News & Rumors: Xbox One (codename Durango)

Status
Not open for further replies.
this was posed by Don Mattrick today. https://twitter.com/donmattrickxbox

345mn15.jpg

seems like PR mangers do read the forums. anyways, made a promise to lay off of the posting till the 21st. so, catch yall later.

"President, Interactive Entertainment Business, Microsoft. Parody Account."
 
Indeed..............and now begins the long weekend. Prepare for anything :oops:

I'm working all weekend so you guys better come up with some stuff to argue about so I can get through the day. Weddings are much more expensive than new consoles for some reason so those 2 jobs Ken said I had to work aren't going to cut it !
 
Can you tell me what the Figure number is for the diagram? My count seems to be off.
At least one of the diagrams seems to be a description of the layout and redundancy scheme for R600.


The vector scalar processor in patent US7577869, figure 10 is almost a decade behind what we know is GCN 1.0, and it has a transcendental unit that the rest of the leaked data explicitly said wasn't in Durango.

Which parts of which sources are we cherry-picking?


Just my opinion after reading some of the AMD/ATI patents that are linked to each other..

Method and System for Load Optimization for Power - http://www.google.com/patents/US201...a=X&ei=sAaXUe_MG-ToiAeuuoHYCw&ved=0CF8Q6AEwBg

Integrated Vector-Scalar Processor - http://www.google.com/patents/US201...a=X&ei=sAaXUe_MG-ToiAeuuoHYCw&ved=0CHQQ6AEwCQ

Apparatus with redundant circuitry and method therefor - http://www.google.com/patents/US201...&sa=X&ei=xFSWUZKME4KYPa37gKAD&ved=0CDcQ6AEwAA

Device Discovery and Topology Reporting in a Combined CPU/GPU Architecture System - http://www.google.com/patents/US20120162234

(plus other linked ones linked to the inventors)

Just my interpretation of what GCN 1.0 is based on the gcn white paper and what is defined in these patents, anyway we'll find out in 3 days .. I still believe Marc Berry is on the right track
 
Sorry to interrupt this stimulating VSP conversation, but MS expects next gen to increase in console sales by 28%.

ProjectedConsoleGenerationalGrowth.jpg


http://news.xbox.com/2013/05/x360-aaron-greenberg-industry-growth

With Wii U dead on arrival, I guess they expect Durango (and PS4?) to pick up the slack, BIG time.

Let's assume they figure Wii U tops out at a very generous 30M. That leaves 355M consoles sold between PS4 and Infinity...

So is that MS hubris or a complete misread on the market? Either console maker will be lucky to even repeat their current gen's performance.
 
Let's assume they figure Wii U tops out at a very generous 30M. That leaves 355M consoles sold between PS4 and Infinity...

So is that MS hubris or a complete misread on the market? Either console maker will be lucky to even repeat their current gen's performance.

28% growth by 2016? That's very ambitious.
 
Just my opinion after reading some of the AMD/ATI patents that are linked to each other..

Method and System for Load Optimization for Power - http://www.google.com/patents/US201...a=X&ei=sAaXUe_MG-ToiAeuuoHYCw&ved=0CF8Q6AEwBg

Integrated Vector-Scalar Processor - http://www.google.com/patents/US201...a=X&ei=sAaXUe_MG-ToiAeuuoHYCw&ved=0CHQQ6AEwCQ

Apparatus with redundant circuitry and method therefor - http://www.google.com/patents/US201...&sa=X&ei=xFSWUZKME4KYPa37gKAD&ved=0CDcQ6AEwAA

Device Discovery and Topology Reporting in a Combined CPU/GPU Architecture System - http://www.google.com/patents/US20120162234

(plus other linked ones linked to the inventors)

Just my interpretation of what GCN 1.0 is based on the gcn white paper and what is defined in these patents, anyway we'll find out in 3 days .. I still believe Marc Berry is on the right track

Your right that it isn't GCN.


From what I can tell, has to do with AMD's older arch's (VLIW ones) that used vector processors instead of what GCN uses (scalar processors, that can be arranged as vectors).


