hardware accelerator

A hardware accelerator is a specialized processor designed to perform a specific class of computations more efficiently than a general-purpose digital computer’s CPU. Efficiency here means some combination of higher throughput, lower latency, or lower power consumption for the target workload.

The general pattern is: identify a computation that a CPU performs by executing many generic instructions, then build circuitry that performs that computation directly. The trade-off is generality for efficiency — an accelerator that excels at matrix multiplication cannot run arbitrary programs.

Common hardware accelerators include graphics processing units (GPUs), neural processing units (NPUs), FPGAs (field-programmable gate arrays), and DSPs (digital signal processors). Each targets a different workload profile. A modern system-on-chip (SoC) like those in laptops and phones typically integrates several accelerators alongside the CPU, sharing the same memory pool.