So if I’m understanding everything that 2.5 multiplier of SSD performance on average is talking about the Raw number right? So the average would be 6 gigs per second? Could be higher or lower depending on content though.
Are you speaking about XvA and the "effective multiplier" stuff MS mentioned?
Truth is I don't think anyone is 100% knowing of exactly how it works, MS simply haven't mentioned enough of the setup for us to know. However, based on what I've seen other knowledgeable people here and on places like B3D discuss in great detail, I figure the "effective multiplier is in reference to:
1: Reduced latency access, which ties into...
2: Highly improves/reduces prefetching frame range (potentially to 2 frames or even 1 frame)
So since it'd appear things like SFS want to focus on bringing in just specific texture file assets "just in time" when they're needed, and blending lower-level mips seamlessly into the higher level-one that is streamed in, you need extremely good latency for that. Bandwidth is less of a need here but it can still be important in its own way, it's just that latency is definitely the more important of the two in serving the purposes of something like SFS.
The big question though is just how good is the latency? Average latency of TLC and QLC NAND is something like 100 microseconds. If MS have been looking to reduce that latency drastically, that serves better purpose for things like what the DiRT 5 developer was talking about regarding texture use/discard/replacement in the middle of a frame.
Obviously they can't get latency figures on NAND down to PCM or RAM levels, otherwise they'd of just gone with more RAM or some PCM in the middle (which I was hoping both systems would do tbh; maybe unrealistically). But MS have consistently been talking about latency reduction in their design. I don't know if their Dynamic Latency Input feature plays into XvA as well, but it does go to show where MS's focus is with their SSD I/O: they're prioritizing latency reduction.
With all that said, Sony, ..... I think they will have very good latency with their setup as well, but it's clear they have chosen to prioritize maximizing bandwidth instead. The thing though is that for vast majority of I/O related tasks, the two essentially cancel each other out. That's why it's been foolish for a lot to directly compare the SSD I/O specs between the two systems because they have had different emphasis of paths to roughly similar end-goals from the beginning. MS's is better suited for stacking on top of any number of SSD configurations (both beastly SSDs and more weak ones) and scalability.
Sony's is more pure hardware-orientated to a specific system design, but it lacks scalability and other solutions that want to be compatible have to play by Sony's specifications which raises the minimum on the 3rd party's end (hence why Sony has to certify 3rd-party SSDs for compatibility with PS5; tbf if MS's solution is doing a setup like the poster 'function' mentioned in their B3D post, you'd need proprietary drives with custom firmware to support that, which could explain their partnership with Seagate).
Fuc yeah!
Bodes extremely well for the gameplay visuals there (I'd assume Project Mara was also in real-time then, too). Can't wait to see more of this game.
news.xbox.com
