It should be noted that the real reason gamers want better framerates is so you get a faster response between you making an input and the game doing something.
Frame generation turning 30 fps into 120 fps may well make things look smooth to an observer, but they’ll still play like a PS4 game in treacle.
Maybe. In my use case it’s bumping from 72fps to 144fps (I have a 40 series card so I can only do 2x frame generation). In that case, the delay to allow for interpolation is likely around 14 milliseconds, potentially slightly longer (although nVidia claims there is a predictive element to DLSS frame generation, so it might not require the next true frame to be generated yet, I’m not sure). I can’t react to anything in that time anyway. Human reaction time is much slower than 14 milliseconds (or, indeed, than the 33ms that 30fps to 120fps with 4x would cause).
I think the benefits to smoother framerate is about more than just latency. I suspect our brains process better when the visual input is smoother. Makes it easier to track movement and such. Even professional F1 drivers and esports players have reaction times at around 150ms minimum, so the latency is something we’re always compensating for (indeed, there is a processing delay in our brains from when visual input is received, latency is literally something our brains compensate for all the time in our daily lives).
But obviously if people are more worried about that latency they can disable frame generation and lower their graphical fidelity to achieve the same framerates. Stuff like frame generation and super resolution is very much about eking every bit of graphical performance out of the silicon as possible, and isn’t strictly necessary.
It should be noted that the real reason gamers want better framerates is so you get a faster response between you making an input and the game doing something.
Frame generation turning 30 fps into 120 fps may well make things look smooth to an observer, but they’ll still play like a PS4 game in treacle.
Maybe. In my use case it’s bumping from 72fps to 144fps (I have a 40 series card so I can only do 2x frame generation). In that case, the delay to allow for interpolation is likely around 14 milliseconds, potentially slightly longer (although nVidia claims there is a predictive element to DLSS frame generation, so it might not require the next true frame to be generated yet, I’m not sure). I can’t react to anything in that time anyway. Human reaction time is much slower than 14 milliseconds (or, indeed, than the 33ms that 30fps to 120fps with 4x would cause).
I think the benefits to smoother framerate is about more than just latency. I suspect our brains process better when the visual input is smoother. Makes it easier to track movement and such. Even professional F1 drivers and esports players have reaction times at around 150ms minimum, so the latency is something we’re always compensating for (indeed, there is a processing delay in our brains from when visual input is received, latency is literally something our brains compensate for all the time in our daily lives).
But obviously if people are more worried about that latency they can disable frame generation and lower their graphical fidelity to achieve the same framerates. Stuff like frame generation and super resolution is very much about eking every bit of graphical performance out of the silicon as possible, and isn’t strictly necessary.
Streamed games over 100ms+ latency before and it’s felt fine though, that part depends a lot on the game. Flight/space sim isn’t going to care much.
I suspect it doesn’t matter too much on joypad controlled games either.
Mouse aiming and especially VR are more likely to feel the difference between real and fake frames.
That’s a good point which I’d not considered. Yeah, generated frames can’t really respond to you in real time.