01
Fix the test scene
Same save, same route, same time of day, no cutscene in the middle. A benchmark that is not repeatable measures nothing.
Log **frame times**, ignore the average and read the **1% and 0.1% lows** plus the graph. A flat 90 fps feels better than 120 with spikes every few seconds.
Requirements
- CapFrameX or MSI Afterburner with RTSS
- The same save or route for every run
- 60 to 120 seconds per run
Instructions
02
Log instead of watching
Capture 60 seconds with CapFrameX, which drives RTSS underneath. Watching an overlay changes the workload and never shows the spike you are looking for.
Capture: 60 s
Overlay: offtipLogging costs a little performance. Compare runs made the same way, not against a run without logging.
03
Read three numbers
The average tells you the ceiling, the 1% low tells you how the worst second feels, and the 0.1% low catches the single hitch that ruins a fight.
Average: 118 fps
1% low: 71 fps
0.1% low: 43 fps⚠ warningA 1% low below roughly 70% of the average is a visible problem, not a tuning detail.
04
Change one thing at a time
Alter a single setting, repeat the run and diff the 1% low. If the average rises while the low falls, the setting costs you smoothness.
05
Read the graph, not the average
Spikes every few seconds point at background apps or shader compilation. A regular sawtooth points at the frame cap or a VRR window that does not match.
tipExport the CSV if you want to see whether the spikes line up with a driver or a scheduled task.
06
Set your own threshold
Decide before testing what counts as acceptable - for example 1% low above 90 fps in a shooter. Without a number, every tweak feels like it helped.
Common issues
The log shows impossible frame rates
The capture hooked the launcher instead of the game. Start the capture after the game is in the foreground, or pick the process explicitly.
Two runs of the same scene differ wildly
Background load, a paused update or a different shader cache state. Reboot, close the browser and run each setting twice.
The 1% low collapses only in the first minute
Shader compilation. Run a warm-up pass and start logging afterwards.
Frame times look perfect but the game feels bad
Input latency, not frame pacing. Look at the frame cap, VSync and the polling rate instead.
Frequently asked
Is 1% low the same as stutter?
It is the closest single number. The graph shows whether it is one hitch or a pattern.
Do I need a paid tool?
No. CapFrameX and RTSS are free; the paid tools mostly add comparison databases.
How to undo this
Nothing to undo - logging writes files you can delete. Remove the RTSS hook if a game refuses to start while the overlay is active.
Changelog · 1 entry
- 2026-10-30First version
Interactive version: checklist, feedback, related guides →
Free · no account needed · tracks your progress