|
Pemalite said:
* Checkerboard isn't happening.
* Checkerboard is not as good as a native 4k image.
But can be good enough considering lower specs. Sony considered that, and marketed as a 4k device.
* Double Rate FP16 is pretty much confirmed.
* FP16 is lower quality than FP32.
In which mean? Lower precision of course.
* FP32 calculations are not not being converted into FP16. That would be counterproductive.
* There is still more to performance than flops, people need to stop focusing on that single *specific* number and ignoring the rest of the system.
I know, but we have to agree that velocity of operations of a GPU play a major role in rendered resolution, the aspect I am considering in this thread. Also, I said that CPU of the switch is good enough, so, Im not considering only the Gflops of the GPU as the total aspect of the system.
* Developers are still free to use frame reconstruction to make a "fake" higher resolution image. Which is what Quantum Break did.
* Super Sampling will likely not happen.
Well, generating higher resolution is super sampling. Some methods are more accurate, other fake. The common state of art super sampling methods, the checker board, the QB fake method, everything that is more than upscaling meant to increase the resolution, so, are all super sampling. Im suggesting that would have a hardware native super sampling method. If developers can do, would be nice if the hardware would make that natively and optimised in hardware by default.
* The Switch is likely going to be a 720P or lower device for games that push hardware.
A hardware push game could be initial rendered at 540p, 360p, or so, than a resolution increasing method native in hardware would do the rest.
* The Switch's performance is still between the Wii U and Xbox One. That hasn't changed.
Yeah, it woundt change the power of the device. Its a way to show better images from less hardware. Still far from xbox one.
There. I think I covered it all?
|
