AMD cuts to the core with 'Bulldozer' Opterons

[w0mbat]

Don't need to check your IP, but if I were you I would check your sources.

why? cause i already had a running system in front of me? the source is me as i am in india.

no offence JF, but u said there wont be any 6-core desktop cpus when we already where running these things

[JF-AMD]

why? cause i already had a running system in front of me? the source is me as i am in india.

no offence JF, but u said there wont be any 6-core desktop cpus when we already where running these things

Well, I work for AMD and i can tell you that you don't have a bulldozer system running in front of you.

And I don't recall saying that there will not be 6-core desktops. What I said was Istanbul is a server die and will not be a desktop die.

As you are probably aware I am a server guy so I don't comment on desktop plans. I just happed to know that they were not taking our die and that Istanbul was going to remain server only.

Thuban does not use server die. It has features that we don't have and we have features that they don't have.

[Nedjo]

I feel that w0mbat will get urge to post some scrshots
After all he has certain reputation when it comes to reliability of his posts!

Yeah he'll need to hide in Himalayas after that, but ho cares! Fun factor is what counts

[Lightman]

I feel that w0mbat will get urge to post some scrshots
After all he has certain reputation when it comes to reliability of his posts!

Yeah he'll need to hide in Himalayas after that, but ho cares! Fun factor is what counts

Seconded!

I have very high regard of both JF-AMD and w0mbat when it comes to AMD products.
Obviously J.F. you are AMD's server product face and you need to protect company secrets which is completely understandable!
To be hones I've learned few things how corporations operate from your posts, which is great!
On the other hand w0mbat is a cool forum fellow which I indirectly know for few years now. He always was quite credible in his predictions/leaks and thanks to him and persons alike these forums are what they are for me! Very interesting places to be and exchange informations!

[JF-AMD]

I feel that w0mbat will get urge to post some scrshots
After all he has certain reputation when it comes to reliability of his posts!

Yeah he'll need to hide in Himalayas after that, but ho cares! Fun factor is what counts

Maybe he is talking about Magny Cours, that is in production now and lots of people have them in their hands....


Screenshots can be faked. Instead, post a picture of the bulldozer processor itself. Every processor we have has the markings on the top.

Pre-production parts have a special model number that, unless you actually had one in your hands, you probably couldn't guess.

The old barcelona I have sitting on my desk at work isn't an "OS23xxxxxxx" model number, IIRC it is something like "ZVX140xxxxx"

So, pictures please. Or, if you are not comfortable posting pictures, give me the OPN on the lid.

I can guarantee, with very high accuracy, that you will not have that number.

[saaya]

Well, I work for AMD and i can tell you that you don't have a bulldozer system running in front of you.

And I don't recall saying that there will not be 6-core desktops. What I said was Istanbul is a server die and will not be a desktop die.

As you are probably aware I am a server guy so I don't comment on desktop plans. I just happed to know that they were not taking our die and that Istanbul was going to remain server only.

Thuban does not use server die. It has features that we don't have and we have features that they don't have.

excuse me, but if amd really created a seperate die for desktop 6core chips they are plain stupid... sorry, but seriously...
with the limited rnd money they have i HIGHLY doubt they use different silicon for server and desktop 6core chips... it makes no sense at all as the differences in silicon are tiny :P

i know that both amd and intel dont want customers to think desktop and server parts are actually the same with a different sticker on the box for obvious reasons... but come on...

Maybe he is talking about Magny Cours, that is in production now and lots of people have them in their hands....


Screenshots can be faked. Instead, post a picture of the bulldozer processor itself. Every processor we have has the markings on the top.

Pre-production parts have a special model number that, unless you actually had one in your hands, you probably couldn't guess.

The old barcelona I have sitting on my desk at work isn't an "OS23xxxxxxx" model number, IIRC it is something like "ZVX140xxxxx"

So, pictures please. Or, if you are not comfortable posting pictures, give me the OPN on the lid.

I can guarantee, with very high accuracy, that you will not have that number.

why would he do that? so he exposes himself or his source and can be legally prosecuted? who cares about model and serial numbers?

and it wouldnt make sense if amd already had bulldozer silicon but wouldnt share it with their close partners/customers... and amd said they already had working samples in their labs and didnt they even mention they were shipping samples to customers?

