I was trying out FSR4 on my RX 6800 XT, Fedora 42. Works really well and it easily beats FSR3 in visuals even on Performance. It does have a significant performance hit vs FSR3 though but it still works out to be a bit faster than a native rendering on Quality.
Input lag is caused by frame interpolation, right? Or nah?
It’s because game logic is calculated on real frames and these things lower the real frame rate even though they give you more rendered frames. If you were getting 40 real FPS, and then you go to 30 real fps, you will feel a significant amount of lag even if you are getting 60 fps in fake frames. Basically the game loop is running slower and stuff like input polling is happening slower even if you have a higher frame rate.
Frame generation shouldn’t be a bottleneck on the CPU though, should it? That stuff is happening on the GPU. I know I saw a video about this stuff but I can’t remember the real reason input lag increases with frame generation/interpolation.
it’s not. The whole point of FG was to take advantage of high refresh rate monitors as most games can’t render 500FPS even on the fastest CPU… alas, here we are with games requiring FG to get you to 60FPS on most systems looks at Borderlands 4 and Monster Hunter Wilds
Right, but FG shouldn’t be touching the CPU in any way, should it? It should be a local thing on the GPU transparent to the CPU, unless I’m misunderstanding how it works.
It’s not using CPU
It’s kinda the same thing. You get input lag based on the real framerate. Since interpolation requires some extra performance the base framerate will likely be a bit lower than the framerate without interpolation which will case an increase in input lag while providing smoother image.
It seems that the input lag is more perceived, rather than actually experienced, from what I understand. Like if you go from 30 to 120 fps, you expect the input lag to decrease, but since it stays the same (or slightly worse), you perceive it to be much more severe.
The frame rate isnt going from 30 to 120 FPS. It’s actually going from 30 to like 20. The rendered frames are different then the CPU frames which handles the game loops, (physics, input, simulation, etc)
Not sure we have the same definition of frames here.
yes, that’s why FPS in this case is not a good measure of performance
Very much so. The very reason why we want more fps is to have less input lag, that’s my personal take anyway. That’s the only reason why I have a beefy computer, so the game can respond quicker (and give me feedback quicker as well).