AMD: 32nm yields ramping worse than expected, Llano delayed, BD didn't sample in H1

Quote:

---

Originally Posted by MrMojoZ

That info hasn't been current for awhile, sorry. This isn't 2007.

---

Quote:

---

Originally Posted by informal

You are late to the news party,3 years late to be precise.

---

Nice logic there. A project is on track compared to what ? What is the reference date? Because if that's constantly changed with every news release from the company, then a project can never be late.

Any sane person would see that BD is already 1 year late and counting compared to the original plan.

Quote:

---

Originally Posted by Solus Corvus

I love how certain you are about BD being late and crap. You have been predicting BD being Q3/Q4'11 or even 2012 for years now. It looks like you are taking every new negative rumor and bit of bad news and trying to fit them into the late/crap narrative. If that's the narrative you want to build then fine, but don't try to come back later and claim you were just being objective.

---

Regardless of what's being discussed, from a simple philosophical and logical point of view, if it ends as he originally predicted "for years" and that is not objective, than what the hell is objective?
It means the guy is either the new Nostradamus or he used his critical thinking and not relied on faith.

Quote:

---

I freely admit the possibility of BD being extremely late. I'll also admit that it could be a dud architecture. But I don't see enough hard evidence to suggest that they are the only possibilities. You seem to have totally rejected the possibility that BD will deliver competitive performance and/or in a reasonable timeframe. If BD taped out in Q2 and things go relatively smoothly then 12 months isn't unrealistic.

---

I don't think he rejected anything. Just that the uarch rework done in 2007-2009 means they originally aimed too low and now the process blips introduce a feeling of uncertainty whether the original plans for a mid 2011 launch are still valid.

Quote:

---

Originally Posted by savantu

Nice logic there. A project is on track compared to what ? What is the reference date? Because if that's constantly changed with every news release from the company, then a project can never be late.

Any sane person would see that BD is already 1 year late and counting compared to the original plan.

---

The reference point is against the most recent roadmap target of "2011". Of course it could change, obviously it has quite a bit already.

Quote:

---

Originally Posted by Solus Corvus

The reference point is against the most recent roadmap target of "2011". Of course it could change, obviously it has quite a bit already.

---

Huh ? If the latest roadmap is always the reference point, please explain how a project can be late ?

What kind of cognitive reasoning is this ?

if we imagine that amd right now said BD is put on pause for 5 years, would intel continue to build to 22nm and lower? or just pause and eat some nice profits? and if they did pause, would you say they are late, or just realigned their schedule because outside factors make this the best choice?

didnt amd get slowed down when intel changed the sse5 specs? (or avx or whatever it was)

Quote:

---

Originally Posted by savantu

?? This is absurd.

Nehalem has more states than atoms in the universe ( that's a quote from the chief designer ); I doubt Bulldozer is any less complex ( ok, maybe it is since it lacks full SMT ), either way, your claims come against anything the validation teams are saying.

---

i'm not an architect so i cant tell you how many states a modern cpu has. AMD has taken Occam's razor into consideration, something intel hasn't done very well. for example look at how they handle OoO/register renaming in current cpu's.

that doesnt really go against what i am saying. if you catch a bug after the cpu is in the field then you are screwed. if it is at wafer or package level it's not nearly as bad. barton would get fired if there was a recall on nehalem.

Quote:

---

http://www.nytimes.com/2008/11/17/te...5547.html?_r=1

Validation is harder and harder with each new generation, the complexity increase is astounding.

---

but they get better and better at it. productivity has been increasing despite the size of design teams staying about the same. increases in complexity has leveled off over the years mainly because they wont build a cpu that they cant verify in time.

an example would be removing all domino logic from the datapaths in nehalem and switching to pure static cmos logic. this is the exact opposite of P4 which had absurdly complex circuits, to the point where they had to add more pipeline stages to keep the FO4 delay at around 10, an amazing feat but useless in the end product.

Quote:

---

Originally Posted by Manicdan

my rule of thumb:
a debate is between 2 winners
an argument is between 2 losers

---

i think im going to save that quote for a future argument

Quote:

---

Originally Posted by Manicdan

my rule of thumb:
a debate is between 2 winners
an argument is between 2 losers

---

...if you would know from where the word argumenet orgines from you would make such nonsense statements... also you can't have a discussion without arguments..

