Hardware limitation of concurrent application access

Now I know that the GPU hardware does have a limit to the number of applications that can access it directly.

My question is, does the limitation the same for different Nvidia cards? And is there a proper method to check whether an error is due to the hardware limitation.

(Sorry for creating a new topic, since no one answered the previous one.
Previous one: https://devtalk.nvidia.com/default/topic/805893/linux/any-limitation-of-concurrent-processes-using-gpu-/ )