Which driver the GPU will call first & Hows the compatibility of the latest CUDA package with the old graphics card

Question 1:When we install the graphics card driver for a certain GPU and the CUDA package at the same time, which driver the graphics card will use first?
Question 2:I am using the graphics card driver in CUDA 11.1 (my GPU is GTX950), but the graphics card often has problems such as disconnection and rendering failure when running 3D graphics. Is there a way to solve it without deleting the CUDA package?

You can use nvidia-smi to show the GPU card enumeration. Typically the most powerful GPU is the 0th device
You set the environment variable CUDA_VISIBLE_DEVICES to the desired GPU (0 for 1st GPU, 1 for 2nd, …)
If you need multiple GPUs, you can common delimitate them (set CUDA_VISIBLE_DEVICES=1,3)

The latest versions of the Nsight Graphics tool only support Pascal and later GPUs. To help isolate and resolve the 3D rendering issues, you may have to use a much earlier version.