The PS5s ssd advantage is being way over exaggerated. Sony have a habbit of doing these type of custom things, they implemented GPGPU physics calculations on the PS4 which did not amount to anything superior. There 4kcb tech on the PS4 pro was rarely used by 3rd parties.
On the contrary here; I'd say their GPGPU calculations actually pushed their PS4 titles to do things you couldn't do on XBO without scaling back on graphical fidelity. The main issue (if you can call it that) for PS4 was that the graphical bar also got raised a ton, leaving less actual GPU budget for GPGPU compute tasks.
I mean, look at a game like Days Gone. Graphically less impressive than something like, say, GOW4 or Horizon, but the sheer scale of zombies in that game (combined with the level of detail at that scale), that's something GPGPU compute was HEAVILY utilized for I'd assume (such as handling the physics calculations of the zombies as they collide and interact with each other and the environment), among other things in that game. People are overstating aspects of the SSDs but we shouldn't undersell GPGPU compute either. In fact, it's the more interesting of the two things for me when it comes to next-gen and how that can influence game design decisions.
NO
Thats no how it works. Its a shared memory bus for all the chips, its not a like a PC where your statement would be true if that were a GPU only bus.
Its shared.
If 50 % of the GAME TIME use access is the slower RAM on XSX, then the bandwidth is 50 % of 560 + 50 % of 336 which is same as Ps5.. Ut does not matter what is being accessed on teh slower RAM at all, could be CPU stuff, could be audio, could be your shopping list.
If the games are less than 10 GB TOTAL and 100 % of the time game is using the fast memory, then yes XSX will have a very big advantage.
Big open world 4k high quality textures and quality audio...maybe not., butw e dont know. Neither do you.
That's literally what I was implying. I know the memory is shared between the chips, and the CPU and GPU have to contest with each other for memory access on the bus. But you are greatly overstating the amount of game time that would be accessed on the slower RAM amount. It won't reach anywhere near 50%.
Both systems are SIMD and MS's implementation of the memory setup is not necessarily hard-wired to the specific tasks they've mentioned i.e you can have non-graphics data in the slower pool and audio data etc. in the faster pool. So most devs will probably optimize to have as much in the faster pool as possible and minimize use f the slower pool for such tasks.
Which, in actuality, isn't so much as "slower" as much as it is less bandwidth for the task. The speed per chip is the same (56 GB/s), but only on six of them, if in fact all six were being used simultaneously for such a task. So bandwidth per chip works out to the same ratio as on PS5. In tasks where the GPU would want the full 560 GB/s rate, it can do so, and with GPGPU-orientated tasks (and physical overhead with the larger GPU), that is an option for XSX that PS5 can't necessarily provide without paring down in other areas.
You're right, I don't claim to absolutely know everything on this front. Neither do you, however. Also in your case you do not take into consideration the same congestion and bandwidth conflict issues with PS5. If your example of 50% game time being CPU-bound to the slower 336 GB/s pool on XSX is true, spread across a similar setup on PS5 at similar physical GB amount (6 GB, 1 GB "lower" bound partitions of six 2 GB modules), then that's six of the eight modules not being used by the GPU, that leaves 112 GB/s effective bandwidth for the PS5 GPU on 4 GB of physical memory for those tasks. Which is why that wasn't a great example on your part, when you do the equivalent example for PS5.
Or, if an equivalent 6 GB of PS5 memory is dedicated to OS and CPU tasks, that leaves 5 chips for GPU. Or, 280 GB/s. Both systems have to deal with the shared bus issues, but implying use cases where you get 50% game time on the slower 336 GB/s bus is kind of ridiculous, especially when you work out the physical GB equivalence (and tasking equivalence) for PS5 to see where PS5's bandwidth would be for GPU operations only.
In that kind of scenario that is likely where XSX's additional GPGPU compute advantages could be leveraged, provided the data it has to work on is suited for types of parallelized tasks.