Why Frame Pacing Beats Average FPS in Hardware Benchmarks

For decades, average frames per second (FPS) has been the standard metric for benchmarking computer hardware, yet it often fails to reflect the actual user experience. This article explores why frame pacing—the measurement of the exact time it takes to render each individual frame—is a far more reliable metric than average FPS. By examining the flaws of averages, the mechanics of frame times, the importance of 1% lows, and how hardware bottlenecks manifest, you will understand why consistent frame delivery is the true indicator of smooth performance.

The Flaw with Average FPS

Average FPS simply counts how many frames your hardware can push to the monitor over the course of one second. While a higher number generally indicates better performance, averaging the data destroys critical information about what happened within that second.

Imagine a scenario where a graphics card renders 59 frames in the first half of a second, and then freezes, rendering only 1 frame in the second half. The counter will still read 60 FPS for that second. On paper, 60 FPS looks completely smooth. In reality, the user just experienced a massive half-second freeze. Average FPS completely masks these micro-stutters and hitches because fast frames mathematically cancel out the slow ones.

What is Frame Pacing?

Frame pacing measures the consistency of frame delivery. Instead of counting frames per second, it measures "frame time"—the exact amount of time, in milliseconds (ms), it takes for the system to render a single frame.

To achieve a truly smooth 60 FPS, a computer should not just produce 60 frames in one second; it must deliver one frame every 16.67 milliseconds. For 120 FPS, a frame must be delivered every 8.33 milliseconds. Good frame pacing means the time between each frame remains steady. If one frame takes 16ms, the next takes 45ms, and the third takes 8ms, the frame pacing is poor. The human eye is highly sensitive to these sudden shifts in timing, perceiving them as jarring stutters, even if the overall FPS counter remains high.

Understanding 1% and 0.1% Lows

Because reading raw millisecond graphs can be difficult, hardware reviewers translate frame pacing data into "1% lows" and "0.1% lows." These metrics isolate the worst-performing frames in a benchmark run and express them as an easy-to-understand FPS number.

To calculate the 1% low, benchmarking software takes all the frame times recorded during a test, isolates the slowest 1% of those frames, and averages them. If a game has an average FPS of 100, but the 1% low is 25 FPS, it means the game frequently dips to a choppy 25 FPS equivalent. If the average is 100 FPS and the 1% low is 90 FPS, the frame pacing is tight, and the gameplay will feel exceptionally smooth.

Exposing Hidden Hardware Bottlenecks

Frame pacing is the ultimate diagnostic tool for finding weak links in a computer build that average FPS would otherwise ignore.

CPU Bottlenecks: A powerful graphics card might push a high average FPS, but if the CPU is struggling to keep up with game logic, physics, or background tasks, it will delay sending data to the GPU. This results in massive spikes in frame time.
VRAM and System RAM Limits: When a graphics card runs out of video memory (VRAM), it must pull data from the much slower system RAM. This causes severe, momentary stutters. Average FPS will barely drop, but a frame time graph will show violent spikes every time the memory is swapped.
Thermal Throttling: As hardware heats up, it may rapidly drop and raise its clock speeds to manage temperatures. Frame pacing reveals these micro-fluctuations instantly.

Perceived Smoothness vs. Raw Speed

Ultimately, benchmarking is about measuring the quality of the user experience. A game running at a locked, consistent 60 FPS with flat frame pacing (16.67ms per frame) will feel significantly smoother and more responsive to the player than a game averaging 90 FPS with erratic frame times jumping between 8ms and 50ms. Frame pacing measures the actual rhythm of the hardware, making it the definitive metric for real-world performance.