Cuda installation issues on Ubuntu 22.04 laptop

Accelerated Computing CUDA CUDA Setup and Installation
1

I followed the instructions from the following link (I have Ubuntu 22.04 installed on a separate SSD on a Lenovo Legion5 pro with a mobile RTX 4070) :

Everything went fine, but the simple code example at the end didn’t work; the return said CPU only. I checked nvidia-smi and got an error (see below). Following is the output of a command line session. Any idea what I did wrong and what to do?

ken@ken-Legion-Pro-5:~$ sudo lspci | grep -i vga
01:00.0 VGA compatible controller: NVIDIA Corporation Device 2860 (rev a1)
ken@ken-Legion-Pro-5:~$ dkms status
nvidia/535.86.05, 6.2.0-26-generic, x86_64: installed

ken@ken-Legion-Pro-5:~$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Thu_Nov_18_09:45:30_PST_2021
Cuda compilation tools, release 11.5, V11.5.119
Build cuda_11.5.r11.5/compiler.30672275_0

ken@ken-Legion-Pro-5:~$ gcc --version
gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

ken@ken-Legion-Pro-5:~$ g++ --version
g++ (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

ken@ken-Legion-Pro-5:~$ dkms status
nvidia/535.86.05, 6.2.0-26-generic, x86_64: installed


ken@ken-Legion-Pro-5:~$ lspci -nnk | grep -i vga -A3
01:00.0 VGA compatible controller [0300]: NVIDIA Corporation Device [10de:2860] (rev a1)
	Subsystem: Lenovo Device [17aa:3b5a]
	Kernel modules: nvidiafb, nouveau, nvidia_drm, nvidia
01:00.1 Audio device [0403]: NVIDIA Corporation Device [10de:22bd] (rev a1)

ken@ken-Legion-Pro-5:~$ nvidia-smi
NVIDIA-SMI has failed because it couldnt communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running

Fyi, I also tried an Anaconda installation where I created a conda environment with Python 3.9 and installed Pytorch. A simple check of torch.cuda.is_available() failed. One thing that struck me is that the 4070 is listed as NVIDIA Corporation Device 2860, which seems strange.

2

Finally got it installed (cuda toolkit 12.2, driver 535.104.05) and working. The following was a huge help (especially for a novice):
https://www.youtube.com/watch?v=ttxtV966jyQ
Had to try it several times before getting it to work. I took the advice in the video to just reinstall Ubuntu each time rather than try to purge cuda. In the end, I was able to download, compile, and run the samples. I initially had trouble with samples that required graphics where I got the following error: “OpenGL GLX extension not supported by display”. This puzzled me for some time, but I eventually solved it by a bios setting (switching from hybrid graphics to discrete graphics). All good now.