Tensorflow and cuda issues

Hi All, Need some help here!
Just built a system msi 670e wifi mobo, ryzen 9 9750x3d, palit geforce rtx 5080 gpu and 64gb ddr5 6000mhz.
I have managed to get the gpu recognized with the following combo: Python 3.10, Cuda 11.2 cudnn 8.1, tensorflow_gpu==2.10.0 and it only works with Numpy==1.21.6, Pandas==1.5.3 and transformers==4.20.1 with pip left at 22.0.1. This is all under win 11 pro as i know shame on me could not get used to ubuntu…
I am currently training a dqn however it does not seem to use the gpu and it rather uses the cpu power which is a lot slower. Anyone can help me please with possible different config or setup.