I'd like to think it could be matched if not surpassed, considering it's a tech demo before the generation's even officially started. And that applies to both systems; if we can't expect games even a year or two from launch offering the level of fluidity in the UE5 demo at the resolution and framerate it ran in, then these systems will probably be more disappointing than anyone could've guessed.
As to XSX having anything to this demo, well I'll just preface this by saying the UE5 demo itself is still just a tech demo. There's almost no true game-like AI, complex physics systems, scripting logic or enemies to deal with. Collision systems I would also assume are pretty light in the demo from what we've seen. If this were a slice from a game actually in development it would've been even more impressive than it already is. Also worth mentioning that MS should've definitely considered showing some type of gameplay demo slice, even just a two or three minute teaser, of a game actually in development (preferably Halo Infinite) showing off the graphical strengths of their platform and some sort of high-speed asset streaming segment, and on the SlipSpace Engine just to dab a bit of salt in the wound. Especially seeing now that the PS5 reveal event's been delayed (to what date we don't even know), it would've been a great marketing opportunity for MS to drop such a demo on, say, Friday, or maybe they're planning sometime next week? They need something aside from the hardware event later this month IMHO.
All that said...I'd probably say something in July since that's the gameplay event. How? Well, firstly I'll just speculate that the fast asset streaming in the UE5 demo, I hardly think it was taxing the SSD's full speed or capabilities. I've seen estimates of around 1 GB/s at most for that segment, so it would be capable on any system with a drive of at least that fast. And yes I'm sure the demo was also using other aspects of the PS5 SSD I/O, but it's not like the XSX doesn't have equivalents for most of those features, regardless of how "powerful" those equivalents might be.
There's always combinations of techniques leveraging the additional GPU power on XSX that could be used to simulate many of the approaches for the streaming segments in the UE5 demo. For example, if the ML modifications to the GPU are robust enough they wouldn't need to stream the native high-quality textures; simply sample lower-quality ones (reduce the bandwidth footprint), and then scale them to a higher resolution (DLSS-type techniques) during processing. There's already many good cases for this including textures upscaled this way looking even better than the native high-resolution ones. There's also clever ways that one could simply have a standard base (or set of base) mesh and texture models for environmental details, the statues, etc., and simply utilize transformation algorithms on the GPU at render time to morph and alter those assets. There's a range to which they can be morphed depending on the original data; I guess some people would call this procedural generation, but it could also be comparable to granular synthesis where your base determines the number of "seed" permutations (my understanding on granular synthesis comes from studying some sound design/engineering).
I don't think it was ever actually debunked. The Epic China guys played back their demo on-stream and watched the video file on the stream itself, but beforehand had ran the demo on a laptop fitting certain specifications.
Confusion in terms of branch communications is nothing new, it's been happening since the 1980s. companies like SEGA became a bit infamous for it during the mid/late 1990s. I'd assume the Epic China guys did in fact run the demo on a laptop fitting a particular specification but during the stream itself they watched recorded footage of that build run on said laptop. Members of teams among the different branches are never really 100% aware of what people on other branches are doing, so that could explain why Tim was a bit surprised by it when the word got around about that demo.
Regardless it's not like he'd be in a position to affirm the validity of what the Epic China guys did, for legal reasons.
Dontero
The fact we don't have random read speed/latency figures on the NAND for PS5 or XSX isn't surprising, but those are very important things. The way I see it is part of Sony's solution was to dedicate a flash channel for each module, so if data is known to be on a given module then you simply select that module and off you go. That doesn't say anything to any latency, but it as a setup does help a lot in increasing random access capabilities of the NAND.
XSX, imho, has a very different setup, and in some ways that setup doesn't facilitate for increased ability in random NAND access quite the way Sony's does (this has to do with modules to channels), but for all we know they could be using NAND modules with better latency figures on random access. They're unknowns for now and will probably remain so for a very long time.
Bryank75
You messed up your figures a lot. XSX uses 2.5 GB for the OS, not 3 GB. PS5 very likely reserves 2 GB for its OS.
For the rest of your calculations, it's just mumbo jumbo. If you're trying to factor out the RAM bandwidth the OS will occupy, first off you need to understand that the OS will spread its physical reserve across multiple chips on both systems (2 on PS5, 3 on XSX) so as to ensure the other processors can have data in the RAM in such a way full bandwidth utilization can still be achieved if required. There's almost no point in trying to pettily factor out the OS physical RAM reserve from overall system bandwidth because the systems aren't putting entire GDDR6 chips to OS data reserves, and it's not like having the OS occupy some RAM modules means those modules are now only usable by the OS altogether.
So by your own calculations you'd have to remove 2 GDDR6 modules from PS5, or 112 GB/s, and by your own logic that'd leave the system with 336 GB/s bandwidth, which is ridiculous. That's not how you account for system bandwidth (doing that I also noticed you assumed because XSX's OS is reserving some space on the lower-bound 1 GB of 3x 2 GB modules that somehow means the upper-bound 1 GB on those 3x 2 GB modules are inaccessible whatsoever which is....just extremely flawed idea). So there's no method you can calculate PS5 having more system RAM than XSX since they both have 16 GB; you can only claim it having an additional 512 MB of usable memory for games, and again that's on the assumption the PS5 OS reserves 2 GB of physical memory.
... but let’s not distract from what you admitted is concern trolling essentially as you are spreading doubt and uncertainty (without much evidence) around something you are not interested in the least beside to put it down.
