Sony patents AI that predicts a game's next frame before the GPU finishes rendering it

You realize that generated frames can't respond to input by definition, which means your input lag gets weird, right? FG can make a game look faster but internally it will still be running at the base frame rate, which is what input is linked to.
Right. But if the base frame rate is good and input lag is well within acceptable specs , then FG can make it look smoother. This doesn't have to be an all-or-nothing scenario.
 
Right. But if the base frame rate is good and input lag is well within acceptable specs , then FG can make it look smoother. This doesn't have to be an all-or-nothing scenario.
FG is only a good thing if you're already getting good frames. The math says it can make 120 FPS look like 240 while feeling like 60. The problem I see with that is that if you are already getting 120+FPS then framegen is kind of unnecessary.

What makes it really dumb, and that the people who need this the most are probably getting less than 60FPS already. If you're getting 30FPS and want to get 60 then turning framegen on in those situations can make the game feel like 10. And I'm guilty of doing that. I've cranked up all the settings in CP2077, turned on framegen and tried to play for a bit and it's fun for about 10 minutes until I have to turn everything back down to get the ~90FPS that makes the game playable.
 
You realize that generated frames can't respond to input by definition, which means your input lag gets weird, right? FG can make a game look faster but internally it will still be running at the base frame rate, which is what input is linked to.
I notice zero input lag in the games I play. I also prefer playing with a controller, which probably helps with this, but I also don't play anything PvP at all as I despise rando people, and so I stick to co-to-quad-op and single-player games.
 
There's an interesting distinction from conventional frame generation. Instead of waiting for the next rendered frame, Sony wants to predict it using previous frames and potentially controller inputs. The big question is what happens when the prediction is wrong, especially in fast-moving games.

We're slowly approaching the point where the GPU renders half the image, AI invents the other half, and marketing calls it 240 fps. I'm not necessarily against the technology, but I would love to see more emphasis on responsiveness and actual image quality.
 
It may (or may not) look better.
But it does not react better to players input.


And that is the reason why it may or may not look better.
But definitely plays worse.

1) game running at 60fps, latency equal to 60fps, no artifacts from (M)FG
2) game showing 90fps, latency equal to 45~50fps, some FG generated artifact
3) game showing 170fps, latency equal to 40~45fps, more MFG generated artifact

I would pick the 1st one.

BTW
You may NOT get 240fps from 60fps base using 4x MFG.
Even 5x may not reach 240 from 60.
And using 6x MFG has its consequences.
You forget one crucial element. At higher framerates each frame is shown and seen on the screen for much shorter time.
Thus the time to see the artifacts is also reduced. In order to persistently see it across thousands of concurrent frames, it would have to be something major.

FG works best in a ~70-120 window (native framerate). Enabling it in there increases latency minimally and allows the user to reach their monitors max refreshrate without seeing most artifacts, if any.

Enabling it even from native 60 is not a good experience and neither is using extreme modes like 6x, even from 120 native.
 
And fresh from the other post re the RTX 6060 being 'slower than the PS6'... Well, this is might end up being a large part of why, who knows?

That is, as well as whatever else we might accept as reasons, causes, fair prices or even if this new tech or even the RTX 6060 are any good at all.

All we can tell with any kind of accuracy is that the beatings will continue until... something or other.
 
Back