Can't install anything but cuda 10.1 with driver version 430

Accelerated Computing CUDA CUDA Setup and Installation
1

I’ve been trying to roll back to cuda 10.0 with no success. I previously had Cuda 10.1 on my system, which had been installed using a runfile. I uninstalled it using the uninstall script. However, now whenever I try to install a different version of CUDA, I get Failed to initialize NVML: Driver/library version mismatch.

I have tried completely removing all versions of cuda and nvidia drivers using

sudo apt --purge remove nvidia-*
sudo apt --purge remove cuda*

Then I reinstalled just the driver to see if I could get nvidia-smi to work. However, if I install 410 or 418, I get the same error message. nvidia-smi only works if I install 430.

All I can think is that there are some files that were put in place by the runfile that never got removed and that’s corrupting my installation. I have been using .deb files to install cuda 10.0 and the ubuntu repository for the drivers.

I’m running ubuntu 18.04 and I have two rtx 2080ti.

2

Just keep your 430 driver that is working.

You can use CUDA 10.0 with it. Use the runfile installer method and select “no” when prompted to install the bundled driver - keep the 430 driver intact. It will work with CUDA 10.0

Also, the 430 driver will show a “CUDA Version” of 10.1 even if you install CUDA 10.0. That is OK.

3

If I try to do that, I get the message Failed to initialize NVML: Driver/library version mismatch

4

You’re doing it wrong, then.

Get your 430 driver working again. Make sure it works correctly after a reboot as well.

Then download the runfile installer for CUDA 10. When prompted to install the driver, select “no”. It will not disrupt your driver.

5

Thanks!! Works when I use the runfile. Does this mean I won’t be able to upgrade from 430?

6

You can upgrade from 430 to any driver you want. Any newer driver will also work with CUDA 10.

7

Hi, I followed the above steps of changing it to 430 drivers and installing Cuda 10.0 through .run file.

After installation of Cuda , Nvidia-smi is working fine, nvcc --version gives the Cuda version but when do device query its stating "./deviceQuery Starting…

CUDA Device Query (Runtime API) version (CUDART static linking) "

after this nothing happens, it returns nothing, terminal get stuck. Running docker images trough Nvidia-docker also have the issue

I’m running ubuntu 18.04 and have one V100