I have NVIDIA RTX A2000 Laptop GPU, and on CUDA GPU Compute Capability | NVIDIA Developer it’s said, that RTX A2000 works with CUDA 8.6, but on Build from source on Windows | TensorFlow there is no any info which version of tensorflow works with CUDA 8.6. The closest it CUDA 8 for which tensorflow_gpu-1.4.0 and python 3.5-3.6 is compatible. Can anyone advise me what versions of python, tensorflow, CUDA and cuDNN should I use for my NVIDIA RTX A2000 Laptop GPU?
You’re interpreting the Compute Capability list incorrectly.
The A2000 has a compute capability of 8.6, which is the architecture of the GPU. The Cuda Toolkit version is not directly related and Cuda Toolkit 11.1 was the first version to support it.
If you look at the “GPUs supported” table here, you can see which Cuda version supports which Compute Capability.
Now I work on python 3.10 and tf 2.10 with CUDA 11.2 and cuDNN 8.1. Tensorflow found and is using my RTX A2000 with compute capability 8.6 well. Just during tf 2.10 instalation, numpy 2.2.5 followed it and I had to downgrade it to 1.26.4.