I recently read an article that went into great detail about why frame rate is the wrong metric to determine the fluidity of a game experience. Though I may agree with the general concept, I can’t completely agree with the sentiment that Frames Per Second (FPS) is a worthless metric that needs to be replaced.
If we remove some obvious conditions from the equation, you’ll find that FPS is a brute but useful method to improving responsiveness. The most obvious condition is large stutters caused by resource loading or video drivers suffering from bandwidth issues. Assuming that the game has properly balanced their resources for the machine that is being used, we can get down to the root of the issue.
These numbers are very different for all people so please consider them only for the purpose of this argument. The cinema experience is currently only running at 24 frames per second, or 24p. This is considered the standard despite historical claims that a minimum of 46p is required for a fluid viewing experience. Lots of tricks are being used in modern cinema to continue filming at 24p but playing with shutter speeds of the player to emulate speeds of 48p and even 72p, despite those projectors still rendering the same frame multiple times. The important thing to note about the low frame rate in film is that the cinema experience is non interactive, it is a passive experience. We go where the camera takes us.
In video games you have a controller, or a mouse and keyboard. The issue of frame rate becomes more evident because the game’s logic and it’s rendering are synchronized. If a player moves their mouse, there is an expectation that the game will immediately respond. Assuming that the claims are true, that most people can not visually distinguish above 48 frames per second, and that human reaction times may also sit somewhere in that same window, this means that we have around 20 milliseconds to miss the player’s input and have to process it in the next frame, thus causing this perception of lag or stutter.
FPS is a brute force blanket approach to solving this issue. If the simulation runs at a minimum of 2x the average human reaction time, we are given 1 frame to make mistakes. As John Carmack once said, “somewhere between 60 and 120 is just don’t care anymore,” and he may be right. If the assumption of 48 frames is true, then it means that we only need to sustain a frame rate of around 96 FPS to never have to worry about an unresponsive game.
It’s important to note the distiction that I made however. I am talking about 96 FPS simulation speeds. There is the possibility that a game could render their games at 48-60 FPS and simulate their physics and process inputs at 96 FPS, but that sounds like a nightmare. We may have to settle with consuming brute power by synchronizing frames, even if our eyes may only process 60% of the content being displayed.