Xbox One (Durango) Technical hardware investigation

[dagamer]

I am confident that both MS and Sony will have very well thought out business models that fit into their vision and long term strategy. From what we know the hardware decisions seem perfectly acceptable to me, the machines appear to be reasonably balanced in terms of computational and graphical power and should be available at reasonable price points.

I am little curious about 3rd parties and their approach this next generation, in many ways it reminds me of the line in Karate Kid "walk on this side side ok (low cost, highly innovative, social gaming), walk on this side ok ( Hollywood budget, big explosions, story and character development) but walk in the middle crushed like grape....".

With many games disappointing in sales, I don't think they continue making such spec tales and getting disappointing results. Tomb Raider was a great game, but Square-Enix was disappointed. Sleeping Dogs was a great game, S-E again disappointed. Crysis 3, Dead Space 3, God of War: Ascension, Gears of War: Judgement, all shadows of previous gams in the series. This can't keep going on forever.

 

[bagofsuck]

Can someone explain me the differences between

12 CU array vs 12 CU vs 18 CU ?

 

[SlickShoesRUCrazy]

Can someone explain me the differences between

12 CU array vs 12 CU vs 18 CU ?

12 cu array would actually be 48 CU if I have understood everything correctly.

 

[bagofsuck]

12 cu array would actually be 48 CU if I have understood everything correctly.

I am asking because at semiaccurate forums some users were talking about this and that vgleaks said the xbox gpu has 12CU array?

 

[Betanumerical]

I am asking because at semiaccurate forums some users were talking about this and that vgleaks said the xbox gpu has 12CU array?

It wasn't CU array and if it was they made a little mistake, after all they also showed us the TFLOPS so unless for some reason the CU's are massively weaker then standard GCN CU's..

 

[Nisaaru]

I flipped the image over on it's side to get a better look at it, This is definitely AMD's tech, but I've never seen them use a diagram designed like THIS before. This can't be VI as some people have been claiming though. It has DDR3, not DDR5. AMD would not be scaling back bandwidth for Hawaii or Curacao. And then you have a security co-processor? Very curious.

It's an APU based on VI for server purposes. You wouldn't use DDR5 there. As I suggested here before a scaled down version could very likely be Durango as it would explain all these weird rumors(yield,delays(20nm), heat(vgleaks specs make a 100W design), big chip) which just don't match with the vgleaks specs, over clocked or not.
That AMD expects to release a 20nm chip at the end of the year which nobody here expected changes the context. If AMD can do it MS might have decided to do it too as 20nm is expected to be mainstream for NV/AMD next year anyway.

 

[Shifty Geezer]

1.2TF would be xbox tv

So all these devs have been developing games for Xbox TV, not the next XBox? Did anyone tell them this?

Not to doubt you Shifty but it was my understanding that the figures you mention (1.2 TF and eSRAM) are simply the ones provided by VGLEAKS. The specs haven't been, to my knowledge at least, confirmed by developers. And for good reason!

They've been independently confirmed by EDGE and Digital Foundry, who have sources. The information is taken from the developer documents, and was accurate regards what developers were being told is in Durango as of January this year.

They're not just unsubstantiated rumours some website posted, but proper leaks of developer information.

I'm not saying the final hardware will be as described by the VGLeaks documents, but discussion of Durango's hardware surely should be based around leaked technical documents describing it rather than pie-in-the-sky, unsubstantiated rumours (which is what the 'overheating' and 'super-powerful console' etc. are). If we apply cold, hard logic, the next XBox will be revealed as 1.2 TF GPU + 8 core Jaguar + eSRAM etc (if any specs are given). If the next XBox is radically different from that, it won't be because people trusted unfounded rumours but because MS have pulled off some amazing conspiracy/ruse, either getting all the devs talking behind closed doors to lie, or feeding them fake information to obscure the real Xbox product.

 

[Rangers]

and was accurate regards what developers were being told is in Durango as of January this year.

one thing i'm not sure of is where you get that January 2013 date.

the vgleaks specs are valid i believe, but there's no conclusive way to solidly date them that i've heard.