[JF-AMD]

The dies are based on the same architecture, but each has different features added. It is not a "grounds up" design, they are based on the same thing. But you can't put an istanbul in an AM3 and call it thuban because there are different features.

As to whether someone has an actual bulldozer running, let me just propose this thought: The guy who would know a.) there were silicon available and b.) if any of that silicon was outside of our walls, is saying that someone can't have silicon in their hands.

I obviously can't comment on specifics, but, if someone was wondering whether there was actually silicon outside of AMD's walls, they might want to put slightly more faith on the person who has access to the master schedules, no?

Everyone can believe whatever they want. It's not the laws of intellectual property that are at debate here, it's the laws of time and physics.

[saaya]

The dies are based on the same architecture, but each has different features added. It is not a "grounds up" design, they are based on the same thing. But you can't put an istanbul in an AM3 and call it thuban because there are different features.

if you have the tools and access codes to re-fuse the die, i bet you could

all departments that are going to sell the part under one name or another will throw in what features they want/need and the final design will be a compromise between all of that, and then each team will fuse things on or off and bin according to what they need want to create a product out of that molten chunk of sand somebody sneezed hazardous metals on

different features, yes, different silicon, no...
i refuse to believe amd is wasting rnd resources on such a useless effort... you dont create a new set of multi million us$ lito masks because less than 1% of the silicon is slightly different... and if you do, then your either making lots of money with it (not the case last time i checked, no offense!) or your really REALLY bad at managing your rnd and mfc budget :P

As to whether someone has an actual bulldozer running, let me just propose this thought: The guy who would know a.) there were silicon available and b.) if any of that silicon was outside of our walls, is saying that someone can't have silicon in their hands.

I obviously can't comment on specifics, but, if someone was wondering whether there was actually silicon outside of AMD's walls, they might want to put slightly more faith on the person who has access to the master schedules, no?

Everyone can believe whatever they want. It's not the laws of intellectual property that are at debate here, it's the laws of time and physics.

hey, im not saying your lying to us! ive read quite some of your posts and seen some interviews and you seem like a very honest and nice guy

but i dont trust that you were informed about everything that goes on... again, no offense... but telling you about something you dont need to know, especially SINCE you are a honest and straightforward guy, would be risky, or at least according to some top guys at amd i guess it could be considered risky.

and correct me if im wrong, but didnt you say something along the lines of "i cant comment on anything past the next xxx months cause thats all i need to know and thus thats all my superiors let me know" during an interview some time ago?

please dont take any offense by what i wrote, i hope i dont sound rude or harsh... and thanks for being so straight forward and honest and sticking around here... it helps to improve amds image as a company that is aware of what is going on around it and listens to the market and customers... something that was lost during the reign of senor ruin errrrr i mean ruiz

[JF-AMD]

No offense taken. It's just that I know silicon schedules. I know the milestones to ~ the end of 2011 at this point. I know Bulldozer schedules well and don't really care to know the follow on product's schedules because that is outside of my horizon.

The earliest silicon comes in with ~1-2 wafers worth of product. These are extremely valuable and have the highest security. A company is going to know exactly where every one of them is (early silicon is always locked away in a lab). They are all internal, they don't go to partners. Highest scrutiny.

When you get to EVT silicon, now, you are building ~1000 or so units. Could one or two get into the wrong hands? Sure. Do these go outside of the company? Yes.

EVT is generally ~6 months prior to launch. If I am launching in 2011, there is pretty good odds that I am not at EVT level silicon, right? So, IF I had silicon (not confirming or denying), it would be locked away in the highest security internal lab, only a few folks would have access to it, and it would all be accounted for. At that point it is critical that every single piece is either in a system, or a tester, not just laying around. People would notice one missing and it would be a big deal.