Has nothing to do with your/marcberrys claim.


as above.

Next time it would be useful if you could also link/tell us the relevant part of the patent. these are long not everyone has time to read the entire thing.
 
I will not get into a war of words with you, 1 day ago you did not know about the VSP, so that's a waste of time. All I can do is show you the door and you have to walk through it. The fact that you did not add in the VSPs tells me alot about your hate for MS. Why? MS spent a lot on the 360 GPU. At the time it was the 2nd most powerful GPU on the market. The GPU had the 1st programmable hardware tessellation unit as well. The GPU was the only one with unified shader architecture. Why?
The Queue Diagram you linked to is not for Durango its for another console.
I know, compare the vertex grouper tessellators (VGTs) to that one. Their is 1 in each SC in the patent GPU and 2 in that one.
In 85 hours you will be on my side or you won't.
On my side there already is:
VSPs
VGleaks
Patent
Yukon docs
Was SuperDAE ?
and soon May21.

GPU: Use this to help.
sc_durango1.jpg

12 SCs that have,
4 SIMDs inside each. (4SIMDs: this will not be the name of this 2nd step, it will be called CU array (CUA).)
4 VSPs inside each one of the CUA . The patent shows their are 4 16wideSIMDs inside each VSP with 64ALUs on 16 OPS * 4 VSPs = 64 threads/clock * 4 CUAs = 256 OPS.
12 SC * 4 CUA * 256 threads/clock = 3072 ops/clock
3072 ops/clock * (1 mul + 1 add) * 800 MHz = 4.9TFLOPS or 8x that of the 360GPU. This is without the 360soc 22nm chip.
We will know after the first DEMO.
See you all after MAY21 at 10AM

OUT!
 
I will not get into a war of words with you, 1 day ago you did not know about the VSP, so that's a waste of time. All I can do is show you the door and you have to walk through it. The fact that you did not add in the VSPs tells me alot about your hate for MS. Why? MS spent a lot on the 360 GPU. At the time it was the 2nd most powerful GPU on the market. The GPU had the 1st programmable hardware tessellation unit as well. The GPU was the only one with unified shader architecture. Why?
I know, compare the vertex grouper tessellators (VGTs) to that one. Their is 1 in each SC in the patent GPU and 2 in that one.
In 85 hours you will be on my side or you won't.
On my side there already is:
VSPs
VGleaks
Patent
Yukon docs
Was SuperDAE ?
and soon May21.

GPU: Use this to help.
sc_durango1.jpg

12 SCs that have,
4 SIMDs inside each. (4SIMDs: this will not be the name of this 2nd step, it will be called CU array (CUA).)
4 VSPs inside each one of the CUA . The patent shows their are 4 16wideSIMDs inside each VSP with 64ALUs on 16 OPS * 4 VSPs = 64 threads/clock * 4 CUAs = 256 OPS.
12 SC * 4 CUA * 256 threads/clock = 3072 ops/clock
3072 ops/clock * (1 mul + 1 add) * 800 MHz = 4.9TFLOPS or 8x that of the 360GPU. This is without the 360soc 22nm chip.
We will know after the first DEMO.
See you all after MAY21 at 10AM

OUT!

Its a waste of time yes, because you keep aruging against simple maths.

You are also misintepreting things, as I have shown and others have shown you are incorrect, and quite frankly its getting annoying all you are doing now is spitting fanboy rubbish in a very incoherent manner, hell you don't seem to have mastered the use of paragraphs yet.

Now lets address the other issues.

VSP's. Vector Scalar Processor. A Vector of Scalar Processors. Now that was easy wasn't it.

Vgleaks. As shown and pointed out by me many times, you are simply cherrry picking which part of vgleaks you believe. After all I have shown how silly it would be to have these magical VSP's that are 4x as powerful, yet reduce cache sizes by 4x.

You also ignoring that vgleaks displayed the total power of the GPU and it was 1.2TFLOPS. So you are saying that the very souce you are using as evidence is incorrect can noone else see the insanity in this?.