These tags, must be just winding me up

12gb, 15% overclock, 17.3% overclock, 384-bit,

384 bit doesn't seem viable, would require total redesign right? 8GB/68GB/s can only correspond with 256 bus I assume.

 

[Brimstone]

So all these devs have been developing games for Xbox TV, not the next XBox? Did anyone tell them this?

xb360 .25 TF

xbtv 1.25 TF or a modest 5x over the 360

xb720 5 TF or 20x the 360 performance hence the $500 price tag that is hinted at.



Why would a 1.25 TF console using ddr3 be $500? Of course the $500 rumor could be bogus.

 

[Betanumerical]

xb360 .25 TF

xbtv 1.25 TF or a modest 5x over the 360

xb720 5 TF or 20x the 360 performance hence the $500 price tag that is hinted at.



Why would a 1.25 TF console using ddr3 be $500? Of course the $500 rumor could be bogus.

Well it could be that they want to turn a nice profit, and including nice storage mediums. I do not think a 5TFLOP monster would be practical, heats a a thing, and for APU's its even more important.

 

[SlickShoesRUCrazy]

xb360 .25 TF

xbtv 1.25 TF or a modest 5x over the 360

xb720 5 TF or 20x the 360 performance hence the $500 price tag that is hinted at.



Why would a 1.25 TF console using ddr3 be $500? Of course the $500 rumor could be bogus.

doubt the 720 will be 5tflops....even with a $500 dollar price tag, I would think Microsoft would still be losing money....I doubt any of these companies want to be losing money right off the gate on their product.


my estimate has been since the Yukon leak to be between 1.2tflops(vgleaks) and roughly 2.5tflops since the Yukon leaks specifically says 8x the 360 in performance(and yes, 8x 360 is not 2.5tflops, I just lowered it).

 

[dragonelite]

xb360 .25 TF

xbtv 1.25 TF or a modest 5x over the 360

xb720 5 TF or 20x the 360 performance hence the $500 price tag that is hinted at.



Why would a 1.25 TF console using ddr3 be $500? Of course the $500 rumor could be bogus.

XBTV that plays 360 games at 1080p for $150~180 im all for it.
5Tflops seems high but a healthy 2.5 Tflops seems oke for at a $500 price point.

 

[french toast]

20x the performance of the 360?? 5TF?? Someones having a lucid dream.

More likely scenario is this.
http://www.edge-online.com/features/microsoft-ceo-steve-ballmer-unveils-xbox-720/

 

[Xenio]

5 TF is too much, but I agree that the final hardware will be somewhere in the 1.2-2.5 TF range
as nobody really know just keep the discussion seriously please.

in the leaks they talk about 8x the 360, but if they are talking about specs, then Infinity will be a 2 TF machine, if they was refering to overall performances, it can be 1.2 TF that have the performances of a 360@2TF.
I can't say "IT IS A 1.2 MACHINE" or "IT'S A 2+ TF MACHINE" because within 9 day I can be hardly backslapped by Microsoft's reveal day,

of course I will very surprised if they can launch an absurd 5 TF machine, I think it will remain a dream for technology nerd (as we are), maybe my possibilities are:

35% 1.2 TF
45% 1.5-1.6 TF
19% 2-2.5 TF
1% 5TF

 

[SKYSONY]

I'm afraid the fact that EDGE contacted various sources that agreed with vgleaks tells us that the leaked 1,2 performance number should end up being quite accurate.

Of course, changes can be done, but performance os not going to be 4x what every rumor claims, not even 2x I believe.

 

[Xenio]

performance os not going to be 4x what every rumor claims, not even 2x I believe.

sorry but when the talking is about the microsoft console, your name.... how to tell..

 

[Grall]

5 TF is too much, but I agree that the final hardware will be somewhere in the 1.2-2.5 TF range

You know, you're precisely right... It'll be 1.2TF exactly, pretty much.

