I think MSFT should have added the XB1S to the that listing. For the performances it delivers the XB1S SOC is simply ridiculously huge.
Either way it was clear Scorpio was not to be a overgrown XB1 but a new architecture but you did a great find
Middle Generation Console Upgrade Discussion [Scorpio, 4Pro]
- Status
Either way it was clear Scorpio was not to be a overgrown XB1 but a new architecture but you did a great find
Wonder how that would work since render targets can spill over (partial render targets) into DDR3, and there's potentially no way to tell how many unique buffers a developer will render first in ESRAM then swap out to DDR3 and so on (can't just dedicate, for example, xXx MB in DDR3, since # of buffers might exceed that.).
Intercept the ESRAM er... management?
saw this on windows central this morning. http://windowscentral.com/project-scorpio-will-require-xbox-one-emulation
it's just a badly written job description.
I never thought for a second it would be emulating the xo, the virtualization avoids 'all' that.
Is it badly written? I'd trust MS to be able to do that part of their job right, as they deal with the best of engineers. The specific wording is - "We are the Xbox Compatibility team. Our mission is to bring the 360 game catalog to the Xbox One, and Xbox One catalog to Project Scorpio." It's punctuated explicitly, so no confusion. Whatever's going on, the XB1 catalog is only coming to Scorpio via the work of the Xbox Compatibility Team. And the only clarification we have is the line, "Backward Compatibility is built on advanced emulation technology." So really, it's unambiguous in saying they use advanced emulation, and it's implied they need to do this get XB1 games on Scorpio. Which confuses us all as we've been talking about forwards-compatible hardware ever since the x86 nature of the consoles was known! But there it is.
edit: Your link's Update 2 suggests it's like Neo -
edit: Your link's Update 2 suggests it's like Neo -
I guess the post is for someone to work on the emulator, while the Team works on both emulated 360 games and compatibility with XB1 games on Scorpio.
Yeah. But note no-one was really thinking it'd be full-scale emulation; just a matter of how much needs to be done per title, and why. We had similar intrigue with Neo's requirement for devs to rework the titles, and they aren't just plug-and-play (although I think they are when Neo runs in 'compatibility mode').
Will Scorpio have its own discs or will XB1 and Scorpio games be sold on the same disc (like PS4 / Neo)?
Someone should ask this question to Phil. It's simple, really, if Scorpio games somehow end up having their own discs (immediately, after 6 months or 1 year or more), then Scorpio is Microsoft next gen and not what they want you to believe now.
Someone should ask this question to Phil. It's simple, really, if Scorpio games somehow end up having their own discs (immediately, after 6 months or 1 year or more), then Scorpio is Microsoft next gen and not what they want you to believe now.
really depends on what components are used, jaguar cpu, same gcn level, some form of esram/edram then hardly any change to vm, any of those parts changed then the related parts of the vm will possibly need to be tweaked, or updated to make best use of hardware or move from deprecated functions.
So Imo it isn't a per title change.
so right now pretty hard to say Imo, although most people do believe there will be a single pool of ram, but even that is conjecture.
so if it is the same bineries then it's not a next gen machine then? Guess that's one way to decide if you want.
at this point I'm not even sure what difference it makes if it is next gen or not, if they xo and Scorpio can play same games as has been said many times already.
Well that is damage control as they don't want people to see Scorpio as a new system (which I think it is).
XB1 games are not abstracted so they can be run on other hardware otherwise they would have passed on a cloning Durango @14/16nm.
Most likely they will develop a virtual machine it should not be difficult compared to what they achieved for BC in past, more like a walk in the park: the CPU and the GPU are going to be really close relative of the ones in Durango.
Imho MSFT should not have unveiled Scorpio that early, it is stupid, they will deal with that type of issue till launch, so for a while and as the whole thing smells of if not improvised but of an accelerated decision process, lets hope there are no delay to add up on what could turn into a complicated situation.
hadn't thought of that, and sounds viable.
not sure how they could actually market that though.
some games work on xo others don't.
could mean only the xo games need to be put through the lengthy qa process though, which would be a big saving in time and resources.
I don't think they will be able to emulate the 360, though you are right to assume that faster PU should translate into improved BC performances.
D
Deleted member 11852
Guest
Where have I seen that type of phrasing before. Oh yes: "We are the Borg. We will add your biological and technological distinctiveness to our own. Your culture will adapt to service us.”
Logical fallacy here. Scorpio won't have 320GB/s because it lacks EDRAM but because it is needed to tackle 4kish rendering.
Scorpio is likely to pass on ESRAM because the size requirement for the various render target @4ish resolution make its unpractical nabd because there are memory solution that can satisfy the need of the system both for the memory speed and capacity.
She could have simply said no comment or that it will be revealed at a later date if she did not want to reveal something MS did not want her to. But she specifically said no Zen in semi-custom solutions until 2018. Thats pretty much a clear cut statement. Now if she was misquoted or the quote is a lie thats one thing. But the quote itself is not confusing whatsoever
As for valid reasons to use Jaguar, well size and price are by far the most valid. Su has stated Zen is a premium product, Jaguar was a dirt cheap netbook CPU like Atom, just because AMD can scale it down doesnt make it cheap. Core M is not exactly cheap for example. Dresdenboy from AT forums calculated Zen core size is roughly 4,9 mm2 at 14nm LPP, in comparison Jaguar core is 3,10 mm2 at GloFo 28nm
If MS is going for more CUs but lower clocks then that will be a big GPU for a console, 14nm LPP is smaller and denser than 16nm+ so its not like TSMC process would make a size difference.
Your RAM comparison is flawed since they talked about bandwith speed. That makes it easy to deduce wich RAM they will be using. They revealed nothing about the CPU other than amount of cores, either because their plans are not finalized yet or because its the same CPU
Both of you are of course assuming you'd want to use esram for holding multiple 4K buffers.
No mention of ram was made, type or speeds. Just an amount of bandwidth that will likely change closer to release. But evidence and logic does lean heavily towards a pure GDDR solution.
It's just that, well, MS has a knack for doing things people don't expect.
I don't think esram would be in such a size for 4K native. But that doesn't imply it couldn't exist. 32MB esram would be sufficient in providing backwards compatibility as well as providing a very fast scratch pad for developers; at the cost of what I assume some minor die space in the grand scheme of things.
Only developers that have shipped top tier titles truly know the ins and outs of embedded ram. If the benefits outweigh the cons I don't see a reason to exclude it unless the sacrifice of computational power was too much.
Sent from my iPhone using Tapatalk
they didn't say anything about technology in regards to any components, bandwidth for ram, cores for cpu, terraflops for gpu.
so your conclusion is pure speculation and maybe they've decided and it's new tech.
it was an early teaser of what to expect, not the launch reveal etc, purposely vague on every component. Maybe they've decided everything, maybe nothing. (although bit late for the nothing)
That's an interesting idea, sram is supposed to be a structure that shrinks extremely well, maybe 32MB on 14nm could be only around 25 mm2?
It's not as big of a compromise as it was on XB1.
I suppose... assuming same bandwidth & ROP interconnects. Not sure how complicated things would get if they wanted >32ROPs attached.
- Status
