K8 launch delay caused by SOI verification and design change.

[NFS4]

http://digitimes.com/NewsShow/Article.asp?datePublish=2002/09/19&pages=PR&seq=203

 

[KF]

I didn't know the Barton Athlon was going to use SOI too. No wonder it's taking a while.

 

[Rand]

Forget that--- there is another MUCH more interesting factoid in that article.

The Barton-core processors, which will also adopt the SOI manufacturing technology, will hit the market in the first quarter of 2003 as well, Heye said.

Look's like SOI is again planned to hit in Barton.

 

[microAmp]

From article:

Heye added that AMD is also planning to introduce the 64-bit K8 processors into the mainstream desktop market, where prices are set at US$999 on average, by the end of 2003.

WOW! That'll help their ASP and my... their stock price too, if they sell a lot of them, of course.

 

[NFS4]

Originally posted by: motoamd
From article:

Heye added that AMD is also planning to introduce the 64-bit K8 processors into the mainstream desktop market, where prices are set at US$999 on average, by the end of 2003.

WOW! That'll help their ASP and my... their stock price too, if they sell a lot of them, of course.

REREAD that sentence again...it fooled me too before I went back and read it again

 

[FishTankX]

Heh, yeah, that tripped me up the first time I read it too, before reading NFS4's post.

The computers, the whole machines themselves, cost 999 dollars. Heh.

 

[Dulanic]

SOI on barton? They took that off the roadmap months ago... its back? For some reason I dont think so... but I guess its possible.

 

[dullard]

Originally posted by: FishTankX
Heh, yeah, that tripped me up the first time I read it too, before reading NFS4's post.

The computers, the whole machines themselves, cost 999 dollars. Heh.

I really dislike them avoiding the real question: what will the price point be for the processors themselves. I just bought a great lower end 2.4B P4 computer system for $700 (includes shipping and monitor but not tax). The 2.4 GHz P4 sells for $200. So I could have put a $499 processor in that machine to bring the total to $999. Thus using those numbers, AMD expects a Hammer to be at $499 by the end of 2003. But what if they were thinking of mid-end parts instead of my low end parts, then maybe it is only a $399 price by the end of 2003. That is the problem, AMD never defines what they consider an "average desktop machine". I highly doubt they meant high end parts in the average machine.

Anyways it is looking like Hammer will be $399 to $499 by the end of 2003. But what price will it be in the first half of 2003? One would have to assume AMD plans at least one price cut during that time. So lets add 25% to those prices. That means Hammer starts in the $498 to $624 range. But what if AMD plans two price cuts? Lets add another 25% for that senario: Hammer starts in the $624 to $779 range.

That just makes me frustrated. The total range is $498 to $779 for a Hammer chip. Well I've been estimating that Hammer will start at $500 to $650 for over a year now - we are no closer to where we started.

Originally posted by: Dulanic
SOI on barton? They took that off the roadmap months ago... its back? For some reason I dont think so... but I guess its possible.

I think it would be silly for them to delay Barton by a quarter just to put SOI back in after they just took it off. Even if they really want SOI on Barton, why not release Barton now, and then a SOI version later? For that reason I tend to agree with you. It's possible but I'll wait and see.

I find it funny that that link has these two sentences quite close:
"The Barton-core processors...will hit the market in the first quarter of 2003 as well"
and on the bottom of the page
"AMD Barton-core Athlon XP on market in October"

 

[dexvx]

Wow Dullard, thinking a bit hard there are we now? But you're right. It all depends on what kind of hardware is surrounding that $1000 machine, and if it included a monitor or not. Given the "average" $1000 machine, they use about a $300 processor.

By any standards, I believe that the K8 desktop processor at launch will be over the $500 USD mark.

 

[microAmp]

Originally posted by: NFS4

Originally posted by: motoamd
From article:

Heye added that AMD is also planning to introduce the 64-bit K8 processors into the mainstream desktop market, where prices are set at US$999 on average, by the end of 2003.

WOW! That'll help their ASP and my... their stock price too, if they sell a lot of them, of course.

REREAD that sentence again...it fooled me too before I went back and read it again

Yeah, you're right

 

[KF]

Since the pricing analysis doesn't include any real-life factors, I'll launch into my own.

The selling price of Hammers, or any other CPU, is not determined by AMD's cost of production , and even less what price AMD would like to get. It will be what people are willing to pay. AMD does not sell their top-of-the-line Athlon for $150 because that is all they need to charge or the price they prefer. They can't get any higher price.

