Skip to main content
All terms
Hardware & Systems

GPU

Massively parallel hardware that does the heavy math for AI training and inference.

Definition

A GPU is a processor with thousands of cores designed for the parallel matrix and vector math that dominates deep learning. Its throughput on these operations vastly exceeds a CPU's, making it the workhorse of AI. In AI contexts, GPU usually means a data-center accelerator (e.g. NVIDIA H100/H200) programmed through CUDA.