GA9Z — Gaming Performance Check | ga9z.net
An ordered procedure: pick a repeatable scene, fix resolution and quality preset, run for 60 seconds, log average FPS, the 1% low and frame time, then repeat after every change.
Start by choosing a scene you can return to: a fixed camera position in a game world or a scripted benchmark loop. The point is that run two loads the same scene as run one, so the only difference between the runs is the variable you are testing.
Lock resolution and the quality preset before the first run and write both down beside every result. A preset swap between runs makes the figures incomparable, and an unnoticed resolution change can move the average by tens of percent.
Let each run last 60 seconds. Short runs swing on shader warm-up and background loading rather than on your hardware; a full minute lets those transients settle out of the average.
On narrow screens, swipe or scroll the plate sideways.
Log three figures from the same pass: average FPS, the 1% low and frame time. The average alone hides stutter — a game may average 60 FPS and still look choppy on a poor frame time — so the 99th-percentile floor is what reviewers report alongside it.
Repeat the whole sequence after every driver update, settings change or hardware swap, one change per run. Two changes at once can cancel each other out and leave you with a number that explains nothing.
Four failure modes produce a flat or misleading figure. VSync left on caps the frame rate at the display refresh rate and raises input lag; an in-game frame cap pins the number below your GPU's real ceiling; a thermal throttle decays clocks over minutes so the average drifts downward; and a recorder running in the background steals frames and adds stutter your hardware would not otherwise show. Rule each out in turn before drawing a conclusion about the machine.
Further reading