TensorFlow CPU vs GPU vs TPU

CPU are for general purpose computational processor, GPU were designed to accelerate rendering of graphics while TPU were designed to accelerate deep learning tasks

Coming to, CPU, GPU and TPU they can serve the same computational task given the appropriate compiler support.

In the current situation, one can use GPU(Responsible for computing the pixels shown on the screen (in general; CUDA on NVidia allows some somewhat general computations to be done)) and CPU(Responsible for doing general computation and for orchestrating the other hardware components of the system so they work as supposed.) for the general processor but for TPU, other dependent support such as compilers are not available. TPU provides better functionality for the deep learning task involving TensorFlow.