Patent. This is completey unrelated to GCN at all, it has to do with VLIW architectures and making the yields better using redunant ALU's.

Yukon docs. These are older then the information vgleaks has, by far. It was a target.

superDAE. One of your sources/proof/evidence has flipped there stance, if anything that should show you something

Now for the GPU.

12SC's. Yep we agree on that
4 SIMD. Yep we agree on that
4 VSP. Here is where your going wrong and so did vgleaks/your intepreting it wrong.

From here on out the rest of your maths is also incorrect as you have interpreted things incorrectly.

Heres more evidence that your way off the mark.

Xbox 360 GPU = 240 GFLOPS. http://en.wikipedia.org/wiki/Xenos_(graphics_chip)

240GFLOPS * 8 = 1.920TFLOPS.
Thats less then 1/2 of your crazy prediction. But more rooted in reaility.
 
I'm working all weekend so you guys better come up with some stuff to argue about so I can get through the day.

eastmen, I doubt very seriously if there will be a lack of debate around here over the weekend.

But seriously, SuperDaE and Xbox are now tweeting each other. I wonder, I really do. It could be the biggest sucker punch since the WoW funeral massacre.
 
So what have we learned so far.
Nothing.
1. Most did not know what the hell VSPs are.
You still haven't shared that with us. Incorrect links don't count. You must mean the "Vector Scalar Processor (primary math unit)" from Fig.1 here.
2. This is not the same old GCN we know of.
3. SC is not a CU.
I seriously hope we're not going to rehash this.
4. Each SC has 4 SIMDs. VGleaks. More later on this one.
5. Inside each of those SIMDs have 4 VSPs VGleaks. More later on this one.
6. Inside each of the 4 VSPs are 0-4 16wideSIMDS. (see last post)
SIMDs within SIMDs! Yo dawg, alert the meme generators!
SuperDAE " no the PS4 is better." said the man going to jail.
Wait, he's going to jail for leaking fake info? Not sure how jail hurts his credibility. In fact, leaking NDA'd info seems worthy of some legal action. That would only enhance the leaker's credibility, no?
vgleaks:
Instruction issue rate
12 SCs * 4 SIMDs * 16 threads/clock = 768 ops/clock
FLOPs
768 ops/clock * (1 mul + 1 add) * 800 MHz = 1.2 TFLOPS
YES! We got through to you! I'd almost feel like we accomplished something if you hadn't come back for a second round of acronym theater. Or if you didn't keep using ops and threads interchangeably, and especially CUs and SCs (since according to you they're not the same thing).
next:
Instruction issue rate
12 SCs * xxxxxx * xxxxxxxx/xxxx = xxx ops/clock
FLOPs
xxx ops/clock * (1 mul + 1 add) * 800 MHz = xxx TFLOPS
xxxx, this sounds xxx!
The next xbox is not GCN1.0.
I'm honestly glad Beta and 3dil are in on this conversation. Maybe they can help me understand how anything you've posted supports this assertion. (Durango turning out to be GCN1.1--or whatever 7790 is--doesn't count.)
 
Wait, he's going to jail for leaking fake info? Not sure how jail hurts his credibility. In fact, leaking NDA'd info seems worthy of some legal action. That would only enhance the leaker's credibility, no?

Microsoft sent FBI to buste him, his last rant was "i like ps4 more", but ok, you've no doubt about the credibility, maybe we shoud have doubt in your logical capability or in your fairness.

I'm honestly glad Beta and 3dil are in on this conversation. Maybe they can help me understand how anything you've posted supports this assertion. (Durango turning out to be GCN1.1--or whatever 7790 is--doesn't count.)

I'm glad too that there're are eta-beta in this thread, because I've saved all his claims about the "impossible to up clock" "wishuful thinking" "1.5-1.6 TF via overclock? ridicolous" "impossible to change anything about memory" "no way, bla bla"
I've saved all the replies for some few days

funny that he says to marcberry "you are simply cherrry picking which part of vgleaks you believe." because he's doing the same thing.
Anyway if he is your support, good luck
 
The fact that you did not add in the VSPs tells me alot about your hate for MS.
Skepticism is hate?

