We normalize CPU gaming capability and GPU rendering capability onto comparable planning scales, adjust GPU demand for resolution, and report the likely limiting side with a confidence label. The output is an estimate for component planning, not a measured FPS guarantee.
What is the model, in plain terms?
The calculator uses a gaming model (version 0.3 on the current site) built from public gaming data. For each CPU, it estimates a gaming ceiling: roughly how many frames the processor can prepare in demanding games. For each GPU, it estimates rendering pace, then adjusts that pace for resolution, because higher resolutions ask the GPU to do more work per frame while the CPU's work per frame changes much less.
The comparison is therefore not clock speed versus clock speed, or core count versus core count. Architecture, per-core speed, cache and memory performance all feed into the CPU ceiling, and architecture, VRAM and rendering performance feed into the GPU pace. When the CPU ceiling is clearly below the GPU pace at a given resolution, the result leans CPU-limited. When the GPU pace is clearly below the CPU ceiling, the result leans GPU-limited. When they are close, the pairing is balanced for that resolution.
Why does the same pairing change with resolution?
| Resolution | What changes | Typical lean |
|---|---|---|
| 1080p | GPU work per frame is lowest, CPU work is relatively high | More likely CPU-limited, especially at high refresh rates |
| 1440p | GPU work rises, CPU gets more time per frame | Often balanced |
| 4K | GPU work per frame is highest | More likely GPU-limited |
This is why a single percentage for a pairing is misleading. A pairing can be CPU-limited at 1080p, balanced at 1440p and GPU-limited at 4K, and all three statements can be true at the same time. The calculator shows the result by resolution for that reason.
Where does the data come from, and what is provisional?
The database is built from public benchmark-scale data, with model-specific calibration where available. Some derived CPU ceilings and resolution values in the current database are provisional until competitor calibration fixtures are populated. That is stated here openly because it affects trust: a provisional value is an estimate built from the available scale, not a measured result for that exact part in that exact game.
We do not copy competitor scores or republish review charts. Facts are used with credit where appropriate, and our own measurements, where we publish them, are labelled with their date and method. The accuracy test planned for the site will compare calculators against measured results for a set of pairs and publish its method and raw data, so readers can check the claim instead of taking it on trust.
Calculator estimate vs real benchmark
| Question | Calculator estimate | Measured game benchmark |
|---|---|---|
| What does it answer? | Which component is more likely to set the pace? | What performance did one exact test system produce? |
| Inputs | CPU, GPU, resolution and normalized performance data | Game build, scene, settings, drivers, RAM and cooling |
| Best use | Shortlisting parts and spotting an obvious mismatch | Confirming expected FPS in the games you play |
| Main limitation | Cannot know every workload or system condition | May not transfer to a different scene or configuration |
A useful buying decision uses both. Start with the estimate to narrow the choices, then check reproducible benchmarks that match your intended games and settings.
What does the calculator not know?
- Your game: esports, AAA, simulation and strategy games load the CPU very differently. See CPU bottlenecks in esports versus AAA games for why one number cannot fit both.
- Your settings: ray tracing, view distance, crowd density and upscaling change which part is the limit.
- Your RAM, cooling and drivers: single-channel RAM, thermal throttling and old drivers can all look like a CPU or GPU limit.
- Your monitor: a 60 Hz display hides a CPU limit that a 240 Hz display would expose.
For those reasons, use the calculator as a shortlist, not a verdict. Confirm a suspected processor limit with the 10-minute CPU bottleneck test and with benchmarks for the games you actually play.
Then try the exact pair in the PC bottleneck calculator . A component guide such as the Ryzen 7 9800X3D analysis can add platform context.
The RTX 5070 analysis separately shows how the GPU-side recommendation changes with resolution.
Limitation: every result on this site is an estimate. If a result looks wrong for your game, the game, settings or system conditions above are the first places to check, and we would rather you test than trust a percentage.
Editorial and model transparency
This page documents the current model version and its known limits. Derived values are not presented as laboratory measurements. Material model changes should update this page and its modified date so readers can distinguish an editorial revision from a change to the calculator itself.
Frequently asked questions
Are PC Bottleneck Lab results exact?
No. Results are estimates based on public benchmark data and a resolution-aware gaming scale. Your game, settings, RAM, cooling, drivers and monitor all change the real result, so treat every result as guidance, not a guaranteed frame rate.
Why do you show ranges instead of one percentage?
Because the same pairing can be CPU-limited at 1080p, balanced at 1440p and GPU-limited at 4K. A single percentage hides that change and misleads builders who play at a different resolution from the one the number was calculated for.
Where does the data come from?
From public benchmark-scale data, with model-specific calibration where available. Some derived ceilings and resolution values in the current database are provisional until calibration fixtures are populated, and that limitation is stated openly on this page.
Does the calculator know which game I play?
No. It estimates a representative demanding-game average, not a result for one title. Esports, AAA, simulation and strategy games load the CPU very differently, so confirm the result with benchmarks for your own games.
What should I do if a result looks wrong for my game?
Test with monitoring software first, using the resolution test in our CPU and GPU check guides. If the game, settings or system conditions explain the difference, trust the test. If not, the result may reflect a provisional value in the database.