To make the best use of mixed precision training paradigms, we need to harness the right technology. In the latest Paperspace blogpost, we cover the Tensor Core in depth to show how each generation of GPU microarchitecture has improved as a tool for deep learning