Zen4 peaks at 5.7 for an ST workload in a 170W socket. That's about 3/4 of what a console consume and then there's the GPU to consider.
There is no way it will run at such clocks. Expect something way lower, like 4.5 or so.
Not really. 45TFs on RDNA3 is faster than 22.5TFs on RDNA2, not equivalent. Due to how this is implemented it is far from being 2X faster and is closer to being 1.25x or so. Still it's not equivalent, and on a console it may end up being better utilized than on PC.
Also not sure what RDNA3/4 means. We don't have anything on RDNA4 yet so we can't just lump them together like this.
For a console which is in development now I think it's unlikely to be using anything but RDNA3 (or even 2, since Sony's b/c is still h/w based for all we know).
2-3x gains in RT are impossible without a similar gain in memory bandwidth. I think you're missing the critical piece there - bigger caches. It is very likely that the Pro will get IC for the GPU (instead of the useless 4GBs of LPDDR5). Even then though I'd expect RT to improve in line with FLOPs gains (accounting for 2 vs 3 differences) since we're talking about RDNA.
All in all the way PS5 can scale is the biggest reason why I doubt that PS5Pro is necessary or a thing even. A "Pro+slim" is an interesting concept though and it does make some sense. Just don't expect any "xX gains" from it, these will likely be in +XX% range.
Thanks man. It seems that one can op for higher bw using gddr6x and gddr7 instead of using infinity cache or vice versa. Since both configs have die size under 300mm2, maybe it makes sense to go for slower ram and 128mb of infinity cache.