AMD realizes that Bulldozer has 800 million LESS transistors than it thought!

[qubit]

THIS THREAD HAS NOW BEEN SUPERCEDED BY THIS NEWS ARTICLE. PLEASE COMMENT THERE.


So, not only have AMD produced a mediocre processor, it turns out they didn't even know how many transistors it has!

This is another one to file in the "you couldn't make this stuff up category", lol.

This is a bit unusual. I got an email from AMD PR this week asking me to correct the Bulldozer transistor count in our Sandy Bridge E review. The incorrect number, provided to me (and other reviewers) by AMD PR around 3 months ago was 2 billion transistors. The actual transistor count for Bulldozer is apparently 1.2 billion transistors. I don't have an explanation as to why the original number was wrong, just that the new number has been triple checked by my contact and is indeed right. The total die area for a 4-module/8-core Bulldozer remains correct at 315mm2.

AnandTech

[kid41212003]

Fhat the wuck?

[Jizzler]

856,482
856,483
856,484
856,485

Hey Bob! You wanna get lunch?

Oh damn, you made me lose count. Oh well, it looks like 2 billion. I'm done.

[Vincy Boy]

Originally Posted by kid41212003

Fhat the wuck?

My thoughts EXACTLY

[Super XP]

Oh boy, looks like AMD fired all the good guys

[newtekie1]

[kid41212003]

Will their stock drop because of this?

[Benetanegia]

Hmm I'm sorry but I call BS. Damage control. My internal logic tells me that despite being different architectures, there's no way there's such a big difference between Llano transistor density and BD transistor density (66% difference!!). Specially since BD was designed specifically to improve in this metric.

1.2 billion + 60% == we are back to around 2 billion.

I'm skeptic by nature, and considering AMD's track record at blatantly lying about Bulldozer, there's no way in hell I'll believe this change.

Also it's so convenient. Hmm let's see... the (old) PR guy screwed up and provided a number that was wrong and no one noticed it. But ey! No problem because we just fired the PR department. In the several weeks between when this number was first mentioned in previews and the launch/first reviews, no one noticed it was wrong. On the several weeks AFTER the launch no one noticed it was wrong either... As if launch information and PR weren't carefully planned for weeks... NO WAY I can believe this, sorry.

[qubit]

@Benetanegia

Yeah, good point. That's what makes it all the more fail isn't it?

Now just imagine if there were actually 800 million more transistors than they realized. It would be doing less while using even more resources...

[ensabrenoir]

....and start this changes everything about Bd/ true performance/once fixed rant..........NOW!



TOLD YOU SO SEE

[unsmart]

The PR guy,s where right it's the fab that screwed up and forgot the other 800million AMD's just saving face. Once they figure where they go BD[netburstII] will live up to the hype.

[Super XP]

Well I can see AMD turning this around for the betterment of the company and hopefully fixing then releasing Bulldozer II or should I say Piledriver.

[15th Warlock]

Originally Posted by Super XP

Well I can see AMD turning this around for the betterment of the company and hopefully fixing then releasing Bulldozer II or should I say Piledriver.

I wish I could share the "glass half full" attitude, but I just fail to see how this mistake could be used for any good and/or affect the performance of AMD future processor offerings...

[cadaveca]

And people are complaining about AMD firing people?

[qubit]

Originally Posted by 15th Warlock

I wish I could share the "glass half full" attitude, but I just fail to see how this mistake could be used for any good and/or affect the performance of AMD future processor offerings...

Indeed, it's an unmitigated embarrassment. I think they should have left the "error" in, rather than try to make their processor look better.

[Super XP]

Originally Posted by 15th Warlock

I wish I could share the "glass half full" attitude, but I just fail to see how this mistake could be used for any good and/or affect the performance of AMD future processor offerings...

You know why? Because AMD has no choice but to make many years of R&D work the way it was meant to.

PI was a bad chip, but PII turned out to be really good. Now they need to repeat this with Bulldozer but with more aggressive improvements. With Bulldozer this is achievable, but with PII it was not. This old K10 is just that, too old and needs to be put to rest.

[bencrutz]

Originally Posted by Benetanegia

My internal logic tells me that despite being different architectures, there's no way there's such a big difference between Llano transistor density and BD transistor density (66% difference!!)

Llano have a GPU in it while BD don't?

[15th Warlock]

Originally Posted by Super XP

You know why? Because AMD has no choice but to make many years of R&D work the way it was meant to.

PI was a bad chip, but PII turned out to be really good. Now they need to repeat this with Bulldozer but with more aggressive improvements. With Bulldozer this is achievable, but with PII it was not. This old K10 is just that, too old and needs to be put to rest.

Yes, AMD better step up to the plate and starts delivering, I don't want a monopoly that will do nothing but harm the PC industry, I agree with you on that; however, I still fail to see how the mistake reported in this thread will make AMD better or even affect the design of their future processors, these CPUs are designed by multi-national teams of some of the most brilliant engineers in the world, and quite frankly a stupid PR mistake like this has nothing to do with whether their processors deliver or not, unless so said PR mistake is so big as to affect the financials of AMD to the point where they can no longer invest as much on R&D.

[Benetanegia]

Originally Posted by bencrutz

Llano have a GPU in it while BD don't?

So? SB has a GPU in it while SB-E doesn't and that does not prevent the bigger chip from having a higher density. GPUs may have higher density, but they do not make up for a 66% difference period. And definitely not when the GPU only ocupies 1/3rd of the die.

A 25% difference, I would have bought that. 66% no way. Especially since 16 MB of cache, with 6T/bit SRAM would already take around 900 million transistors. Add IMC, add hypertransport and other I/O logic. Add the cores! 1200 NO WAY.

Plus the justification holds no water. Oh it's Interlagos that is 2.4 billion and Valencia/Zambezi is 1.2 billion. PR staff told the numbers for Interlagos and 2.4 suddenly became 2. Also 8C part having exactly half the transistor amount as the 16C part, that alone already stinks as a lie. As does the fact that they are giving rounded up numbers when in the past they never did. Thuban 904 million. Deneb 758 million.

They will never convince me that such a mistake could have been made in a PR world that is so carefully planned these days and they will never convince me of a transistor amount that is unrealistic.

[de.das.dude]

maybe this is why it has such crapy performance.

[Zen_]

Originally Posted by de.das.dude

maybe this is why it has such crapy performance.

No. The reason has been right in front of us the whole time from the Real World Tech article...inefficient branch prediction logic that has difficultly keeping the pipelines full, and hardware prefetchers that are next to useless.

[qubit]

THIS THREAD HAS NOW BEEN SUPERCEDED BY THIS NEWS ARTICLE. PLEASE COMMENT THERE.