You should better exchange argument with opinion, cause an argument you can prove or disprove by hard facts and or logical reasoning... and opinion on the other had is always an opinion regardless of facts and logic. :p:

Quote:

---

Originally Posted by savantu

Nice logic there. A project is on track compared to what ? What is the reference date? Because if that's constantly changed with every news release from the company, then a project can never be late.

Any sane person would see that BD is already 1 year late and counting compared to the original plan.



Regardless of what's being discussed, from a simple philosophical and logical point of view, if it ends as he originally predicted "for years" and that is not objective, than what the hell is objective?
It means the guy is either the new Nostradamus or he used his critical thinking and not relied on faith.



I don't think he rejected anything. Just that the uarch rework done in 2007-2009 means they originally aimed too low and now the process blips introduce a feeling of uncertainty whether the original plans for a mid 2011 launch are still valid.

---

Well one thing is for certain, opinion raises a lot of interesting question about the failure that is larabee. I assume you realise that?

Quote:

---

Originally Posted by Manicdan

if we imagine that amd right now said BD is put on pause for 5 years, would intel continue to build to 22nm and lower? or just pause and eat some nice profits? and if they did pause, would you say they are late, or just realigned their schedule because outside factors make this the best choice?

---

So you build an argument from an absurd premises and then use it infer on Intel's action ? Allow me to refuse entering in this logical fallacy.

I work with projects which span over 5-7 years, from the rfq to end of production. Every project has an internal time lime established from the start by the PM and they typically use "gates" ( important ones called milestones ) to measure progress. Everything is referenced to original gate or milestone.

We do not know the original time schedule for BD. We do know however when the company offered the first external schedules ( roadmaps ) for BD. The original roadmap is our reference point.That the project couldn't keep to it from various reasons, is another issues altogether.

Nehalem was late by 2 years ( planned for 2006 ) - it was redone in 2003, Sandy bridge is also late by 2 years - redone in 2007 IIRC, nothing new.

It simply means they aimed too low and had to readjust their goals.

Quote:

---

didnt amd get slowed down when intel changed the sse5 specs? (or avx or whatever it was)

---

They use the original format which is more complex and includes the one Intel adopted.

Watches thread..:D

So how many of the Pro-Intel guys in this thread are like savantu; guys who are linked to Intel projects? Seems like way too many on here.

Quote:

---

Originally Posted by savantu

Nehalem was late by 2 years ( planned for 2006 ) - it was redone in 2003, Sandy bridge is also late by 2 years - redone in 2007 IIRC, nothing new..

---

exactly ... iirc phenom I had some problem that they pushed back phenom II so in that same way bulldozer could be late 2 years .... 2011 instead of 2009 because of problems along the way with another cpu in the past ..... they respect life cycle of each generation so they maximise the proffit on each generation to recoup the R&D money right?

Quote:

---

Originally Posted by Hornet331

...if you would know from where the word argumenet orgines from you would make such nonsense statements... also you can't have a discussion without arguments..

You should better exchange argument with opinion, cause an argument you can prove or disprove by hard facts and or logical reasoning... and opinion on the other had is always an opinion regardless of facts and logic. :p:

---

the idea has very little to do with facts, and much more about the emotion of people involved (and no i have not a clue for its definition, english is not my strong point, just my first and only language)

a debate is 2 people with facts and for helping direct a decision, very restrictive, and very formal (excluding politics, lol)

an argument is 2 people who think talking louder makes them more right, i say its between 2 losers because the moment a debate turns into an argument, cops get involved and your both on tv looking like idiots.

Quote:

---

Originally Posted by informal

You are late to the news party,3 years late to be precise.

---

Quote:

---

Originally Posted by informal

.AMD always stated 2011 as a target for BD launch,looking at how they executed starting with Deneb(pull-ins all the way with the roadmaps) you can bet they will execute equally if not better. Original post.

