In my opinion, Microsoft played way too safely with the Xbox ONE, Xbox One X and the Series X. They went with DDR3, because it was the surest way to have 8 GB of RAM in their system. They went with UMA GDDR5 in the Xbox One X, because that's what SONY did. Now they're going with UMA GDDR6 in the Series X.
If you really think about it, the Series X is a pretty conservative architecture.
The real question is whether SONY's gamble with HBM will pay off.
Again, it's funny that people have this impression because Sony has been extremely conservative in their approach to console architecture with the PS4. It is not in the same league as PS3 in terms of exotic architectural features; those days are dead. GDDR5 was not necessarily a gamble, just the amount Sony thought they could pack into PS4 was (you can even tell it was a gamble because the 8GB was packed in clamshell mode style; if they planned 8GB at the start then Clamshell Mode would not have been necessary; the bus would have been thicc enough to host eight chips in full 32-bit mode).
I don't know where the HBM talk is coming from but there's almost no chance it will be present. If so, it'll likely be a very small amount (2GB) for OS and background tasks. They aren't going for a split pool towards gaming purposes because that nullifies one of the big advantages of a unified memory architecture. It's also not like HBM has a league of performance over GDDR6; the costs for implementation and scale of the implementation may or may not outweigh what can be done with GDDR6, that's something Sony always considers when designing theirs systems these days.
In terms of memory the only genuine big surprise anyone can expect is persistent memory; its inclusion would outweigh choosing "only" a small pool of HBM over GDDR6, mainly in terms of capacity sizes. A cache of persistent memory around 64-128GB in size operating at ~ 25 GB/s over a DRAM controller would bring significant performance gains for next-gen gaming. Production of scale is already there, future congruence with high-end PC features is already there, the obvious advantages over SSD are already there, the capacity advantage is already there, etc. And at the scales MS and Sony can afford them, the pricing will be advantageous to them, as well.
The only question is, are either of them taking this approach? Is it only one? Because if there's a situation where only one is taking the approach and the other isn't, you're going to see MASSIVE disparities in overall game performance next gen, regardless if other specs match up (which we already know they won't exactly, i.e PS5 with slightly more TFLOPs). But the reality of those advantages are more or less what should be leading people into thinking if Sony (or MS) are doing some radical memory setup next gen, it WON'T involve HBM over GDDR6. It'll be GDDR6 and inclusion of persistent memory (which can also be used as a marketing bulletpoint, i.e can claim "128GB of memory", and they'd technically be correct because persistent memory offers levels of alterability and random access like conventional DRAM).
Let the HBM dream die guys; there's not enough advantage over GDDR6 to justify the price premium, and they could reallocate that price premium towards persistent memory instead (partially), which has obvious advantages over the fastest SSDs married with DRAM performance, in large capacities. That's the route either Sony or MS will go for unique memory setup next-gen.
www.howtogeek.com
www.resetera.com
