Hopefully some here in the know can shed some light on these, but here are my thoughts.
Ok, so Sony has a patent on the tech they use for PS4pro "checkerboard" upscaling or uprendering or whatever. Basically, a native 4k image is made up of 4 1080p images. This method natively renders just two of those 4 images and then reconstructs the other two images from the native pair. The end result is a "native" 4k frame buffer that is made up of 50% native pixels and 50% reconstructed pixels.
Sony went the extra step of customizing their Polaris GPU from AMD with what I'll call henceforth Up Rez Engines (URE) which makes this process have zero cost to the developers rendering pipeline.
It's left to be seen how effective this method really is without any side by side comparisons, but it's my guess that if its done in hardware and not a software implementation then it would yield all round better results than however it's been implemented in the past. We do however have a number of games shown so far that uses this tech; Horizon, Days Gone, ME:A. Tomb Raider....etc.
I find this all somewhat intriguing. GPUs are constantly implementing things in hardware that automates a number of redundant rendering techniques so they don't all have to be done in software. That's arguably the key advantage of a GPU. But never has a GPU been built with hardware to make this pixel rendering technique possible/inherent. And I'm wondering why....
I've looked at screen shots of the Tomb Raider comparison, and while I still stand by my belief that we need more material to make such comparisons, I think it holds up very very well to native renders.
So why are GPU manufacturers not baking hardware like this into their GPUs, it literally allows you create a passable faux 4k render while using only half the power requirement and massive reductions in the required memory bandwidth. In a machine that has the power overhead to do "native" 4k, this would mean that you have enough power to not just make its the prettiest looking faux 4k possible but even to make it run at 60fps.
As I said, it's all still left to be seen, but if from comparisons it's extremely hard to tell the difference... then shouldnt GPU manufacturers at least consider making this an option? Unless of course sonys patent prevents them from doing that.
Checkerboard rendering used in PS4 Pro is exactly this, render the red pixels, reconstruct the green pixels using different samples provided by the engine (you should post this in the OP):