The patent shows their are 4 16wideSIMDs inside each VSP with 64ALUs on 16 OPS * 4 VSPs = 64 threads/clock * 4 CUAs = 256 OPS.
OK, so going from the patent I linked above (liquidboy's "Method and System for Load Optimization for Power", ), you seem to be basing your love for MS (judging by the opening quote) on Fig 1/1A and the following paragraphs from that patent:

US 20120019542 A1
By way of example, FIG. 1A includes an illustration of a shader pipe interpolator master (SPIM) module 100, along with shader pipe interpolator slave (SPIS) modules 102, and a single SIMD 104. The SPIM module 100 controls the operation of the shader engine (discussed in greater detail below). The SPI slave modules 102 are interpolator controllers and manage work items within the shader engine. Additional details of the SIMDs are also discussed more fully below.

A component within the SIMD that is crucial to delivery of the mathematical horse power is the ALU shader pipe controller (SP4). In the illustration of FIG. 1A, a single SP4 (0—2) and an SPIS (2) form a quad pipe 105. The SP4s serve as the scalable central ALU.

Each SP4, such as SP4 (0—0), contains four Vector Scalar Pipes (VSP) 106. Each VSP, such as VSP 108, can perform four scalar operations per clock, 16 scalar operations per instruction and hence total of 64 scalar operations per SP4.

Instructions are issued to each SP4 of SIMD 104 over four clocks. Each SP4 outputs four data set per clock one each from VSP to a shader export module (discussed below). The VSP is the primary math unit comprising of staging registers, logic, four multiply and accumulate (MACC) units and a transcendental engine. Each VSP also has four sets of General Purpose Registers (GPRs). Each MACC+GPR pair reads operands per clock to execute one scalar operation per clock.

So, a quad pipe consists of an SPIS and an SP4. An SP4 contains four VSPs. Each VSP can perform 4 scalar ops/clk, and 16 scalar ops/instr. Note the difference in labels! Is this the disconnect between your numbers and my understanding? Are you skipping from 4 scalar ops/clk to 16 scalar ops/clk, without realizing that it takes a VSP four clocks to achieve those 16 scalar ops/instr?

This sounds like GCN to me (a vec4 is processed over four clocks in one ALU), but what do I know?

Edit: Not much! 3dil has already said this patent refers to VLIW4/5, and Jawed says it's not GCN.
 
source: Larry Hryb

https://soundcloud.com/xbox/show-476...aming-industry

Special Guest: Aaron Greenberg, IEB Chief Of Staff shares numbers about the health of the gaming industry

Update: There probably only CoD on 21st, every other game will be at E3 with 'tons of exclusives' and 'world premieres' and with 'lot to share between E3 and the holidays'.

Update²: seems my wording made some misleading, there won't be 60 minutos of CoD, lol, 'cmon guys, they just said that CoD will be there with gameplay but other games will be saved to E3

so the reveal day is mostly for hardware and services? good to know
 
I'm glad too that there're are eta-beta in this thread, because I've saved all his claims about the "impossible to up clock" "wishuful thinking" "1.5-1.6 TF via overclock? ridicolous" "impossible to change anything about memory" "no way, bla bla"
I've saved all the replies for some few days

funny that he says to marcberry "you are simply cherrry picking which part of vgleaks you believe." because he's doing the same thing.
Anyway if he is your support, good luck

Really, then present them to me. Because im pretty sure I qualify most of my statements with words like unlikey, or due to x.

I don't really like making statements that are absolute because some things can change. It is pretty unlikely that the memory width is going to change because this requires a new memory controller which would require them to redesign the APU (this takes time).

but they could use either higher density chips, or a higher clock rate on it although, although this make decrease yields/increase BOM.

They could most certainly have a upclock on the GPU, and it would make sense to but once again things like heat and power may prevent it to a certain degree. (1.5TFLOP is 25% overclock) this would introduce more heat, decrease yields and also not play nice with the eSRAM (the speed of the eSRAM is tied to the clock speed of the GPU).

