Can someone explain to me why "checkerboard" upscaling is better than something like bilinear/bicubic upscaling or lanczos?
Is it only performance, or can you not use the latter in upscaling games?
I don't know to be honest. Hopefully someone else can chime in, cause I too am a litle confused.
The end result is worse than upscaling from 1440p. So I can't imagine that they are workign with a 1440p image. From what I kinda understand, they are rendering something close to 1080p but over a spread out grid of 4x4 pixel clusters. The scaler then alternates? between the clusters on a frame by frame basis, and interpolates the in between clusters using temporal data (previous frames).
That's my bare bones, possibly incorrect interpretation of things which would explain why the artifacts, and why it's not as sharp as 1440p upscaled on PC.
But it doesn't explain why this would be any sort of performance hit on the PS4 Pro. If this is done in hardware, and you are essentially rendering 1080p at any one point, then why can't the PS4 run at 60 FPS and or higher quality at this upscaled checkerboard 4k? They've said that the 1080p mode will include better graphics and or frame rate, if my understanding is correct (above) then that shouldn't change just because of upscaling. Which leads me to believe my understanding is flawed
When the results are better by running at 1440p, I would imagien develoeprs would just do that. With a 4.2 TFLOP GPU they should be able to pull off 1440p at current PS4 graphics settings without too much issue.
I'm confused, but definitely, the technique doesn't stand up even to 1440p upscale don PC. Or at least doesn't appear to.