Problem Installing Drivers on Ubuntu 20.04 using: nvidia-driver-455, on Lenovo T490 with MX250 dGPU


I’m a computer science student trying to get nvidia drivers working for machine learning.

My machine is a Lenovo T490 laptop with a MX250 dGPU, but I think it defaults to using the intel integrated graphics to run my desktop windows. It’s a dual boot machine with windows on its own partition, but I am specifically trying to install the cuda and nvidia drivers for my MX250 on my Ubuntu 20.04 system. My machine is 64 bit.

My benchmark for if my install has worked is running nvidia-smi and hoping I don’t get the below error message:
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Tragically, I have tried many different ways of installing nvidia drivers and the cuda toolkit, and I’m afraid my system might be hopelessly tied up with conflicting installs. I have tried using a run file for Linux-x86_64, 450.80.02, but that install complained about nvidia-drm kernel not being able to be found. I have also tried various cuda installs but those also did not seem to work. I also recently upgraded from 18.04 to 20.04 (today) to try and help solve my install problems but that did not seem to make a difference.

My latest attempt was using:
sudo apt-get install nvidia-driver-455

As far as I can tell, I don’t think I have any bad files in /etc/modprobe.d

I could not upload the .gz file, so I just uploaded the entire thing. I hope that’s ok
nvidia-bug-report.log (1.8 MB)

Please let me know if there’s any further information I can provide to help debug my setup.


  • Don’t use the .run installers, use --uninstall to uninstall them
  • purge anything nvidia/cuda
  • add the ubuntu graphics ppa
  • install the driver from that (sudo apt install nvidia-driver-455)
  • download the cuda .deb
  • add the repo to your system
  • don’t install cuda
  • instead, run sudo apt install cuda-toolkit-11-1
  • set PATH variable if necessary
  • have nvidia-persistenced start on boot and make sure it is continuously running

Hi generix,

Thanks for getting back to me. I have run up to and included sudo apt-get install cuda-toolkit-11-1

I’m not sure which PATH variable to set, so I did not set anything.

When I run nvidia-smi I still get:
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Trying to start nvidia-persistenced yields this. I am not sure what the best way to start this on boot would be.
nvidia-persistenced failed to initialize. Check syslog for more details.

I have attached my bug report here:
nvidia-bug-report.log (1.7 MB)

I can’t see any messages of a recent kernel driver install, could you do a clean Ubuntu reinstall?
Afterwards, just install the driver using the Software&Update applicaton and check if the nvidia gpu is functional.
To install cuda then,
downlaod cuda from
and do all the steps from the install instructions but omit the last one:
Don’t install ‘cuda’
but run
sudo apt install cuda-toolkit-11-2
I’m sorry for the previous instruction being incomplete, cuda 11.2 now being released and the install intructions on the downloads page are changing with every release.

Hi generix,

I was finally able to resolve my issue without re-installing Ubuntu!

I installed nvidia-driver-460 thru:
sudo ubuntu-drivers autoinstall

Then, since nvidia-persistenced was unable to start (or nvidia-smi), I found this page that had similar behavior:

After clearing the blacklists per the above link, and:
$ sudo update-initramfs -u

And then rebooted, nvidia-persistenced was able to start as I saw it in htop, and nvidia-smi worked properly. My system already has secure boot disabled.

Then I installed cuda-toolkit-11-2 per your listed instructions, using only
sudo apt install cuda-toolkit-11-2

But nvcc still didn’t work, so then I ran:
sudo apt-get install nvidia-cuda-toolkit

And then nvcc did seem to work.

So now I believe my driver issues are sorted out! Thanks for your help! The MX250 gpu also seems to work well in tensorflow!