FP64

Jensen Huang on GPUs - Computerphile

Computerphile

Double precision floating point, important for scientific computing but increasingly emulated to make space for tensor cores.