Well given that the checkerboard resolution solutionfor PS4 pro is rendering the game twice (once for each half of the APU) and is then combined to make the Checkboard 4K image, is it possible that a game that runs a sub 1080p image maybe doesn't place nice with Sony's solution?
That's not how checkerboard rendering works. A frame is not duplicate rendered. And the two types of rendering combined during checkerboard are not done by "half the APU", they're just run on the entire GPU. Finally, checkerboard rendering's accuracy is not resolution-dependent. (Though the results will appear better at higher resolution, that's because the pixels subtend smaller angles. Each individual pixel is just as accurate as at lower resolutions.)
Why not just run the game at 1440 and checkerboard that out to 2160? ...I mean I guess when checkerboarding is used properly , it's kind of doing this going both ways but usually you get some pretty solid looking still images....
Checkerboard rendering is not a type of upscaling. It can't be applied at the end of the pipeline to boost results; it's integrated with the normal rendering process and is how you get to your initial target. Changes after that must be done with upscaling.
I'm also curious why devs don't attempt a solution like what Guerrilla games did with Killzone Shadowfalls multiplayer back at the PS4 launch. You might recall they were 1080p but... it was 960X1080p stretched out horizontally so the lines were really sharp, the framerate held up better but the vertical columns of pixels got fatter.
That's not what
Shadowfall was doing. Many games have done what you say, but that's just anamorphic upscaling.
Killzone was notable because it was doing something new: not upscaling, but rendering every other column in a different, cheaper manner. This allowed them to maintain a higher framerate in multiplayer with a minimal loss to detail (but some artifacts).
Devs aren't still using that technique because checkerboard rendering is a more advanced, more accurate version of it. Especially the PS4 Pro version, which uses specialized hardware to increase the consistency of the results.
Yeah.. this.
What is it, a portrait oriented resolution?
First, let me add that my pixel counts also showed about 1440x1620, so I can believe it. (I say "about" because there are hideous artifacts all over the image that make counting an incredible chore, and the real results I got very noisy. They hovered around
NX Gamer's numbers, though.)
As for why, it's been quite common for res-constrained games to lower their horizontal resolution and not vertical. This is because artifacts are less pronounced if you're only upscaling in one dimension. And in severe cases, the horizontal res can dip lower than the vertical even though in the final image that's not true. For example, a dynamic game may hit 960x1080.
The problem with Syndicate that makes the results noticeably bad is manifold:
1. The PS4 Pro should be able to render more pixels than 1440x1620. This is less than 1440p, and other titles have no problem hitting that natively, much less in checkerboard as Ubisoft appears to be using.
2. Ubisoft are throwing away the advantage of horizontal-only reduction, because they have to scale 1440x1620 in both dimensions anyway.
3. Whatever scaling method they've chosen for this 8/3 and 4/3 blowup is
awful. This is the biggest culprit in the end results. I don't even know that it's nearest-neighbor, because there's blur and noise everywhere. Maybe they're doing the upscale at a weird point in the chain, and AA is grabbing bad samples? Whatever the reason, it gives abysmal IQ.