I don't even get a recommended option...

I'm sorry I am not very tech savvy, should I restore my computer from factory settings or can this problem be fixed?

Debug output

sudo lshw -C display

*-display

   description: VGA compatible controller
   product: NVIDIA Corporation
   vendor: NVIDIA Corporation
   physical id: 0
   bus info: pci@0000:01:00.0
   logical name: /dev/fb0
   version: a1
   width: 64 bits
   clock: 33MHz
   capabilities: pm msi pciexpress vga_controller bus_master cap_list fb
   configuration: depth=32 latency=0 mode=1920x1200 visual=truecolor xres=1920 yres=1200
   resources: iomemory:600-5ff iomemory:640-63f memory:96000000-96ffffff memory:6000000000-63ffffffff memory:6400000000-6401ffffff ioport:3000(size=128) memory:97080000-970fffff

sudo dkms status
nvidia/550.54.14, 6.1.0-1036-oem, x86_64: installed

What I have tried (Update)

Tried reinstalling the drivers in many ways and even from a root shell ( Nvidia Driver Broken on Update - Unable to Reinstall ) . Tried sudo ubuntu-drivers autoinstall and sudo reboot.

I feel like I am trying random stuff at this point because I don't really know what is broken in the first place...

Jules_JJ
  • 11

2 Answers2

1

I have fixed my issue. I retried this sudo ubuntu-drivers autoinstall and follow the instructions on the screen. At some point, it asks you to create a password to ensure it is you (the user) making the installation.

The problem was actually coming from the MOK menu / secure boot that appears when you restart the computer. Instead of pressing continue, you have to go to the enroll MOK option. I created my password with an azerty keyboard but when I was trying to put the password on this menu, the keyboard acts as a qwerty keyboard...

So keep this in mind when running the updates!

Jules_JJ
  • 11
0

Had the same issue today with the 550.67 driver -- but on Pop OS 22.04. I had CUDA running beforehand, but wanted to update for getting TensorFlow to work. My card is a GTX 1050 ... I can run after cleanup 470 version. There no CUDA is provided. Others I did not try till now.

You can check your driver history like that to find out, which driver you had before:

#!/bin/bash
log_path="/var/log"
grep -h "nvidia-driver" $log_path/dpkg.log* | grep "install " | sort -u
Stephan Strauss
  • 1