Seanspeed said:
90hz headsets will be running at 90fps, though. There will be no cheating
Let's clarify some misconceptions here.
There's always going to be "cheating" - the reprojection corrects for ~10.5ms at 90fps, or 33% of latency you'd have otherwise. All demos on Oculus use time-warp (and presumably so do other HMDs), because otherwise, none of them would be close to that magic 20ms barrier @90hz or lower.
gofreak said:
My point was, though, that even at a native framerate (90, or even 120) you're probably going to be using time warping.
See above - but yes, in a nutshell.
The number in ms, is lower at 120fps - but you're still looking at a significant reduction regardless.
This should also answer people suggesting ghosting is somehow related to time-warp (it's not). That being said - these low-persistence displays
DO still exhibit ghosting (at least the one(s) I tested to date), with or without Time-warping.
Indeed, because 45fps is probably too low a threshold to use it like this.It's either 45fps or 90fps, and one is probably a bit too low.
That's just being silly.
With async time-warping your render-updates aren't required to be v-synced below screen-framerate. 73fps will reproject to 90 just as nice as 45 - the difference is you will be correcting for a smaller time-delta in ms, meaning potentially less artifacts to deal with.
I'll also note here that we're still in very early days of exploring possibilities. For instance, on modern GPUs render-pipeline where you go single-buffered/chasing VSync is actually viable(on fixed hw, I doubt this will ever work on PC), which would allow us to hit 20ms treshold on 60fps, without time-warping.