Yesterday I was in the power lab (watch for a video in the future) and there were plenty of Magny Cours laying around. I picked one up out of the tray for the video. That would never happen with Bulldozer. I doubt they would even let me in the lab, I have no business in there; pretty sure my badge wouldn't work

[Nedjo]

Maybe he is talking about Magny Cours, that is in production now and lots of people have them in their hands....

well the thing about someones reputation is based on the facts that he knows exactly what does he talking about, and w0mbat certainly doesn't have reputation of being n00b

Screenshots can be faked. Instead, post a picture of the bulldozer processor itself. Every processor we have has the markings on the top.

Again there's certain trust that someone is posting true pictures. eg. people know that say... chew* or Sampsa, or saaya... and bunch of other people would never post faked pictures... and w0mbat is in that trustworthy group...

Pre-production parts have a special model number that, unless you actually had one in your hands, you probably couldn't guess.

I know I had some Agenas - they've end-up being bellowed gifts, after polishing and photo engraving

The old barcelona I have sitting on my desk at work isn't an "OS23xxxxxxx" model number, IIRC it is something like "ZVX140xxxxx"
So, pictures please. Or, if you are not comfortable posting pictures, give me the OPN on the lid.
I can guarantee, with very high accuracy, that you will not have that number.

No one who has access to the any type of NDA samples will not post picture with OPN

So, point of my quoting is that we have brilliant situation where trustworthy member is in the position of losing hard earned credibility 'cos other member with equal level of reliability is daring him to prove its trustworthiness... this has huge potential in terms of forum communities importance in contemporary marketing and PR strategies.

excuse me, but if amd really created a seperate die for desktop 6core chips they are plain stupid... sorry, but seriously...
with the limited rnd money they have i HIGHLY doubt they use different silicon for server and desktop 6core chips... it makes no sense at all as the differences in silicon are tiny :P

don't think you've understud JF correctly! He was speaking about silicon features, not the separate MASK that is used to create Server/Desktop part!

Mask is the same, but production recipe and Enabled/Disabled features ARE different (eg. desktop part doesn't need snoop filter part of the L3)!

As to whether someone has an actual bulldozer running, let me just propose this thought: The guy who would know a.) there were silicon available and b.) if any of that silicon was outside of our walls, is saying that someone can't have silicon in their hands.
I obviously can't comment on specifics, but, if someone was wondering whether there was actually silicon outside of AMD's walls, they might want to put slightly more faith on the person who has access to the master schedules, no?

So there IS silicon INSIDE walls? j/k

Everyone can believe whatever they want. It's not the laws of intellectual property that are at debate here, it's the laws of time and physics.

I'll personally believe ONLY when Dirk say what there is and what there's not!

different features, yes, different silicon, no...
i refuse to believe amd is wasting rnd resources on such a useless effort... you dont create a new set of multi million us$ lito masks because less than 1% of the silicon is slightly different... and if you do, then your either making lots of money with it (not the case last time i checked, no offense!) or your really REALLY bad at managing your rnd and mfc budget :P

Behind this statement I expect knowledge of:
a) leading the manufacturing process of highly integratet x86 processing units
b) economics management of the "a"

So saaya any backgrounds on these departments in your CV?

but i dont trust that you were informed about everything that goes on... again, no offense... but telling you about something you don't need to know, especially SINCE you are a honest and straightforward guy, would be risky, or at least according to some top guys at amd i guess it could be considered risky.

I'd say that JF has extensife knowledge of what he knows, can knows, does say and can say... it's not like he's with in the company for the same time as being member of this forum

and correct me if im wrong, but didnt you say something along the lines of "i cant comment on anything past the next xxx months cause thats all i need to know and thus thats all my superiors let me know" during an interview some time ago?

so now you've busted his neat excuse not to use boring "no comment"

please dont take any offense by what i wrote, i hope i dont sound rude or harsh... and thanks for being so straight forward and honest and sticking around here... it helps to improve amds image as a company that is aware of what is going on around it and listens to the market and customers... something that was lost during the reign of senor ruin errrrr i mean ruiz

I think that JF has figured out that this is the forum with the best crew that really appreciate his posts and the fact that he is so proactive, although we don't hassle with server parts here


Oh and I just recollected one person name who was proactive in spreading Iranian freedom movement, on some other social community service... but the problem is he's not in India

BTW
I don't believe that w0mbat will post any pictures

[Nedjo]

No offense taken. It's just that I know silicon schedules. I know the milestones to ~ the end of 2011 at this point. I know Bulldozer schedules well and don't really care to know the follow on product's schedules because that is outside of my horizon.
fixed...

