Measurements & stability 6 min
How to measure your PC’s performance properly
An optimisation is only credible when measurement conditions are comparable. The aim is not the most spectacular number, but a reproducible observation.
Establish a comparable baseline
Use the same machine, resolution, preset and test scene and, where relevant, the same driver and system versions. A change in test procedure can hide or exaggerate a difference.
- Record important version numbers
- Close tasks unrelated to the test
- Repeat the test several times
- Keep the raw results
Choose the right metrics
Average FPS does not tell the whole story. The 1% lows and frametime consistency help assess smoothness, while temperatures and effective clock speeds explain hardware behaviour.
- Average FPS and 1% lows
- Frametime and percentiles
- Temperatures and effective clock speeds
- Any errors or instability
Interpret without overgeneralising
A synthetic score measures a specific scenario. It does not automatically translate to a universal gain in every game. Likewise, DPC latency is not a direct measurement of end-to-end input latency.