Installation error with Quadro T2000 and CUDA10.1, Nvidia driver 418.87

Hi, I am Hotae. I use Quadro T2000 GPU on my laptop and wanna use CUDA. So, I installed CUDA 10.1 and Nvidia driver version 418.87 on my 16.04 Ubuntu.
But, when I tried a deviceQuery test after I installed it, it told me that test failed. I already checked every compatibility, but there is few Quadro T2000’s data on internet. So, please check CUDA10.1 and 418.87 driver are compatible with Quadro T2000.
Moreover, my ultimate goal is to use nividia/cudagl on docker. It also doesn’t work now and outputs the error, ‘docker: Error response from daemon: OCI runtime create failed: container_linux.go:346: starting container process caused “process_linux.go:449: container init caused “process_linux.go:432: running prestart hook 0 caused \“error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: cuda error: unknown error\\n\”””: unknown.’

I guess this problem is also from CUDA and my GPU’s mismatch / compatibility.
So, please check this because Quadro T2000’s official data is insufficient now.
Thank you so much.

That GPU is very new. When you have a GPU that just came out yesterday, there’s a possibility that even the latest CUDA toolkit doesn’t support it. The issue here is not general CUDA support, but instead a need for the latest driver. In these situations, the process to use is as follows.

  1. Go to https://www.nvidia.com/drivers and download the latest driver for your GPU, and install it.
  2. Download the CUDA toolkit installer of choice (for a Turing GPU such as this, I would recommend 10.0 or later).
  3. Read the CUDA linux install guide. Follow the instructions there with the following modification:
  • For a package manager install, install the cuda-toolkit-10-1 package instead of the cuda package (or change 10-1 to match the version of CUDA you want to install)
  • For a runfile install, deselect the option to install the driver.

With either of these methods, the driver (e.g. 418.87) in the CUDA toolkit installer will not be installed, and your existing driver (which might be for example 430.50 https://www.nvidia.com/Download/driverResults.aspx/151568/en-us ) would be left intact.

On windows, generally your laptop should have shipped with an appropriate windows driver for your GPU. Use the windows CUDA installer, but deselect the option to install the driver bundled with the installer. On a new laptop like this, the existing driver should be compatible with any CUDA version.