Do we really need two copies of cudaDNN files?

As per the instructions here, Installation Guide :: NVIDIA Deep Learning cuDNN Documentation,
all the cudaDNN files should be copied into the main cuda directory.

But the instructions at GPU support  |  TensorFlow say that both the main
cuda folder as well as the folder where cudaDNN was installed must be on the path.

Do we still need the folder to cudaDNN on path if all of the files in that directory
were copied to the main cuda directory? Can I delete one of the two sets of these repeated files?