Average FPS tells you almost nothing about how smooth a game feels. CapFrameX is a free capture and analysis tool built on Intel’s PresentMon that records every frame’s timing and turns it into readable metrics — averages, 1% lows, 0.1% lows, stutter percentage, and frame-time graphs. If you want to prove a tweak actually helped, this is the tool that shows it objectively.
Why frame times beat average FPS
Two systems can both average 100 FPS, but one delivers a smooth line at 10 ms per frame while the other spikes to 40 ms every few seconds. Your eyes read those spikes as stutter. CapFrameX reports percentile lows — the 1% low is essentially your worst 1% of frames — which correlate with perceived smoothness far better than the headline average.
Key metrics it reports
| Metric | What it means | Why you watch it |
|---|---|---|
| Average FPS | Mean frame rate | Headline number only |
| 1% low | Worst 1% of frames | Felt smoothness |
| 0.1% low | Worst 0.1% of frames | Severe hitches |
| Stutter % | Frames far above median time | Micro-stutter detection |
Running a clean capture
Install CapFrameX, launch it, and set a capture hotkey and duration (20-30 seconds is plenty for a repeatable scene). Load your game, walk to a consistent, repeatable spot — the same corridor or benchmark run each time — and press the hotkey. Capturing an identical scene is the single most important habit; comparing two different areas produces meaningless conclusions.
Comparing before and after a tweak
Capture a baseline, change exactly one setting (a driver update, a frame cap, an XMP profile), then capture the same scene again. CapFrameX overlays both runs so you can see if the 1% low rose and the frame-time line flattened. A tweak that lifts your average by 2 FPS but improves 1% lows by 15 FPS is a real win for smoothness even though the headline barely moved.
Reading the graphs
The frame-time chart should look like a flat, low line. Regular spikes point to shader compilation or traversal stutter; a stair-step pattern can indicate a power or thermal limit kicking in. Use the L-shaped and percentile views to quantify how much of your session was spent in “bad frame” territory rather than trusting your gut.
Design a fair test
Good benchmarking is mostly about controlling variables. Pick a repeatable scene — a fixed traversal path, a built-in benchmark, or standing in one spot during a busy moment — and always capture the same length of time. Close background apps, let the GPU reach a steady thermal state with a warm-up run, and disable any dynamic power-saving that could skew consecutive runs. Three captures per configuration and taking the median filters out random noise.
Report the numbers that matter
When you share or record results, list average, 1% low, and 0.1% low together, plus a note on your settings and resolution. A change that lifts the 1% low from 55 to 70 FPS is a real smoothness gain even if the average barely moves, and only percentile lows capture that. CapFrameX’s aggregated session view and its frame-time distribution graph make these differences obvious at a glance.
Treat your captures as evidence, not decoration. The whole reason to run CapFrameX is to replace “it feels smoother” with a measurement you can trust, so keep the raw sessions until you are confident a tweak genuinely helped.
FAQ
Does CapFrameX affect performance while recording?
The overhead is negligible — it uses the lightweight PresentMon backend. You can leave it capturing during normal play without meaningfully changing your results.
Is it accurate across GPU brands?
Yes. Because it measures at the presentation layer rather than relying on a vendor overlay, it works consistently on NVIDIA, AMD, and Intel GPUs, which makes it ideal for fair before/after comparisons.
Bottom line: CapFrameX turns “it feels smoother” into hard numbers. Capture the same scene, watch 1% and 0.1% lows and the frame-time line rather than the average, and change one variable at a time. It is the honest way to confirm whether a tweak actually improved your experience.
