← All guides

How we cut click-to-photon latency to 0.4ms: the full stack

MB
Marcus 'Bravo' Lee
Lead Performance Engineer · June 21, 2026 · 12 min read

Click-to-photon — the time from switch actuation to pixels changing — is the only latency number that matters, because it's the one your brain plays against. Stock mid-range rigs measure 25-40ms. Our reference build measures under half a millisecond of controllable overhead. Here's the stage-by-stage accounting.

Stage 1: the input device (0.1–8ms)

A 125Hz mouse adds up to 8ms of polling delay alone; 8000Hz cuts that to 0.125ms. Debounce, switch type and wireless implementation stack on top. This stage is pure hardware + configuration — the cheapest milliseconds you'll ever buy back.

Stage 2: OS input path (0.5–5ms)

Windows pointer processing, timer resolution drift and USB interrupt scheduling all sit between the device and the game. Raw input, a locked 0.5ms timer, and pinned interrupt affinity flatten this stage to near-zero — this is most of what our Zero Delay + pack does.

Stage 3: render queue (1–20ms)

The biggest and most misunderstood stage. A GPU at 99% load with a deep pre-render queue is rendering your past, not your present. The fix trio: FPS cap below refresh, Reflex/Anti-Lag on, pre-render queue tuned per game. This stage alone explains why capped setups out-feel uncapped ones.

  • GPU at 99% = latency, cap it
  • Reflex/Anti-Lag → on wherever it exists
  • Queue depth → 1 for esports titles

Stage 4: display (1–30ms)

Panel response, processing modes and refresh rate close the loop. 240Hz halves frame-delivery jitter versus 144Hz; proper overdrive settings prevent smear without inverse ghosting; and any TV-style processing must die. We verify the whole chain with a photodiode rig, not marketing specs.

What 0.4ms actually means

The 0.4ms figure is our measured controllable overhead — stages 2 and 3 tuned to their floor on reference hardware. Your total will always include your device and display physics. But the tuned-versus-stock delta is typically 15-30ms of real, reproducible latency — more than the difference between a 60Hz and 240Hz monitor.

// The TL;DR
  • Click-to-photon is the only latency that matters
  • The render queue hides the biggest milliseconds
  • 8000Hz polling costs less than a monitor upgrade
  • Tuned vs stock = 15-30ms of real advantage
#input latency#technical#deep dive

Skip the manual work

Bravo applies every tweak in this guide — and hundreds more — in one click, fully reversible. Tuned per game, per rig.

See the tweak packs →