What the price might be I haven't the faintest idea, but I hope the hell it's not $500, that's for sure as it would put a serious crimp in MS's style (the public dislikes expensive consoles, as Sony learned with PS3 and MS gloated over years ago), AND in my wallet I should add. Also, hoping the PS4 isn't $500 either by the way (and both next-gen consoles launching at $400 would make the wuu deluxe at $350 look like a seriously bad deal (which it's been all along one might add)); but those matters are all a topic for a different thread. Sorry for the brief sidetrack.

of course I will very surprised if they can launch an absurd 5 TF machine

Even briefly entertaining the thought they might possibly MAYBE launch a 5TF machine is utterly absurd. Not only does none of their devs know about something like that being in the cards, the hardware costs alone would punt the console into far beyond $500 range. Then add that cooling a monstrous chip like that at anything less than an original launch 360 roar is serious business; look at the ridiculous heatsink in phat PS3s for an example of what would be needed for a consumer device.

It's completely unrealistic; the statistical probability of this occurring is so low it's almost literally negative.

 

[Bagel seed]

Some perspective from someone who is in chip testing with apparent inside info.. back in Feb after the PS4 event:

Nope, everything has been set in stone I can tell you that much for both the PS4 and Xbox.

Inorder for you to really understand what is going on the 1st question to be asked would be ->What do you really understand by the term 'mass production' from specifically semi-conductor stand point?

Pre-silicon world is totally different from the realities of post silicon working.

General consumers are not aware that any ASIC of this complexity along with APUs CPUs etc take 8 months - 1 year in the minimum from the first tape out of the chip. Even to mass produce you need a test program to test chips (my current line of work), this is after the fact that even a minor fix on skews in a logic with a rev spin needs a new GDS and refabbing of that spec'd chip.

Mass production state is achieved after all the logic, each and every block, every tile has been tested. This takes 1 year in the minimum, you change anything in-between and it messes up everything, the whole year's work is to be redone. This is why we have a term called PoR, A team first commits to the design specs once thats locked in its final unless something horrible goes wrong. If its working, that's it. The customer, say Sony, has to get their plans straight. No time to change things at the last moment unless they want to be late by 1.5-2 years again.

You don't just change things and expect to have it working. Change anything on the chip and it has to go through verification. This process has so many phases that are beyond the scope of one post to discuss. But like I gave you the simplified flow a few posts above, it should give you an idea of how much is involved.

People generally have the misconception that mass production means 'making the final chip', They generally assume that you can just throw in a few more shaders or say change cores at the last moment when a company gets to know a rival's plan. General public are not aware of what goes into testing the chips, how chips are tested, how the test program is develop, what is verification, what are the differences between pre-silicon simulation and post simulations, how changing a component can affect critical path and there by affect over all speed of the chip and then time wasted in correcting or compensating the lost speed uplift etc.

The golden phrase in semi-con industry: If you don't test it IT WILL fail guaranteed.

...
I know the actual details of xbox and I knew the details about PS4 well before it was announced. You are under the impression I am sharing an opinion, no it was not opinion. Its information based what I know directly from these projects. I have my means. Both ASICs have their specs locked in and are far advanced into their development.

And even without having the means to such information any design engineer will tell you the same thing that changing specs at this stage is not happening unless some one wants to 1.5-2 years late because they know what goes into making these decision and they also understand people from the outside have a hugely misunderstood concept of semiconductor operations.

From the comments here

 

[SKYSONY]

sorry but when the talking is about the microsoft console, your name.... how to tell..

I'm sure whoever wants to read my post will do it for its content, not for my name. You have been commenting and quoting rumors from the most fanatic sites of the net, and probably are the only one expecting the console be better than the PS4.

What was that brilliant quote you post here a few days ago about arguments and attacks

 

[Strange]

35% 1.2 TF
45% 1.5-1.6 TF
19% 2-2.5 TF
1% 5TF

It should be more like

100% ~1.2 TF
5+5i% 1.5-1.6 TF
-5i% 2-2.5 TF
-5% 5TF

Have to let it add up to 100% somewhere, even with negative possibilities, so