---

It seems you were late to bulldozer related news considering you wrote bulldozer was always scheduled for 2011. ;)

This thing is a year late according to AMD, but it's employee says it's still on track. :shocked:

Debate or debating is a formal method of interactive and representational argument


http://dictionary.reference.com/browse/argument

Quote:

---

Originally Posted by Kuroimaho

It seems you were late to bulldozer related news considering you wrote bulldozer was always scheduled for 2011. ;)

This thing is a year late according to AMD, but it's employee says it's still on track. :shocked:

---

Reference point is the latest roadmap and latest server schedule update.
Bulldozer @ 45nm was a scraped project and it is known for 3 years.The new BD core is improved version,you can call it Version2 if you like and although it's related to the one you linked,is another project done on a smaller node.Changes are done on the various core parts(fpu being the most obvious one) and it truly warrants a new name.
You need to inform yourself better before getting inside these kind of arguments.Remember,this is not AT forumz or THG forumz.

Quote:

---

Originally Posted by savantu

Huh ? If the latest roadmap is always the reference point, please explain how a project can be late ?

What kind of cognitive reasoning is this ?

---

The current target is 2011. So when we are talking right now about the product being on track we are talking about in reference to the current target.

Your "but what if it changes" objection is irrelevant to the CURRENT discussion. Of course it could change in the future, and if it does then BD would be late in reference to the date in the current discussion and on track in reference to the future target. But since we don't know if it will be rescheduled in the future the most logical use of the term on target is in relation to the most current estimates and not on old estimates or future hypotheticals.

By your logic any company that has to reschedule a project can never be on target again. As if companies can't decide on more realistic targets when the real world doesn't cooperate.

Quote:

---

Originally Posted by savantu

Regardless of what's being discussed, from a simple philosophical and logical point of view, if it ends as he originally predicted "for years" and that is not objective, than what the hell is objective?
It means the guy is either the new Nostradamus or he used his critical thinking and not relied on faith.

---

This is a false dichotomy. A person could easily could easily take something on faith and end up being right without having come to that conclusion through logic or magical powers. My team is so great I'm sure they will win. When they do it's proof that I was right all along regardless of how it happened and when they don't it's an exception. Sound like any sports fans you know?

Quote:

---

I don't think he rejected anything. Just that the uarch rework done in 2007-2009 means they originally aimed too low and now the process blips introduce a feeling of uncertainty whether the original plans for a mid 2011 launch are still valid.

---

What "mid" 2011 plans? Last I heard it was all of 2011 so it could be Q1 through Q4 and not be late in reference to the current timetable. That encompasses both my estimate and his (except the 2012 part), but he has yet to acknowledge that. Neither have you.

As for what he has rejected, he can speak for himself. All we have so far seems like a rejection of the possibility, though he's free to say otherwise.

Quote:

---

If I had to guess, I'd say BD will be lucky to make late Q3 on the desktop, and will be Q4, (if not H1 2012) for server products, due to the extra validation required.

---

Quote:

---

You guys should just come out and move BD to "H2 2011" at this point, at least, obviously "H1 2011" ain't happening.

---

Quote:

---

BD isn't going to ship in Q2 2011.

---

Quote:

---

Originally Posted by informal

Reference point is the latest roadmap and latest server schedule update.

---

You know, there is huge pink elephant on the sofa and nobody seems to notice it.

You apply a logical fallacy called "circular reasoning" : since the reference point is the latest roadmap, a project can never miss the target because the target is linked to the latest roadmap.

Roadmap -> target -> new roadmap -> new target. You're always on target.

Is it that hard to grasp ?

Quote:

---

Bulldozer @ 45nm was a scraped project and it is known for 3 years.The new BD core is improved version,you can call it Version2 if you like and although it's related to the one you linked,is another project done on a smaller node.Changes are done on the various core parts(fpu being the most obvious one) and it truly warrants a new name.
You need to inform yourself better before getting inside these kind of arguments.Remember,this is not AT forumz or THG forumz.

---

I don't frequent any of those, but you seem much more familiar to them. Guess it explains the reasoning then.

Quote:

---

Originally Posted by richierich

So how many of the Pro-Intel guys in this thread are like savantu; guys who are linked to Intel projects? Seems like way too many on here.

---

Who said I'm linked to Intel projects ? I'm not. I work in a different industry.There are other Intel employees around here, some even posted in this thread.

Quote:

---

Originally Posted by Solus Corvus

The current target is 2011. So when we are talking right now about the product being on track we are talking about in reference to the current target.

