Performance

Latency

The time delay between initiating an operation and receiving the result.

Detailed Explanation

Latency measures responsiveness—how long from request to response. Memory latency is time from address to data. Pipeline latency is cycles from instruction fetch to result. Network latency includes transmission and processing delays.

Latency differs from throughput—a pipeline may have high throughput (one result per cycle) but multi-cycle latency. Latency matters for dependent operations; throughput matters for independent parallel work.