Average FPS can hide a game that feels rough. If most frames are fast and a few take much longer, the average stays healthy while you notice hitches. One-per-cent lows and a frame-time graph expose that uneven delivery.

Quick Answer

Use average FPS for broad speed, one-per-cent lows for the slow tail, and frame time for the shape and timing of each hitch. Run the same benchmark several times after shaders have warmed. Compare the full trio, since no single number can explain every pacing problem.

Once the limit is known, compare upgrade candidates across Evetech's GeForce catalogue and its wider graphics-card favourites. A stronger card helps a GPU-bound low; it may do little for a traversal stall or saturated CPU thread.

What a One-Per-Cent Low Means

Benchmark tools order or analyse the captured frame times and summarise the slowest portion. The precise calculation differs between tools, so treat results from separate applications with care. It is not always the literal average of the worst one per cent under every method.

One-per-cent lows remove much of the flattering effect of fast frames. A game averaging 120fps with a 45fps low can feel worse than one averaging 100fps with a steady 80fps low.

Very rare events may show up in a 0.1 per cent low, though that figure becomes noisy in a short capture. Longer, repeatable runs produce more useful tails.

Convert FPS to Frame Time

FPS is a rate; frame time is the duration of one frame. At 60fps the ideal interval is about 16.7ms. At 120fps it is about 8.3ms. A graph makes a 40ms spike visible at the exact point where the game hitched.

Use the graph to classify the pattern. Evenly high frame times suggest a steady performance limit. Regular spikes can point to background polling or a cap problem. Spikes at new areas may involve shader compilation or asset streaming.

Make the Comparison Fair

Keep the route, save, settings and capture length fixed. Close background updates and let launch-time shader work finish. Run at least three passes, discard an obvious interrupted run and report the spread rather than choosing the prettiest result.

Check GPU utilisation, CPU core load and VRAM beside frame time. The relationships tell you whether to lower graphics settings, investigate the processor or wait for a game patch.

FAQ

Is a higher one-per-cent low always better?

Yes within the same sound test, because it means the slow portion completed faster. Different tools or runs still need context.

Why can two games with 100fps feel different?

Their frame-time variation, input path, engine and display presentation can differ despite matching averages.

Should I use 0.1 per cent lows?

They can reveal rare stalls, but capture length and run-to-run variation make them noisy. Pair them with the graph.

Capture the same route three times, read average FPS beside one-per-cent lows and use the frame-time graph to find when each hitch begins.