Wait...they shipped these to my state? Hmm...maybe I can try finding them...
Jokes aside, pretty interesting. If they really are doing a PS5 Pro, there's still two ways that could go. Either something basically doubling base PS5's TF perf and increasing RT and ML capabilities significantly, or something sticking with baseline PS5 TF perf, pixel fillrate etc. but revamping the GPU architecture with additional hardware for RT etc.
Each would have their advantages and disadvantages...but the latter might be better approach IMHO. The current consoles aren't even being taxed so much in terms of optimization as could, there's still a lot of performance with even those and considering most of that perf is really tied to game scope & budget, we shouldn't see the base systems (aside from Series S) start getting genuinely taxed until maybe 2025.
But in any case, would like to know more about these prototypes and what they are specifically. Hopefully there's a leak or two.
P.S: I hope one of those prototypes is a base PS4-level PlayStation portable. I think that makes even more business sense than a PS5 Pro right now, especially for certain markets like Japan.
I'm inclined to agree. 2023 is aggressively early considering base consoles are still extremely hard to find in stores regularly. It's also why I personally think (tho I have no evidence to back this up) why a future "PS5 Pro" is actually likely to be a base PS5 with a newer GPU design based on RDNA 3 (or possibly RDNA 4) and adding a good deal of dedicated hardware accelerated silicon such as things for improved RT.
Remember, raw TF is not a good barometer for game performance; while some things like mesh shading do benefit from wider raw compute performance, they've also barely been used as-is on Series systems (or PS5, which may not have mesh shaders in name but have something virtually identical in function with their revamped primitive shaders). If you want something with more rasterization throughput for example you don't really need more CUs; if scalability allowed for it you could just add more ROPs and TMUs and adjust clocks to be slightly faster and now you have a good deal of more rasterization perf, and only slightly more raw TF perf, yet you get a lot more immediate boost for game-related processing tasks versus taking the other route.
Another advantage would be that die size could be managed better, which is especially important if they're going to remain on 7nm (again there's nothing showing this, but take it as an example). A PS5 Pro with a GPU doubling the size of PS5's (and the same would apply for a Series X "Pro" doing similar in its case) means now a wafer only gets you probably 1/3 less chips than you'd normally get (the GPU portion of the chip will be bigger but not everything is being doubled i.e the CPU cores for example). A "Pro" refresh adding additional hardware acceleration for specific tasks while keeping the CU count essentially the same would make better use of wafer budgets, which are already being taxed as-is by the electronics industry as a whole.
Didn't the PS4 Pro change GPU architecture from base PS4? Polaris vs. GCN 2.0? And it's not like RDNA 3 won't be BC with RDNA 2, as RDNA 2 is BC with RDNA 1 in spite of certain changes.
I mean I guess RDNA 3 does change up the arrangement of SIMD units, ALUs, etc. and isn't using CUs in the sense that RDNA 2 does, though, so maybe that does constitute as a bigger architectural shift?