I don't cherry pick the vgleaks article, i take it all as a truth, im just interpreting one part differently to him.

On the other hand he takes 90% of the article as incorrect and wrong, and says that only one part is correct.

Which is more likely?.

I know you like to assassinate peoples character but atleast be more subtle about it.
 
Let's assume they figure Wii U tops out at a very generous 30M. That leaves 355M consoles sold between PS4 and Infinity...

So is that MS hubris or a complete misread on the market? Either console maker will be lucky to even repeat their current gen's performance.

yeah, the wii u flopping so much makes that target very aggressive. it means basically ps4 and 720 now have to sell > ~150m each.

I'm an optimist and can almost believe in it, though. I dont think hardcore gaming is in trouble. Everybody expected wii u to do poorly imo it's not a cause for alarm.

GAF typed up some more of Greenberg's comments about it. Seems they are counting on emerging markets.

It's not just going to be penetration growth this next generation, we think it's going to be global growth. So we're seeing a lot of markets around the world that are really booming. Latin America for example is a huge growth market, we're seeing growth in Asia, obviously China is a big market. So there's a lot of growth in not just the functionality of the devices but also more countries around the world bringing in consoles. But even with that, what I can tell you is it's just a projection. It won't be right, but I think it's a good sense of where we see the next generation going.

Also maybe this is why MS focusing on entertainment

EntertainmentRevenueGrowth.jpg


ComparingConsoleGenerations.jpg


By MS slides contrary to doom and gloom gaming is a huge bull market lol.
 
"President, Interactive Entertainment Business, Microsoft. Parody Account."

that may be a parody account, but aron greenberg and major nelson real twitter accounts have definitely watched gaf. in fact greenberg took a gaf made avatar on twitter and then major nelson asked for one too (but i just checked and apparently never followed through).

dd41e8c4c61fa68575c864fdd4a960d6.png
 
source: Larry Hryb



so the reveal day is mostly for hardware and services? good to know

no, it makes sense that the only game already known and announced is the one greenberg mentioned. it wont be just cod imo.

greenberg cant say "We're gonna show COD, the new PGR5, tease a new Gears, etc etc" because all that stuff except COD would still be secret.
 
Microsoft sent FBI to buste him, his last rant was "i like ps4 more", but ok, you've no doubt about the credibility, maybe we shoud have doubt in your logical capability or in your fairness.
I was questioning the logic of the FBI going after someone NOT leaking NDA'd info. In no way did I specify or even imply what I believe about SuperDAE. I don't even remember what he leaked!

Let's say SuperDAE leaks that XBN has x CPU, y GPU, z RAM. MS sends the FBI after him. Would they go after him because he's leaking random info, or because he's leaking actual, NDA'd info? Logic dictates the latter, assuming the FBI is going after him about his leaks. If the argument is that he's under investigation for something other than his leaks and therefore that lessens his credibility, I might buy that. But wasn't whatever he leaked corroborated by multiple sources, according to Digital Foundry? Was he shown to be lying about his leaks?

I'm not sure where fairness comes into it. Are we back to skepticism being hate? I admit I'm very skeptical of marcberry's posts because I don't understand his math and I don't believe the sources he has chosen to share with us support his ideas. Not sure my skepticism qualifies as unfairness. I'm not going to treat all rumors as equal. XBN using GCN? Sounds plausible. XBN quadrupling GCN SIMD width (4x the compute power without attendant texturing or ROP upgrades)? Especially when PS4 isn't doing anything like that? Color me skeptical, rightly or wrongly. And maybe explain how such a change will mesh with future games instead of calling me names.
 
Betanumerical said:
Which is more likely?.

