3 ms·Transistor density/cost still helps with the inherently-parallel ML workloads.by kemiller 5y agoTransistor density/cost still helps with the inherently-parallel ML workloads.