Each frame has a delivery time. A 60 FPS average implies 16.7 milliseconds per frame, but the experience can still stutter if a small group of frames takes far longer. Percentile lows summarize that tail more clearly than the average.

Recommended evidenceA repeatable frame-time or completion-time trace
Shortcut to avoidComparing different scenes or warm-up states

What to record before testing: frame time and 1% lows

Average FPS tells you speed. Frame-time distribution tells you whether that speed feels consistent.

Signals that support frame time and 1% lows

  • Frame time is the inverse of frame rate, so spikes are visually important.
  • 1% lows should be compared in the same scene and test length.
  • Shader compilation, storage, CPU scheduling and VRAM can all create spikes.
  • A stable cap can feel better than an oscillating higher average.

A repeatable test for frame time and 1% lows

Test frame time and 1% lows with the following controlled sequence and record the response after each change.

Controlled checks for frame time and 1% lows

  1. Graph frame times and inspect spike frequency and magnitude.
  2. Repeat a route after shader caches are warm.
  3. Change one subsystem-sensitive setting at a time.
  4. Compare capped and uncapped runs.

How to interpret evidence about frame time and 1% lows

For frame time and 1% lows, apply the lowest-risk relevant change first and repeat the original measurement before moving on.

Low-risk actions for frame time and 1% lows

  • Use an FPS cap the system can sustain consistently.
  • Reduce CPU-heavy crowd, simulation or draw-distance settings for CPU spikes.
  • Reduce textures if VRAM pressure is present.
  • Treat one-off loading transitions separately from continuous gameplay.

False positives to rule out for frame time and 1% lows

Before concluding that frame time and 1% lows is the cause, exclude these competing explanations in the original test scene.

  • A short capture can make percentile values unstable.
  • Loading screens and menus can contaminate an otherwise repeatable sample.
  • Frame-generation display rate and application-present rate are different metrics.

Scenario matrix for frame time and 1% lows

Target FPSFrame budgetOne-millisecond share
60 FPS16.67 ms6%
120 FPS8.33 ms12%
240 FPS4.17 ms24%

Worked example: Frame Time and 1% Lows

Apply the evidence, not the label

Two presets both average 100 FPS. The first holds most frames near 10 ms; the second alternates between 6 ms and 18 ms. Their averages match, but the second feels less consistent and produces weaker percentile lows. The frame-time trace reveals the difference immediately. Capturing the same sixty-second route three times shows whether those spikes repeat or were an isolated loading event.

Questions about frame time and 1% lows

Is a higher average always better?

No. A slightly lower but more consistent frame rate can feel smoother.

What is a good 1% low?

There is no universal number; it should be close enough to your target that dips are not disruptive.

Do monitors cause stutter?

Heavy logging or multiple overlays can, so validate with a minimal setup.

Continue with evidence

Next steps for Frame Time and 1% Lows

Compare frame time and 1% lows with related causes in the diagnostics evidence hub. Keep the original workload and settings unchanged. Then choose a diagnostics tool or guide for the next question.

01 Diagnostic actionOpen the PC bottleneck diagnostic checklistStart with a repeatable test for frame time and 1% lows on the real system.Open the tool 02 Supporting guideHow to Check a PC Bottleneck in 20 MinutesA repeatable test plan using utilization, frame time, temperatures and controlled setting changes.Read the guide 03 Supporting guidePC Bottlenecks at 1440pThe practical middle ground where CPU and GPU choices most often need to be balanced together.Read the guide

Technical sources and further reading

The references below support the specific measurement method or technology discussed for frame time and 1% lows. Check the documented software version and system configuration before transferring a result to another PC.