And the worst thing is that with GPU rendering it's pretty easy to introduce extra latency - if you buffer too much, miss sync events etc - and there's inherently more latency to the rendering process than just writing a pixel color into a framebuffer and having it scanned out to the screen.

The whole point of >60FPS would be to able to reduce the delay between the user action and the results appearing on the screen.

The irony is just like with manufacturing and CPUs, the secret to getting high throughput necessary for high FPS is deep pipelining - meaning more delays compared to just drawing stuff.

The parallel of this in games has been frame generation, and fake frames - the way this is implemented that the game waits an extra frame's worth of time, then starts generating an interpolated frame between the two states, which adds extra time as well.

I propose, that instead of FPS, gaming benchmarks should move to the metric of action-to-pixel (ATP). For this metric, the FPS(or frame time) is a lower bound - if your game runs at 60FPS, your frames take 16.66ms to render - so by definition you can't react any faster than that.

Not sure if anyone's working on this, but this would be a nice way to combat the framegen BS that's going on in the games industry right now.