First hints of AMD going fabless

[xpresso]

Here we go, I mentioned this a few times in the past...AMD cannot compete in process technology and let's face it: half the company is fabless (ATI). I think I am right

http://www.marketwatch.com/news/story/amd-swings-loss-weighs-restructuring/story.aspx?guid=%7BFB6128FB%2DAE93%2D424C%2D9E6D%2D105CEE504CE9%7D&siteid=yhoof

 

[qcmadness]

Here we go, I mentioned this a few times in the past...AMD cannot compete in process technology and let's face it: half the company is fabless (ATI). I think I am right

http://www.marketwatch.com/news/story/amd-swings-loss-weighs-restructuring/story.aspx?guid=%7BFB6128FB%2DAE93%2D424C%2D9E6D%2D105CEE504CE9%7D&siteid=yhoof

In fact I don't think this is a wise idea at all.

 

[xpresso]

Here we go, I mentioned this a few times in the past...AMD cannot compete in process technology and let's face it: half the company is fabless (ATI). I think I am right

http://www.marketwatch.com/news/story/amd-swings-loss-weighs-restructuring/story.aspx?guid=%7BFB6128FB%2DAE93%2D424C%2D9E6D%2D105CEE504CE9%7D&siteid=yhoof

In fact I don't think this is a wise idea at all.

Wise or not this is business reality. AMD buying ATI forcing them down this route even more. When half the company is fabless, it gets interesting. AMD is in danger of falling 18 months behind in process technology. Simply AMD cannot compete even with IBM providing the R&D 8O

 

[qcmadness]

Wise or not this is business reality. AMD buying ATI forcing them down this route even more. When half the company is fabless, it gets interesting. AMD is in danger of falling 18 months behind in process technology. Simply AMD cannot compete even with IBM providing the R&D 8O

In contrast, I would like AMD to produce their CPU / GPU / Fusion processors in house.

 

[xpresso]

I dont see a hint.

While company executives declined to offer specific details, such as whether it would involve layoffs or major cost cuts, Chief Executive Hector Ruiz hinted the restructuring could involve an "asset-light" strategy, suggesting AMD may outsource more of its chip production, an approach Texas Instruments Inc. pursued several years ago

The talk of asset-lite is an understatement by Ruiz, here's the facts:
1. ATI division is fabless, so 100% of graphics products are outsourced
2. AMD has no true modern Si R&D division, they are more into development, true R&D services provided by IBM to stay in the game.
3. AMD relies already on Charter for a portion of their microprocessors

So asset-lite? what else is there left to outsource other then the internal portion of fabbed products?

 

[qcmadness]

I dont see a hint.

3. AMD relies already on Charter for a portion of their microprocessors


In fact Fab 36 alone should have enough capacity to supply all processors needed by AMD.

 

[xpresso]

Wise or not this is business reality. AMD buying ATI forcing them down this route even more. When half the company is fabless, it gets interesting. AMD is in danger of falling 18 months behind in process technology. Simply AMD cannot compete even with IBM providing the R&D 8O

In contrast, I would like AMD to produce their CPU / GPU / Fusion processors in house.

then better find a way to fund it...this business is not kind if you have high exposure in leading edge process development and you're 1-2 years behind your biggest competitors. Barcelona is going to make or break the company, AMD will not report a profit for how long?? Fuzion will not bring in premium money in my opinion and will be delegated to low cost entry to maybe mid level PC's. AMD even admitted that.

 

[xpresso]

I dont see a hint.

3. AMD relies already on Charter for a portion of their microprocessors


In fact Fab 36 alone should have enough capacity to supply all processors needed by AMD.

ummm, then why use Charter? low yields?

 

[qcmadness]

In fact Fab 36 alone should have enough capacity to supply all processors needed by AMD.

ummm, then why use Charter? low yields?

I don't know the exact reasons. But I think the ramp-up of Fab 36 is too slow.

The capacity of Fab 36:
20K 300mm wafer by the end of 2007.

Assume 60% yield with 65nm process.
The die size of a Athlon64 x2 processor: 120mm^2

20,000 * (300mm / 2)^2 * Pi * 60% / 120mm^2 = 7.1M per month

You can produce ~22M CPUs in a quarter, which accounts for about 25-30% of total CPU shipments.

 

[xpresso]

In fact Fab 36 alone should have enough capacity to supply all processors needed by AMD.

ummm, then why use Charter? low yields?

I don't know the exact reasons. But I think the ramp-up of Fab 36 is too slow.

The capacity of Fab 36:
20K 300mm wafer by the end of 2007.

