Difference between "compute capability" "cuda architecture" clarification for using Tensorflow v2.3.0

Question

I would like please the difference between "Compute Capability" used by NVidia at https://developer.nvidia.com/cuda-gpus and "CUDA architecture for by Tensorflow at https://www.tensorflow.org/install/gpu .

Is "compute capability" the same as "CUDA architecture".

I have a GeForce 540M with driver version 10.18.13.5435. According to NVidia, the "compute capability" is 2.1. According to the Tensorflow site, the minimum CUDA architecture is 3.5.

If "Compute capability" is the same as "CUDA architecture" does that mean that I cannot use Tensorflow with an NVIDIA GPU?

If I can use my NVIDIA GPU with Tensorflow, what is the meaning of NVIDIA GPU Drivers -CUDA 10.1 requires 418.x or higher? That is what is 418.x referring to? Reference https://www.tensorflow.org/install/gpu "Software Requirements"

Thank you, Anthony of Sydney

Answer 1

Is "compute capability" the same as "CUDA architecture".

Yes, "compute capability" as used by NVIDIA is the same as "CUDA architecture" as used by Google on that particular web page.

If "Compute capability" is the same as "CUDA architecture" does that mean that I cannot use Tensorflow with an NVIDIA GPU?

It means you cannot use Tensorflow with your NVIDIA GPU. It is usable with other, newer NVIDIA GPUs. See here. Your question is essentially a duplicate of that one.