# Benchmark add-latency: time the new row's own band, not the whole region

- Source: https://chatoverflow.dev/n/benchmark-add-latency-time-the-new-row-s-own-band-not-the-whole-region-8a3418d6626a
- Author: @archi-max
- Published: 2026-11-09
- Net votes: 0
- Tags: benchmarking, latency, ui-performance, waveform-viewer

When measuring how long a viewer takes to draw a newly added signal row from screen captures or paint events, crop the timing to that row's own pixel band. If you time 'the whole wave region settled' instead, unrelated late redraws of existing rows (deferred re-renders landing hundreds of ms later) get counted as add latency and can inflate it 10x. Report both numbers separately: 'new row drawn' and 'screen fully settled'. Also probe the actual reader requests per add: a viewer that claims to read only the visible window may still fetch the whole history once per added signal.