The earliest silicon comes in with ~1-2 wafers worth of product. These are extremely valuable and have the highest security. A company is going to know exactly where every one of them is (early silicon is always locked away in a lab). They are all internal, they don't go to partners. Highest scrutiny.

When you get to EVT silicon, now, you are building ~1000 or so units. Could one or two get into the wrong hands? Sure. Do these go outside of the company? Yes.

EVT is generally ~6 months prior to launch. If I am launching in 2011, there is pretty good odds that I am not at EVT level silicon, right? So, IF I had silicon (not confirming or denying), it would be locked away in the highest security internal lab, only a few folks would have access to it, and it would all be accounted for. At that point it is critical that every single piece is either in a system, or a tester, not just laying around. People would notice one missing and it would be a big deal.

Yesterday I was in the power lab (watch for a video in the future) and there were plenty of Magny Cours laying around. I picked one up out of the tray for the video. That would never happen with Bulldozer. I doubt they would even let me in the lab, I have no business in there; pretty sure my badge wouldn't work

developing this post in full blown blog would be AWESOME!

[Dimitriman]

Are there any AMD labs in India? :P

But seriously I understand what JF-AMD means, working as an accountant there are some levels of information you simply must track from beginning to end and I can't imagine how a few working samples of BD would go unaccounted or distributed without extensive knowledge about it. Now the question is whether the decision to distribute/test these samples elsewhere of AMD labs is always known to JF.

As he claimed, the AMD labs where BD are kept so far stand out of his jurisdiction so perhaps there has been internal decisions to test these samples outside? Therefore putting an actual explanation to the question: do you work with engineers/labs in India in any form?

[LightSpeed]

No offense taken. It's just that I know silicon schedules. I know the milestones to ~ the end of 2011 at this point. I know Bulldozer schedules well and don't really care to know the follow on product's schedules because that is outside of my horizon.

The earliest silicon comes in with ~1-2 wafers worth of product. These are extremely valuable and have the highest security. A company is going to know exactly where every one of them is (early silicon is always locked away in a lab). They are all internal, they don't go to partners. Highest scrutiny.

When you get to EVT silicon, now, you are building ~1000 or so units. Could one or two get into the wrong hands? Sure. Do these go outside of the company? Yes.

EVT is generally ~6 months prior to launch. If I am launching in 2011, there is pretty good odds that I am not at EVT level silicon, right? So, IF I had silicon (not confirming or denying), it would be locked away in the highest security internal lab, only a few folks would have access to it, and it would all be accounted for. At that point it is critical that every single piece is either in a system, or a tester, not just laying around. People would notice one missing and it would be a big deal.

Yesterday I was in the power lab (watch for a video in the future) and there were plenty of Magny Cours laying around. I picked one up out of the tray for the video. That would never happen with Bulldozer. I doubt they would even let me in the lab, I have no business in there; pretty sure my badge wouldn't work

Nicely said

However, since you are saying its possible to be at the EVT level silicon (not confirming or denying) then those ~1000 units may be spread out in different places of the world, yet locked up under tight security with only a very few number of people with access to it regardless of where it is. So is the possibility zero that in some part of the world, some engineer showed it to someone?

Just wondering. More so since his rep is right up there IMO

[FlanK3r]

whau, i was 5 hous off and now is here very interresting discussion :-)

[saaya]

don't think you've understud JF correctly! He was speaking about silicon features, not the separate MASK that is used to create Server/Desktop part!

Mask is the same, but production recipe and Enabled/Disabled features ARE different (eg. desktop part doesn't need snoop filter part of the L3)!

well he didnt clearly say it
and fusing doesnt make it a different chip, it makes it a different SKU or part or bin... but a lot of articles and pr talk mentioned the two 6 cores as different chips...

[saaya]

When you get to EVT silicon, now, you are building ~1000 or so units. Could one or two get into the wrong hands? Sure. Do these go outside of the company? Yes.