If the Hammer out-performs any other mass-produced chip, AMD will be able to charge more than Intel does for their top-of-the-line, and they most likely will. If Hammer falls within the range of performance of an Intel chip, the best AMD can hope to get is what Intel gets, and they won't get that because people would then buy the Intel chip.

No doubt AMD has a price target. But since what Intel will accomplish in the next 3 to 6 months is unknown, AMD can't possibly announce a target price they could stick to. Exactly how much extra people will be willing to pay for a 64 bit chip just because it is touted as such, no one can say for certain, but realistically they won't pay extra for it at all unless it is with an Intel chip. Those are the facts-of-life for AMD.

Although the under $1000 figure for a system is presented as if it came from an AMD representative, you can be sure the number was simply in a question put by the reporter, and all AMD did was agree. That's how reporters pump up stories. As you point out, the number is essentially meaningless, so how could AMD deny it? It is just the art of the reporter that makes it almost sound like it means something.

As for the statement about Barton and SOI, I wonder if the AMD rep didn't just answer yes to a confusing, convoluted question from a reporter, and the reporter took this unexpected bonus and ran with it.

 

[Wingznut]

Originally posted by: KF
The selling price of Hammers, or any other CPU, is not determined by AMD's cost of production , and even less what price AMD would like to get. It will be what people are willing to pay.

Exactly. It's just basic economics.

Originally posted by: KF
If the Hammer out-performs any other mass-produced chip, AMD will be able to charge more than Intel does for their top-of-the-line, and they most likely will.

While having performance will help, it certainly doesn't guarantee that people will want to pay top dollar for it. Other factors (such as branding, availability, etc...) weigh heavily. The general buying public (both home and especially business) still trust the Intel brand name more than AMD. And if AMD can't supply the cpu's in volume, OEM's are less likely to dedicate a line of product around them. (Which, as far as I know, is the main reason Dell doesn't carry AMD. They simply cannot guarantee to supply Dell the way they require.)

Originally posted by: DigiTimes
The Barton-core processors, which will also adopt the SOI manufacturing technology, will hit the market in the first quarter of 2003 as well, Heye said.

That sounds like Mr. Heye said, "The Barton-core processors will hit the market in the first quarter of 2003 as well." and DigiTimes added the part about SOI. To date, I haven't heard AMD say anything but no SOI for their Barton cpu's, so I wouldn't put much stock about that in this article.

Originally posted by: dullard
I really dislike them avoiding the real question: what will the price point be for the processors themselves. ...

Well, I can definitely see how you are very curious to know how much these will go for, but of course they aren't going to lock themselves into a price point more than six months before they are ready. Who knows how much cpu's will be selling for in six months or more??? From their perspective, nothing can be gained from them setting a price at this time.

 

[ST4RCUTTER]

Even if they really want SOI on Barton, why not release Barton now, and then a SOI version later? -dullard


Simple reason...cost. The design considerations for using SOI are substantial. Many seem to be under the impression that migrating a core design from bulk substrate to SOI is a simple matter of changing out the wafers, it's not nearly that simple. Besides dealing with self-heating issues, decoupling capacitance, increased device leakage and floating body effects, which all have to be designed around, you have changes to the core to consider. Given the current economy, they just don't have the capital to take on two iterations of Barton.

 

[Aquaman]

What is SOI and what does it do?

Cheers,
Aquaman

 

[Wingznut]

SOI is Silicon On Insulator. Basically it's a type of wafer that has Silicon Dioxide built into the substrate to help with gate leakage, junctive capacitance, and other such issues that deter performance as transistors get smaller and smaller.

But SOI does have it's drawbacks also. Namely a significantly higher price (for the wafers themselves) and higher native defects.

That's the short answer. There are a couple of different types of SOI. Partially depleted, which is what IBM and AMD is using, and fully depleted which Intel has stated that they are researching.

I'll see if I can dig up a good link for ya.

EDIT: Here you go.

 

[NICKCARTER]

Thanks Wingznut PEZ for the link!
I think that one reasen for the Barton-core to get SOI is HEAT,
The head-room for heat is not that much on the k7.

Isolating the active transistor from the rest of the silicon substrate reduces the electrical current leakage that would otherwise degrade the performance of the transistor.

=Cooler running chip

 

[KF]

Great link, Wingznut Pez

To summarize, by isolating the transistors from the substrate in a more effective way, leakage currents and capacitance are reduced.
This reduces the voltage and current needed to switch at given speed. Power disapation is reduced. Switching at a higher speed means the fastest operating frequency can be higher.