Your "but what if it changes" objection is irrelevant to the CURRENT discussion. Of course it could change in the future, and if it does then BD would be late in reference to the date in the current discussion and on track in reference to the future target. But since we don't know if it will be rescheduled in the future the most logical use of the term on target is in relation to the most current estimates and not on old estimates or future hypotheticals.

By your logic any company that has to reschedule a project can never be on target again. As if companies can't decide on more realistic targets when the real world doesn't cooperate.

---

Precisely, that's how things work in real life. It's late compared to the original plan. Having to reschedule means missing the target.
It is true for Boeing 787 Dreamliner, Airbus 380, Tesla Volt, Intel cpus, AMD cous, tin cans, etc, etc.

There was a game I saw being discussed here which was scheduled to arrive something like 2004 and they constantly changed it and adapted it to new features and it dragged on for so long that they had to cancel it this year and the developer closed. According to your reasoning, they were always on target up to the last moment. Very reassuring....

Quote:

---

This is a false dichotomy. A person could easily could easily take something on faith and end up being right without having come to that conclusion through logic or magical powers. My team is so great I'm sure they will win. When they do it's proof that I was right all along regardless of how it happened and when they don't it's an exception. Sound like any sports fans you know?

---

We agree here. That's why I said " he is either a new Nostradamus ( works on blind faith that AMD will screw up ) or he has excellent critical thinking and feels the way things are going.

Quote:

---

What "mid" 2011 plans? Last I heard it was all of 2011 so it could be Q1 through Q4 and not be late in reference to the current timetable. That encompasses both my estimate and his (except the 2012 part), but he has yet to acknowledge that. Neither have you.

---

True, I would say the consensus went for Q2/Q3 which is mid 2011.

Quote:

---

Originally Posted by savantu

Precisely, that's how things work in real life. It's late compared to the original plan. Having to reschedule means missing the target.

---

Yes, you missed the target. So you make new targets. You don't just sit around and say "I have no clue when we will get it out now". We are discussing on target in relation to the current targets not missed targets or hypothetical new targets because that would be illogical.

Quote:

---

There was a game I saw being discussed here which was scheduled to arrive something like 2004 and they constantly changed it and adapted it to new features and it dragged on for so long that they had to cancel it this year and the developer closed. According to your reasoning, they were always on target up to the last moment. Very reassuring....

---

No, I'd suggest that they missed a huge string of targets and were either behind target or on target for their last projected target before they closed depending on how realistic it was.

Or maybe they had sloppy management that didn't set new targets and push to meet them once the first was missed.

Quote:

---

We agree here. That's why I said " he is either a new Nostradamus ( works on blind faith that AMD will screw up ) or he has excellent critical thinking and feels the way things are going.

---

I see, then we agree. But how was I supposed to know how you felt about Nostradamus? Some people believe he had actual predictive ability (I'm not one of them). I'd also suggest that Nostradamus wasn't operating on typical blind faith and probably more on delusions, hallucinations, mental disorders, etc. Not (usually) a great analogy for the typical fan based thought process.

Quote:

---

True, I would say the consensus went for Q2/Q3 which is mid 2011.

---

Ok, but I'm not really talking about the consensus but what individuals are espousing.

Quote:

---

Originally Posted by savantu

I don't frequent any of those, but you seem much more familiar to them. Guess it explains the reasoning then.

---

No wonder why next to nobody takes you seriously.With that attitude it's a no-brainer.
BTW,I never posted at those places but I know the reputation and poor level of validity and fanboysm that rules there.

Quote:

---

Originally Posted by Solus Corvus

Yes, you missed the target. So you make new targets. You don't just sit around and say "I have no clue when we will get it out now". We are discussing on target in relation to the current targets not missed targets or hypothetical new targets because that would be illogical.

---

Exactly.

First, I think that AMD has executed exceptionally well lately. They have had periods of time in the past where they have performed exceptional, and times when they have fallen flat on their face.

As far as process goes, AMD has a mixed bag. Their move to copper interconnects was arguably the first time in the history of semiconductors when ANY company has ever held a process advantage over Intel. As a result the not only captured the frequency crown (and were the first commercial processor to exceed 1Ghz), they held the performance crown as well.

Since then, Intel has held the undisputed lead in process technology and shrinks. Flat out impressive.

As far as architecture goes, IMHO, AMD has simply owned Intel. Intel has been following AMD for about a decade now. I am looking to see a SB successor that imitates BD in the future.