EVT is generally ~6 months prior to launch. If I am launching in 2011, there is pretty good odds that I am not at EVT level silicon, right? So, IF I had silicon (not confirming or denying), it would be locked away in the highest security internal lab, only a few folks would have access to it, and it would all be accounted for. At that point it is critical that every single piece is either in a system, or a tester, not just laying around. People would notice one missing and it would be a big deal.

only 6 months? there have been leaked samples more than 6 months before the launches in the past... and isnt bulldozer supposed to launch in early 2011 if not end of this year? in that case EVT would have started recently or would be well under way...

[Shadov]

well he didnt clearly say it
and fusing doesnt make it a different chip, it makes it a different SKU or part or bin... but a lot of articles and pr talk mentioned the two 6 cores as different chips...

JF was talking about different functions from the start, additonally Lisbon (-> Thuban) is supposed to be different from Istanbul.

Though you always see what you want in order to support your interpretation.

[Opteron146]

Yesterday I was in the power lab (watch for a video in the future) and there were plenty of Magny Cours laying around. I picked one up out of the tray for the video. That would never happen with Bulldozer. I doubt they would even let me in the lab, I have no business in there; pretty sure my badge wouldn't work

Sounds to me like w0mbat's badge is working for doors that are closed to you

[JF-AMD]

developing this post in full blown blog would be AWESOME!

I could do that, it's pretty simple since I have to run through the silicon with each launch. Let me make that a "post MC launch" blog, I have a bunch of ones in process including one that went live this past week.

[JF-AMD]

Are there any AMD labs in India? :P

But seriously I understand what JF-AMD means, working as an accountant there are some levels of information you simply must track from beginning to end and I can't imagine how a few working samples of BD would go unaccounted or distributed without extensive knowledge about it. Now the question is whether the decision to distribute/test these samples elsewhere of AMD labs is always known to JF.

As he claimed, the AMD labs where BD are kept so far stand out of his jurisdiction so perhaps there has been internal decisions to test these samples outside? Therefore putting an actual explanation to the question: do you work with engineers/labs in India in any form?

Yes, we have development in India, they're awesome folks, I think they did a lot of the Istanbul work, very solid team. But, I am pretty sure that first silicon does not land there, as they are a design organization as I understand it. First silicon goes to test organizations.

The issue is that if we were going to "leak" parts out, it would actually be my call as the director of marketing. And I can tell you that I wouldn't approve that. Especially early Bulldozer silicon. Trust me, the first rev is always pretty rough relative to the final (except Istanbul where first silicon was final silicon.)

[JF-AMD]

Sounds to me like w0mbat's badge is working for doors that are closed to you

Well, maybe he could email me on the company email system then with the OPN, right?

[JF-AMD]

JF was talking about different functions from the start, additonally Lisbon (-> Thuban) is supposed to be different from Istanbul.

Though you always see what you want in order to support your interpretation.

So Lisbon and Thuban are different. I am not aware if this is a different mask or a fusing option. My gut would tell me a mask difference because they don't want HT Assist (it's really not of value for single processors).

We will all use the same cores and memory controllers, but there are distinct difference. Based on the feature sets, that would lead me to guess that they are different masks. But I don't work in development so I can't say for sure.

[Nedjo]

Well, maybe he could email me on the company email system then with the OPN, right?

only if there's one engraved on IHS... if there's IHS at all

[Opteron146]

Well, maybe he could email me on the company email system then with the OPN, right?

I think you underestimate his mental capabilities by a quite high margin

[qcmadness]

So Lisbon and Thuban are different. I am not aware if this is a different mask or a fusing option. My gut would tell me a mask difference because they don't want HT Assist (it's really not of value for single processors).

We will all use the same cores and memory controllers, but there are distinct difference. Based on the feature sets, that would lead me to guess that they are different masks. But I don't work in development so I can't say for sure.

HT Assist, as I recalled somewhere, can be turned on or off by BIOS.
So no big deal.

Using different mask will be a huge waste.
Even Intel does the same thing (server / mobile / desktop - 1 mask for all).