A bottleneck percentage is a model's estimate of mismatch under stated assumptions. It is not a physical property of a CPU–GPU pair. Change resolution, frame target or workload and the limiting stage can change with it.
Learn what a bottleneck percentage can and cannot show, why results change by workload, and how to verify an estimate with real performance data.
Do not accept the hypothesis until the same scene responds to one controlled change.Competing explanations to clear first
Start by separating this symptom from other limits that create a similar graph, utilization reading or user experience.
- The number is not a predicted percentage of lost FPS.
- A different resolution can change the result without changing hardware.
- A capacity or thermal fault can override the model's compute balance.
A repeatable test for a bottleneck percentage
Test a bottleneck percentage with the following controlled sequence and record the response after each change.
Controlled checks for a bottleneck percentage
- Check the calculator inputs for resolution, target FPS and workload.
- Read which component is predicted to limit, not only the number.
- Test two nearby scenarios to see whether the conclusion is sensitive.
- Validate the prediction in the actual game or application.
Use the response to identify the cause
Read across one row at a time: observation, controlled change and interpretation belong to the same test.
| Scenario | Pressure shift | Why the estimate changes |
|---|---|---|
| 1080p at 240 FPS | Toward CPU | Very short frame-preparation budget |
| 1440p at 120 FPS | Mixed | Reference scenario for this model |
| 4K at 60 FPS | Toward GPU | Higher pixel workload per frame |
What to record before testing a bottleneck percentage
Learn what a bottleneck percentage can and cannot show, why results change by workload, and how to verify an estimate with real performance data.
Signals that support a bottleneck percentage
- The denominator and benchmark mix determine the number.
- A small calculated mismatch is normal.
- Percentages should express direction and severity bands, not false precision.
- Observed frame times are the final authority.
Diagnostic walk-through for a bottleneck percentage
The same CPU and GPU may show a small estimated mismatch at 1440p and a larger GPU-side mismatch at 4K. That does not mean performance was lost between two identical tests. It means the question changed. The percentage ranks which path to test first under each stated scenario.
How to interpret evidence about a bottleneck percentage
For a bottleneck percentage, apply the lowest-risk relevant change first and repeat the original measurement before moving on.
Low-risk actions for a bottleneck percentage
- Use the result to prioritize measurement.
- Avoid replacing hardware for a single-digit estimate.
- Compare total upgrade cost, platform changes and your target experience.
- Recalculate when the workload or display changes.
Questions that arise during diagnosis
Is 0% possible?
Not in every workload. A perfectly matched score is still only matched for the model's assumptions.
Is 10% bad?
Not automatically. The effect may be invisible if performance already meets your target.
Why do calculators disagree?
They use different data, scaling assumptions and formulas.
Technical sources and further reading
The references below support the specific measurement method or technology discussed for a bottleneck percentage. Check the documented software version and system configuration before transferring a result to another PC.
