Average FPS can look excellent while a game still feels bad. The useful evidence is how consistently frames arrive and what changed during the spike.
Why average FPS can hide a bad experience
Average FPS compresses an entire run into one number. It can therefore look healthy even when a small number of frames arrive much later than the surrounding frames. Those late frames are experienced as hitches or stutter.
The TWW diagnostic model starts with frame time: how long each frame took, when the spikes occurred, and what the system was doing at those moments.
The first question: is the stutter repeatable?
Repeatability is one of the strongest clues. If a hitch happens at the same doorway, effect, cutscene or first encounter with a rendering feature, game-engine or shader work becomes more plausible. If it appears randomly across games, the investigation should move toward drivers, power, thermals, background software or system stability.
Do not change graphics settings randomly between tests. A diagnostic comparison is useful only when the test conditions remain comparable.
| Stutter pattern | Likely direction | Useful evidence |
|---|---|---|
| First encounter with an effect or area | Shader compilation or asset preparation | Repeat the same path after caching/building |
| CPU usage spikes while GPU usage falls | CPU-side bottleneck or waiting | Frame-time graph + CPU thread/load evidence |
| GPU stays heavily loaded during the spike | GPU workload or thermal/power behavior | GPU clocks, temperature, utilization and frame time |
| Storage activity rises exactly at the hitch | Asset streaming or storage latency | Disk active time, throughput and repeatability |
| Occurs across many games | System-level cause | Drivers, thermals, power, memory and background tasks |
A controlled TWW stutter test
A useful test changes one variable at a time and repeats the same route or benchmark section. The point is not to create a perfect benchmark; it is to make the cause easier to isolate.
- Record a repeatable section of gameplay with frame-time data enabled.
- Run the same section twice without changing settings.
- Mark the exact moment of each large frame-time spike.
- Compare CPU load, GPU load, clocks, temperatures, VRAM/RAM pressure and storage activity around the spike.
- Change one suspected variable, repeat the same section, and compare the frame-time pattern rather than only the average FPS.
What TWW will and will not claim
A frame-time graph can show that a hitch occurred and can reveal correlations. It cannot, by itself, prove that a specific component is defective. The strongest conclusions combine repeatability with multiple independent signals.
This is also why TWW will not label every hitch “shader compilation stutter” simply because the game stutters. Shader behavior is one hypothesis that must fit the pattern.
Bottom line
The fastest path to a better gaming experience is not always lowering graphics settings. First determine what is producing the late frames. Once the bottleneck is identified, the setting or hardware change becomes much easier to justify.
TWW research framework. No universal FPS or frame-time result is claimed without a documented test run on specified hardware, software and game settings.
See the evidence behind this topic
These TWW research pieces examine the measurements, mechanisms, or testing questions related to this guide. Research pages distinguish measured results from general technical guidance.
- What Actually Causes PC Game Stuttering? A Frame-Time InvestigationAverage FPS can look excellent while a game still feels bad. The useful evidence is how consistently frames arrive and what changed during the spike.
Sources & further reading
We distinguish documented facts, measured results, and informed guidance. If a claim depends on a specific device, software version, game, or test method, that context should be stated rather than implied.
Meet the author·See our testing methodology·Read our editorial policy