Bulldozer Die Shot!

[JumpingJack]

The new design is fully modular(M-Space), in theory they should be able to swap diffrent core types/GPU cores round in almost any combination.

Obviously, all they need is Adobe Photoshop.... (it's a joke people, don't throw stones).

[kl0012]

I don't get the reason of this. If AMD so afraid of its competitors why bother to show anything at all? How are this fake picture can confirm that they indeed have BD taped-out?

[JF-AMD]

Check this two roadmaps :




I think AMD choosen to Show us the die of Orochi. A BD core seem to be canceled, i would say BD1.
It's clear the second roadmap is less old, with thuban and AM3r2 for BD.

So i think Orochi was BD1, Zambezi is BD2, the real BD incoming.
And lynx maintream have a GPU on die, it's not for only low end. !!!

What you are looking at is 2 different roadmaps for 2 different years. In 2008 there was probably no code name for the processor so they used the core code name. Then by 2009 they probably had the name.

[Motiv]

I don't get the reason of this. If AMD so afraid of its competitors why bother to show anything at all? How are this fake picture can confirm that they indeed have BD taped-out?

It was at the glo-fo talk, demonstrating they are well on with the 32nm, rather than an AMD presentation, showing off their tech.

[ohnoitseddy]

Also, JF said AMD was asked to show a die shot (with this I presumed by an important company, not the thousands of curious enthusiasts). If they can hide all the important bits, then why not? It was probably just to show that they have silicon in house, that Bulldozer is real now, rather than just a blueprint. It was just part of a presentation too - it made for a nice looking slide.

Really, don't make this any more important than it is.

[Zibi]

Some intresting about die size. Look at the structures, highlighted in the blue rectangle - this is the same IMC at little normalized size. Orochi look's like Agena - this is big chip compared to Deneb. One Bulldozer module (without L2 cache) is about 80-100% larger than the Deneb core (without L2) and probably has 2MB L2 cache. The expected die size of 45nm orochi is about 450 mm2, but at 32nm it will be about 250 mm2. Too many... I expected that it will be no more than 200 mm2.


Regards. Sorry for my english.

[hurleybird]

except the modules and L2 aren't to scale...

[BatteryOperated]

I don't get the reason of this. If AMD so afraid of its competitors why bother to show anything at all? How are this fake picture can confirm that they indeed have BD taped-out?

Darn. Another, evil, nefarious corporation not bending to the demands of a forum of impatient enthusiasts.

[freeloader]

Darn. Another, evil, nefarious corporation not bending to the demands of a forum of impatient enthusiasts.

I detect the sarcasm in your post, but the poster does have a point. Intel is at the point in Sandy Bridge where it's a little late to add features to take on BD when it comes out. Show the real deal and let your cards fall where they may.

[-Boris-]

I detect the sarcasm in your post, but the poster does have a point. Intel is at the point in Sandy Bridge where it's a little late to add features to take on BD when it comes out. Show the real deal and let your cards fall where they may.

It isn't about putting stuff in SB, It's about adding features to the architecture one or two steps down the line.

[BatteryOperated]

I detect the sarcasm in your post, but the poster does have a point. Intel is at the point in Sandy Bridge where it's a little late to add features to take on BD when it comes out. Show the real deal and let your cards fall where they may.

I remain as unsympathetic as ever. And allow me to tell you why.
Each time a snippet of information pertaining to Bulldozer or Fusion is released, an veritable avalanche of bad information floods the relevant hardware forums. What follows are numerous threads of ideas, counter-ideas, theorizing, and plethora of ideas from enthusiasts who confidently assume they can predict the future.

Why, on the surface this appears to be good. Everyone is thinking and involved. The downside, as I see it, is that for every mistaken guess, bad idea, and incorrect assumption, is another day members of AMD's marketing staff, like JF-AMD, has to spend debunking craziness.

[JF-AMD]

Well, I am actually the only member out there debunking and 90% of my forum work is after hours. So, in terms of "AMD resource" it is pretty small.

I am releasing details because server customer are making their buying decisions about next year right now. That is why you don't see client info.

