One number, three conditions

FPS always depends on the game, the resolution and the graphics settings. The same build reaches roughly three quarters of its 1080p result at 1440p, and less than half at 4K. Settings behave much the same way: an ultra preset costs about a quarter of the frames against high, while medium hands back about a fifth.

What the processor limits and what the graphics card limits

A build gives you the lower of two figures: how many frames the graphics card can draw and how many the processor can prepare. At 1080p and in esports titles the processor usually decides, at 4K almost always the graphics card. That is why the same processor upgrade changes everything in one game and goes unnoticed in another.

The average hides the stutter

Both our estimate and most published results are averages. They do not show the short dips you feel as stutter - those come more often from memory speed, background programs or the game streaming in its own data than from the raw power of the card. Treat an average of 60 frames as a floor, then, not as a target.

VRAM: the limit an average will not show

In texture-heavy games, cards with 8 GB of graphics memory can fall apart at higher resolutions and high settings even when their raw speed looks sufficient. If you are aiming at 1440p or 4K with high textures, take the amount of VRAM as seriously as the performance itself.

Match the number to your monitor

Frames above the refresh rate of the monitor never reach the screen, so with a 144 Hz or 180 Hz panel it is worth aiming at an average close to that figure. On a 60 Hz screen 200 frames gain you almost nothing; it works the other way round too - a fast monitor without enough frames goes to waste.

What the estimate does not tell you

Our calculator is a model calibrated on the medians of published tests, not a measurement of your PC. Drivers, game updates, memory speed and background programs push the real result in both directions. Use the estimate to compare options, not as a promise of a particular number.