I think it will end that we are all wrong and we all are partially right, nobody has the whole truth, this is why is funny for a tech lover to speculate over it.
some speculates with more salt, some with a lot of salt as 3Dilettante, shifty etc, some less, what transform a funny and good discussion is someone that want to make war, someone that say something as "you monkey, don't understand a "£$"%, your english is pathetic, you're ridicolous, what you say is moronish, give us your credentials, tard, you're high" and such approach
Shifty is right, there's some that weights more as argument, yukon, dae/vgleaks etc but this is still open to discussion, if blizzard or ubisoft employers says something is changed in memory/specs, you can believe they or not, I'm fine with this but you have to be fine with different thoughts too
this is why some users came here, then go away. and this is sad because shifty does a lot of work to talk even with who have little tech background, cudos to him (and a big sincere thanks); then being aggressive push them to escape;
I know that a lot of people on forums is emotionally sensible, who loves one brand of console, who loves a brand of graphic card, who loves a brand of cpu, if all the users will lose the calm who will win?
Some users as me or other have problems with the english written, too (sorry, I know, this is annoying) and it adds huge difficulties when others search an aggressive approach
I would like to give some contributes, some ideas, but just to make it clear, because some few users I (and others) feel the envirorment as hostile

so can we from this point to start the talk with a better SNR and peaceful approach?


no, it makes sense that the only game already known and announced is the one greenberg mentioned. it wont be just cod imo.

greenberg cant say "We're gonna show COD, the new PGR5, tease a new Gears, etc etc" because all that stuff except COD would still be secret.

but he can say, we will talk about COD and other games, not "it's the E3 for games" leaving to understand that the reveal event is not for games..
I don't care anyway about games now, I agree that E3 fits better for those, I hope that they will talk about services and hardware (and no time filler as people that dances the "love xbox dance" of musicians that plays or such things


I was questioning the logic of the FBI going after someone NOT leaking NDA'd info. In no way did I specify or even imply what I believe about SuperDAE. I don't even remember what he leaked!

I think that his leaks was legit, he cracked the microsoft net and stolen documents, so they are legit. after that he selled one dev kit and trying to sell another one on ebay.
of course microsoft sent the fbi... he know the final name and a lot of things that only few people around him knows too...

he said on his twitter, and personally to some persons, me included, that the specs are old. this can means that they are changed or that are old and unchanged, but the vgleak info are old even if they says otherwise

he enjoyed a lot the rumors-storm that he start in vgleaks and his twitter account



I'm not sure where fairness comes into it. Are we back to skepticism being hate? I admit I'm very skeptical of marcberry's posts

just to make it clear, I don't believe his math at all, but it's interesting and worth to debate about, at the same time I feel the hate against some brand in some users
 
Last edited by a moderator:
So what have we learned so far.
nothing. You said VSPs are big and there's a patent about them. Then you cite a patent about redundant silicon that has a diagram with VSP on like the Durango leaks. that can be explained as GCN 1.0 SIMD units that work on scalars. Until you present something that shows the existence of a new SIMD design that incorporates a scalar unit alongside the vector pipes other than a TLA without any literary substantiation, you have presented nothing.

Now no one has posted a counter to this, with proof..
The proof is in the understanding. The counter-argument to yours is solidly backed with coherent reference to the major rumours and leaks. It's logical to favour that interpretation than "some guy said one console is better than the other," especially when a quick consideration of that reference proves it nonsense - that'd mean a mythical VSP would have to have at least 50% more performance than a standard SIMD (to get 1.2 TF up to 1.8 TF to match Orbis), which would mean 50% more ALUs, something scalar units don't provide and aren't useful for.

All that fits together. The idea of VSPs just being GCN SIMD units fits. Even VGLeaks explanation and the AMD patents you linked to do nothing that change the meaning of the VSP unit. Ergo it's for you to find proof that VSPs aren't just GCN 1.0 SIMD units given another name to differentiate them from older architectures.

Cayman VLIW4 = SIMDs > SPs
GCN = CUs > SIMDs > VSPs

As the ALU units in GCN are different to those in Cayman, it makes sense to give them a different name. Until we have some evidence in a patent or technical document or AMD spokesperson or somesuch saying otherwise, all you are offering is a guess based on an interpretation that doesn't fit with the other information, and just repeating yourself with more and more colours won't change that.
 
Status
Not open for further replies.
Back
Top