Server customers don't need die shots, which is why I was against putting that out there.But I don't control all of the information. Had I not been in meetings solid yesterday I could have gotten to the die shot comments earlier before it started to take off.

I see very particular patterns of people who are deliberately spreading FUD, over and over. I can only guess about their motives.

[chew*]

I look forward to taking a real die shot of bulldozer

[blindbox]

I can't believe you did that!

You should've done it on Cypress instead!

[Chumbucket843]

It isn't about putting stuff in SB, It's about adding features to the architecture one or two steps down the line.

which started several years ago. what's your point? i'll say it as many times as i need to. architecture plans are not directly correlated to competition. this is particularly true with intel. they do a lot of research independently.

[thomasxstewart]

vondrashek md

[HelixPC]

Hope these are good, if they are, i will be making the switch back to AMD!

[JumpingJack]

Some intresting about die size. Look at the structures, highlighted in the blue rectangle - this is the same IMC at little normalized size. Orochi look's like Agena - this is big chip compared to Deneb. One Bulldozer module (without L2 cache) is about 80-100% larger than the Deneb core (without L2) and probably has 2MB L2 cache. The expected die size of 45nm orochi is about 450 mm2, but at 32nm it will be about 250 mm2. Too many... I expected that it will be no more than 200 mm2.


Regards. Sorry for my english.

Your english is fine ... better than mine actually.

Hans is the real expert on deconvoluting die shots and sizes. As has been beat to death, the die shot has been altered for good reasons.

However, Hans did make a guess based on the io pads, which appear to have escaped the photoshop treatment:
http://aceshardware.freeforums.org/p...40.html#p14040

performing the same activity I estimated about 300 mm^2 for a 4 module BD -- so he seems about right, again assuming that the right structures were selected for the scaling. This may seem large to some, but for the markets that this chip would target the margins will be very good.

[WaterFlex]

some day i try AMD

[Andrew LB]

First of all, I'm quite sure Intel has a very good idea of what AMD is building. And considering the lack of any drastic changes in Intel's lineup... they're not worried.

The fact alone that AMD named the chip Bulldozer, photoshopping images way before release, etc kind of reminds me of how nVidia behaved during Fermi's year long delay.

[JohnJohn]

First of all, I'm quite sure Intel has a very good idea of what AMD is building. And considering the lack of any drastic changes in Intel's lineup... they're not worried.

The fact alone that AMD named the chip Bulldozer, photoshopping images way before release, etc kind of reminds me of how nVidia behaved during Fermi's year long delay.

Yes because intel is in the business of doing drastic changes to their roadmap . Because in semiconductor industry you can make changes in architectures as you please . This is intel not jensen

Also if my memory is correct i recall Ajaidev saying some time ago that intel DID changes to what was "original sandy bridge" in response of possible threat of amd's new architecture and introduced what we now know is the 8c/16t LGA2011 SB... maybe now that has passed some time he can say something about this.

[RaV[666]]

First of all, I'm quite sure Intel has a very good idea of what AMD is building. And considering the lack of any drastic changes in Intel's lineup... they're not worried.

The fact alone that AMD named the chip Bulldozer, photoshopping images way before release, etc kind of reminds me of how nVidia behaved during Fermi's year long delay.

Yup. and thats why Intel hasnt done any changes in the first Athlon era, and then with Athlon 64 era.
However if you look how sandy 2011 looks in comparison to 1355 sandybridge you may think that intel IS trying to move as far as quickly as they can.8 cores, Large caches, 4 channel memory(!) thats pretty insane.

[Oliverda]

The fact alone that AMD named the chip Bulldozer, photoshopping images way before release, etc kind of reminds me of how nVidia behaved during Fermi's year long delay.

nV showed a not working fake card. AMD showed us the highly distorted original die for understandable reasons. Different stories.

[generics_user]

nV showed a not working fake card. AMD showed us the highly distorted original die for understandable reasons. Different stories.

and AMD alway said that this was photoshopped; nvidia claimed the card is the real deal until someone noticed the wood scres

[-Boris-]

and AMD alway said that this was photoshopped; nvidia claimed the card is the real deal until someone noticed the wood scres

Besides, it wasn't AMD who showed it, it was GF talking about 32nm.