My conclusion is this. AMD usually makes very good design decisions. They usually execute the design well (with the notable exception of the TLB bug); however, AMD spends its life in the shadow of Intel's process lead ..... as they have done with 32nm. It is remarkably difficult to compete from a full process node behind ..... and yet AMD is still in the running.

It is not difficult for me to believe that AMD may flounder on 32nm HKMG. Many thought that K10 was a failure; however, IMHO it was simply a process failure. At 2.2Ghz, K10 was no match for Core 2. Today 45nm to 45nm, PhenomII is more than a match for Core 2. Imagine if AMD had released Phenom at 3.0Ghz and greater.

The BD design details that I have been able to see or speculate show a very competent and potent design. If it can be released in 2011 (H1 would be better than H2), I suspect it will be a strong competitor to SB.

If AMD again falters on their process transition, things won't look pretty for them with respect to BD.

For those that believe that Intel is not fallible, remember i740? How about Larrabee? Itanic anyone? P4?

I am hoping that AMD executes well. The last system I purchased was a Core 2 which I thought I could update inexpensively to a quad in the future. Time has passed and passed. Had I went with AMD, I could have a quad core on the cheep that would kick butt in my video encoding tasks. (AMD 3.4Ghz ~$170. Intel equivalent at this price = Core 2 quad @2.66Ghz).

For those that prefer Intel, you should be rooting for BD too. Wouldn't you like to have a Core i7 at 3.2Ghz for under $150.00?

well said OneEng

Quote:

---

Originally Posted by OneEng

First, I think that AMD has executed exceptionally well lately. They have had periods of time in the past where they have performed exceptional, and times when they have fallen flat on their face.

As far as process goes, AMD has a mixed bag. Their move to copper interconnects was arguably the first time in the history of semiconductors when ANY company has ever held a process advantage over Intel. As a result the not only captured the frequency crown (and were the first commercial processor to exceed 1Ghz), they held the performance crown as well.

Since then, Intel has held the undisputed lead in process technology and shrinks. Flat out impressive.

As far as architecture goes, IMHO, AMD has simply owned Intel. Intel has been following AMD for about a decade now. I am looking to see a SB successor that imitates BD in the future.

My conclusion is this. AMD usually makes very good design decisions. They usually execute the design well (with the notable exception of the TLB bug); however, AMD spends its life in the shadow of Intel's process lead ..... as they have done with 32nm. It is remarkably difficult to compete from a full process node behind ..... and yet AMD is still in the running.

It is not difficult for me to believe that AMD may flounder on 32nm HKMG. Many thought that K10 was a failure; however, IMHO it was simply a process failure. At 2.2Ghz, K10 was no match for Core 2. Today 45nm to 45nm, PhenomII is more than a match for Core 2. Imagine if AMD had released Phenom at 3.0Ghz and greater.

The BD design details that I have been able to see or speculate show a very competent and potent design. If it can be released in 2011 (H1 would be better than H2), I suspect it will be a strong competitor to SB.

If AMD again falters on their process transition, things won't look pretty for them with respect to BD.

For those that believe that Intel is not fallible, remember i740? How about Larrabee? Itanic anyone? P4?

I am hoping that AMD executes well. The last system I purchased was a Core 2 which I thought I could update inexpensively to a quad in the future. Time has passed and passed. Had I went with AMD, I could have a quad core on the cheep that would kick butt in my video encoding tasks. (AMD 3.4Ghz ~$170. Intel equivalent at this price = Core 2 quad @2.66Ghz).

For those that prefer Intel, you should be rooting for BD too. Wouldn't you like to have a Core i7 at 3.2Ghz for under $150.00?

---

...this is one of the best posts i have read on here! i would sig you......but its too long.

Quote:

---

Originally Posted by OneEng

As far as architecture goes, IMHO, AMD has simply owned Intel. Intel has been following AMD for about a decade now. I am looking to see a SB successor that imitates BD in the future.

---

Did we already forget about Conroe aka C2D?

Quote:

---

Originally Posted by OneEng

As far as architecture goes, IMHO, AMD has simply owned Intel. Intel has been following AMD for about a decade now. I am looking to see a SB successor that imitates BD in the future.

---

K10 a better arch then Nehalem? Definitely not, IMO of course.