Why Benchmarks Use Frame Time Percentiles Over Minimums

This article explains why computer hardware reviewers rely on frame time percentiles rather than absolute minimum frame rates when benchmarking. While average frames per second (FPS) only tell part of the performance story, measuring the lowest drops is crucial for understanding gameplay smoothness. Absolute minimums are easily skewed by single, unrepeatable background events, making them unreliable for hardware testing. By utilizing 1% and 0.1% low percentiles, reviewers can filter out random system anomalies and provide an accurate, repeatable measurement of a component's true worst-case performance and stuttering.

The Flaw with Absolute Minimums

Absolute minimum FPS measures the single slowest frame rendered during an entire benchmark run. This metric is fundamentally flawed because computers are complex systems running hundreds of background processes simultaneously.

If a benchmark renders 10,000 frames and 9,999 of them run at a smooth 100 FPS, the user will have a great experience. However, if a sudden operating system background task, a storage drive hiccup, or a game engine quirk causes just one single frame to drop to 8 FPS, the absolute minimum for that test is recorded as 8 FPS. Because these isolated anomalies cannot be consistently replicated from run to run, absolute minimums are scientifically useless for comparing one piece of hardware against another.

Understanding Frame Times

To accurately measure smoothness, reviewers must look at frame times rather than just average frames per second. Frame time is the exact duration, measured in milliseconds (ms), that it takes for the system to render a single frame.

For example, a perfectly smooth 60 FPS requires a new frame to be delivered every 16.67 milliseconds. If a system delivers 59 frames in 1 millisecond each, and the 60th frame takes 941 milliseconds, the average is still 60 FPS, but the user will see a massive nearly one-second freeze. Tracking the exact millisecond delivery of every frame allows reviewers to see the stutters that average FPS hides.

How 1% and 0.1% Lows Work

To capture stuttering without letting a single rogue frame ruin the data, reviewers use percentiles—specifically the 99th and 99.9th percentiles of frame times, which are commonly converted to FPS and labeled as "1% lows" and "0.1% lows."

Instead of looking at the single worst frame, benchmarking software records the frame time of every single frame, sorts them from fastest to slowest, and isolates the worst 1% or 0.1% of the data.

For example, if a benchmark records 10,000 frames:

  • Average FPS looks at all 10,000 frames.
  • 1% Low averages the slowest 100 frames.
  • 0.1% Low averages the slowest 10 frames.
  • Absolute Minimum looks only at the single slowest frame.

Once the software calculates the average millisecond time of those slowest 100 or 10 frames, it converts that number back into an FPS value so it is easy to compare alongside the standard average FPS on a chart.

Reflecting the Actual User Experience

Frame time percentiles directly correlate to what a person actually feels when playing a game. A single dropped frame lasting a fraction of a second is virtually imperceptible to the human eye. Therefore, the absolute minimum does not reflect what the user sees.

Conversely, if 1% or 0.1% of all frames are consistently taking much longer to render than the rest, the user will experience noticeable micro-stuttering and choppy gameplay. By isolating this small percentage of the worst frames, reviewers successfully ignore one-off background flukes while accurately capturing sustained performance drops. This ensures that benchmark data is both repeatable in the testing lab and meaningful for the consumer.