That said, I'm not sure what any of this has to do with XB. Can someone direct me to why we're talking PS hardware here?
A poster said that PS5 has RDNA2 compute units, while Xbox Series has RDNA1 compute units, which is pretty much the opposite of what is in fact the case.
The reason this whole "full RDNA2" bollocks about the PS5 is so maddening is that there's a really interesting relationship between console vendors and PC tech roadmaps. PS5 was supposed to be a contemporary of the 5700 XT, and originally had the same feature set (barring probably cache scrubbers - but that's a feature not directly exposed to developers). Navi 10 is probably shaped as much by PS5 as PS5 was shaped by it.
It's unclear if MS have a relationship as close as Sony do with AMD. The next Xbox(es), if they actually appear, will come after RDNA4 and that's supposed to come with a big revamp of AMD's graphics architecture. If MS have been working very closely with AMD on specific products they may reflect eachother in the way PS5 and Navi 10 did (to the benefit of both). The thing is, MS's plans are all over the place at the moment and I wonder if they are interested in as deep and long term a committment on future hardware as they would otherwise have.
MS are also in a sense burdened by DirectX as much as they are helped by it. They can't just do what's best for a console in a given period, they have to align with DX even if that hurts the console. Just look at PS5 in comparison Series consoles.
PS5 is an older design that got delayed to add RT. In its extended development there was time to implement a clock boosting feature which automagically boosted performance across the board, and there will have been time for many revisions and respins to improve yields. This means more consoles available to cusomers to take market share. As development kits and software were mature, games at launch came in delivering good results.
Xbox needed to have full DX12U support. This meant waiting for the full RDNA2 featureset. MS got initial silicon much later, final silicon later, had poor yields with no time to rework and improve things before release, had to go with conservative clock speeds, launched with undercooked development tools, and had less time to polish early software. On the plus side, they did get amplification shader (which no-one seems to use) and got DP4a for their AI upscaler (which they never bothered with).
Which one of these seems better in retrospect?
So for MS's next machine, what can they do to improve competetiveness and avoid the self harm they inflicted this generation? One thing they can do is to add valuable features to the hardware even if it won't be there on PC - perhaps something like BVH accelleration hardware, or adding API features for an NPU to the Xbox devkit even if it's not there in PC DX. Something else they might consider is getting silicon earlier so they have more time to get it right, even if it means not getting the absolute latest features, especially if these will make little to no difference in the experience players have.
As I say, the problem is trying to put a number on it when that number is not objective. If they intention is to compare PS5 hardware to XBS, it should be done with objective feature comparisons as you've mentioned there.
I agree. But that can only lead you to one conclusion, and it's one that Sony and AMD marketing moved to successfully obsfuscate, generating legions of footsoldiers that occasionally pollute even here.
PS5 has version 1.0 of AMD's RT hardware btw (PC RDNA2 and Xbox are 1.1). And that was added to PS5 late on in development and so is newer IP than the rest of the chip. That shows both the age of the PS5 design, and also that Sony have a very close relationship with AMD's graphics department.