Problem to run training with the new RTX 3080

Hi,

I used to run an AI trainning on a RTX 2080 and it worked just fine. (Cuda V10.1, cuDNN 7.6.5 and tensorflow 2.3.1)
Latly I upgraded to the 3080 and seems that the program can’t use the GPU (it runs on the CPU).
I’ve tried Cuda V11.2, cuDNN 8.0.5 and tensorflow 2.4.0, Windows 10 64.

Any suggestions?

Hi @avihugoldman,
Please check nvidia-smi command and visible device list from tensorflow API to start with.
In case if the device is not listed, please retry following the steps from below doc.

Thanks!