Comment by lajawfe

Can the FP32 and Tensor core modules compute at the same time, or are they an abstraction over the underlying silicon, and we can only utilize one at a time?