In deep learning, preprocessing can be pretty heavy, and then you find yourself CPU-bounded and not making full use of your 4x2080 Ti.