Besides A100, V100 and RTX 30x0 GPUs, many other GPUs are supported.
But for K80, it is a bit old. It has compute capability of 3.7 only.
Suggest using dgpu which has compute capability of 5.2 or above.
See CUDA GPUs - Compute Capability | NVIDIA Developer
Unable to train | tensorflow.python.framework.errors_impl.InvalidArgumentError: Conv2DCustomBackpropInputOp only supports NHWC - #4 by saketm2408
1 Like