Bottleneck calculator
Check whether a processor and graphics card hold each other back, as a range weighted by resolution and game type rather than a single invented percentage.
Pick a processor and a graphics card and this says whether they hold each other back. The answer is a range weighted by resolution and the kind of game you play, because that is the most a model can honestly claim.
Why a range and not one number
Most bottleneck calculators return a single confident figure: “your build has a 17.3% bottleneck”. No benchmark could produce that number, because hardware does not behave that way.
The same processor and card are processor-limited in a competitive shooter at 1080p and card-limited in a ray-traced single-player game at 4K. One figure cannot describe both, so quoting one to a decimal place is a precision the underlying data does not have.
This tool reports a range instead, and states the resolution and genre the range applies to. Where the two parts are close enough that the verdict could go either way, it says the confidence is low rather than picking a side.
What the controls change
Resolution. At 1080p a fast card finishes frames quicker than the processor can prepare them, so the processor carries more of the load. At 4K almost everything depends on the card. This is the single largest influence on the verdict.
What you play. Esports titles, simulations and MMOs lean on the processor through draw calls, tick rate and unit logic. AAA single-player games lean the other way. The split shifts accordingly.
Upscaling. DLSS and FSR render the game internally at a lower resolution and scale the result up. The card does less work per frame, so the processor becomes the constraint sooner than the output resolution alone would suggest.
Detail preset. This affects only the frame rate estimate, not the balance verdict. Higher settings ask more of the card per frame without changing which part runs out first.
Where the ratings come from
The processor and graphics ratings are a static snapshot published by BottleneckPC under the MIT licence, covering around 310 processors and 140 graphics cards on a calibrated 0-100 gaming scale.
They are not measurements we have taken, and we cannot verify them against a test bench of our own. We can tell you exactly where they came from and when the snapshot was taken, which is stated beside every result.
The ratings ship with this site. Nothing here calls out to a remote server when you use the tool.
What a bottleneck actually is
In any frame, one part finishes its work and waits for the other. Whichever waits is the one with capacity to spare, and the other is the bottleneck for that frame. A perfectly balanced pairing does not exist, because the balance shifts scene to scene within a single game.
A processor bottleneck shows as a frame rate that will not rise when you lower the graphics settings. A card bottleneck shows as the opposite: settings changes move the frame rate a great deal, and the processor sits well below full use.
Neither is a fault. A machine built for 4K will look processor-limited if you play an esports title at 1080p on it, and that is the correct trade for what it was built to do.
Planning a whole machine rather than checking a pair? The PC Builder checks socket, memory, clearance and power across every part you pick.