Performance software is full of numbers nobody can reproduce. Here is what to measure instead, how to measure it, and where tuning honestly will not help you.
Take every reading before you change anything, then again after, on the same machine with the same workload. A benchmark against someone else's PC tells you nothing about yours — different drivers, different thermal headroom, different bottleneck.
How to measure
LatencyMon or the app's own Interrupts panel, 10 minutes idle then 10 minutes under load.
Why it matters
High driver latency is the usual cause of stutter that framerate counters do not show. The Interrupts panel names the offending driver rather than reporting one number.
If your latency is already clean, tuning it will not change anything you can feel. Measure first.
How to measure
CapFrameX or the in-game benchmark, same scene, same settings, three runs before and three after.
Why it matters
Average FPS hides stutter. The bottom percentiles are what you actually perceive as smoothness, and they are what scheduling and power tweaks move.
Average FPS is mostly fixed by your GPU. Expect the lows to tighten, not the average to jump.
How to measure
Task Manager after a clean boot, before and after the startup and services passes.
Why it matters
Fewer resident background tasks means less contention for cache and scheduler time, which shows up in the same percentile lows above.
Freeing RAM is not itself a speed-up on a machine that was not short of RAM.
How to measure
A high-speed camera, or a hardware latency tool if you have one. Same polling rate and same display mode each time.
Why it matters
Input path settings, display pipeline and buffering changes land here rather than in framerate.
This is the hardest one to measure honestly. If you cannot measure it, treat claims about it — ours included — with suspicion.
We do not publish a headline percentage. Any single number would be true of one machine and misleading on yours — the gain depends entirely on what was wrong before, and on a well-maintained system it can be close to nothing.
What the software guarantees is different and checkable: it tells you what it changed, it writes a restore point before touching the registry, and it puts everything back on request. If it does not help your machine, you can undo it and ask for a refund.