Two PCs can both report 90 FPS and yet one feels buttery while the other feels like it hitches every second. The difference is frame pacing: how evenly those frames are spaced in time. Average FPS is a summary number, but your eyes react to the gap between each frame. Understanding frame pacing explains why a lower but steadier frame rate often feels better than a higher, spikier one.
Frame rate vs frame time
Frame rate counts frames per second. Frame time measures how many milliseconds each frame takes. They are two views of the same thing: 60 FPS equals 16.7 ms per frame, 120 FPS equals 8.3 ms, 144 FPS equals about 6.9 ms. Perfect pacing means every frame lands at almost exactly the same interval. Bad pacing means the interval jumps around even if the average stays high.
| Frame rate | Ideal frame time | Feels smooth if variance stays under |
|---|---|---|
| 60 FPS | 16.7 ms | ~3 ms |
| 120 FPS | 8.3 ms | ~2 ms |
| 144 FPS | 6.9 ms | ~1.5 ms |
| 240 FPS | 4.2 ms | ~1 ms |
Why spikes hurt more than a lower average
A single frame that takes 33 ms in the middle of a 90 FPS run reads as a visible stutter, even though the average barely moves. This is why 1% low and 0.1% low numbers matter more than the average when you judge smoothness. A game locked at a steady 90 FPS usually feels better than one that swings between 110 and 70.
What causes poor frame pacing
Common culprits
Shader compilation on first launch, background apps stealing CPU time, a CPU bottleneck in busy scenes, storage stalls when textures stream in, and mismatched frame delivery between GPU and display all create uneven pacing. Multi-GPU setups and some frame-generation implementations can also introduce pacing artifacts if not tuned.
How to improve it
Cap your frame rate a few frames below your monitor’s refresh rate (for example 141 on a 144 Hz panel) so the GPU is never gasping at its ceiling. Enable variable refresh rate (G-Sync or FreeSync) to sync display refresh to render time. Close background overlays and downloads. Turn on Nvidia Reflex or the AMD equivalent to shorten and steady the render queue. Measure the result with a frame-time graph rather than the FPS counter alone.
Measuring pacing instead of guessing
Because your eyes and the FPS counter disagree, the only honest way to judge pacing is to log frame times and look at the distribution. Capture tools record every frame’s time in milliseconds, then report the average alongside the 1% low and 0.1% low, which represent your worst frames.
Reading a frame-time graph
A healthy graph is a nearly flat line hugging your target frame time, say a steady 6.9 ms at 144 FPS. Occasional tall spikes to 20-33 ms are the stutters you feel. If the whole line is fuzzy and thick rather than clean, you have constant small variance that reads as a subtle lack of smoothness even at a high average.
Turning numbers into fixes
Big isolated spikes on level loads usually mean shader compilation or texture streaming from storage. A thick, jittery line often points to background CPU contention or an overloaded CPU thread. Spikes that line up with fan ramps can indicate thermal throttling. Match the pattern to the cause, change one thing, then re-measure to confirm the graph actually flattened.
FAQ
Is 60 FPS with good pacing better than 100 FPS with bad pacing?
Often yes. A rock-steady 60 FPS produces a consistent 16.7 ms cadence your eyes can track, while a jittery 100 FPS that spikes to 30 ms frames will feel like it hitches. For single-player games, smooth pacing usually beats a higher but erratic number.
How do I actually see my frame pacing?
Use an overlay that draws a frame-time graph, such as the one in MSI Afterburner/RivaTuner or CapFrameX. A flat line means good pacing; tall spikes mean stutter. The FPS counter alone hides these spikes because it averages them away.
Bottom line: Frame pacing is the evenness of frame delivery, and it matters more than raw average FPS for perceived smoothness. Cap below your refresh ceiling, enable VRR, cut background load, and judge results with a frame-time graph rather than the average counter.