Assume 60% yield with 65nm process.
The die size of a Athlon64 x2 processor: 120mm^2

20,000 * (300mm / 2)^2 * Pi * 60% / 120mm^2 = 7.1M per month

You can produce ~22M CPUs in a quarter, which accounts for about 25-30% of total CPU shipments.

But we know that Barcelona is like 250 mm^2?? I don't remember exactly other than the die size is huge and the market has turned into a dual-quad core market where the money is.

 

[qcmadness]

But we know that Barcelona is like 250 mm^2?? I don't remember exactly other than the die size is huge and the market has turned into a dual-quad core market where the money is.

Barcelona: ~290mm^2
Kuma: ~180mm^2
Rana: ~130mm^2

So I think AMD should still expand Fab 30 and Fab 36 if they want to get back the market share with a solid product line.

 

[copenhagenlc]

if you havent seen there stocks, there starting to go up a bit

 

[fletch420]

you also have to account for wafer loss- if you start 20k you will not get 20k out. modules take wafer losses daily all over the fab. And yes I work in one so I'm not guessing. I hope they make it just to see another company in the mix but they sure have bet the farm on this one. The purchase of ATI may have been not in the best interest of the company's long term growth. But maybe they have an ace....we shall see.\

 

[exit2dos]

This is a very tricky topic. Asset light could mean a totally different thing than fabless.

Furthermore i think its a bad idea to go fabless ,for them. I think these are desperate words for desperate times and if I had 20 billion I would shut them up and retire many of them.

I doubt they would go "fabless". If anything, I could see them selling their fabs to an investment group or developer - then leasing the fabs back from them.

This is common in industries such as retail - where they sell a warehouse/distribution center (or even the trucks,etc.) - then lease it back. Although it's more expensive in the long-term, It does raise a good amount of near-term cash without having outside investors in the company itself.

 

[Major_Spittle]

Good will and intentions does not make CPU engineering break throughs and fix production yeild issues. A person cannot just say " Engineers, create a better processor than intel in 3 months or your fired " and suddenly get results. AMD is out classed by intel in an industry that is being propelled by multibillion dollar R&D investments creating mordern day technological miricles.

IBM does not even have a HKMG 45nm process to sell AMD right now. This means tooling of a fab cannot even begin yet. Without a fab tooled you can't even Qualify the tools. Without the tools you can't begin to dial in a tool on test chips. Without test chips you can't get sample transistors. Without test chip samples you can't create process specs. Without process specs there are no processes to develop off of to make lead lots to spend months testing/changing/developing to create a prouct you want that yeilds enough to start ramping the product that will take 7-8 months to get to market in number. It is a long difficult processes and as of yet IBM does not even have a HKMG 45nm process to sell AMD right now

AMD is real far behind. They may never get a 45nm process. 45nm may kick their azz and they will never make something yeild that is that small. Nobody even knows yet if 32nm production chips are possible, they have yet to be invented realisticly.

AMD is in a real bad position now. Just got 65nm, but seems to be not very solid for the K10's if they are releasing at 2.2-2.5 like rumored.

 

[xpresso]

This is a very tricky topic. Asset light could mean a totally different thing than fabless.

Furthermore i think its a bad idea to go fabless ,for them. I think these are desperate words for desperate times and if I had 20 billion I would shut them up and retire many of them.

I doubt they would go "fabless". If anything, I could see them selling their fabs to an investment group or developer - then leasing the fabs back from them.

This is common in industries such as retail - where they sell a warehouse/distribution center (or even the trucks,etc.) - then lease it back. Although it's more expensive in the long-term, It does raise a good amount of near-term cash without having outside investors in the company itself.

Thata's very risky though in the position they are in...it is common though. still it screws their fab people.

 

[Major_Spittle]

FYI the IBM partners did develope High k and it was announced this year. I do believe though it is not a completed project, 65nm has been a nightmare for them for sure.

Actually they didn't. Intel showed working processors using the technology. IBM showed a paper they wrote on it saying they think they could build a transitor according to the data.

IBM was told by the media that intel gave them a predated press release that they had the technology. IBM told the media to release that they also have that technology. It was no coincedence that they came out at the same time.

 

[Glacier]

While company executives declined to offer specific details, such as whether it would involve layoffs or major cost cuts, Chief Executive Hector Ruiz hinted the restructuring could involve an "asset-light" strategy, suggesting AMD may outsource more of its chip production, an approach Texas Instruments Inc. pursued several years ago.

I would be just a little worried if I worked for AMD, I'm sure layoffs are also being considered in their asset-light strategy. I will never understand why AMD spent/wasted all that money on ATI, they practically put a nail in their own coffin. With any luck, Barcelona will give AMD the boost they need.

 

[xpresso]

Good will and intentions does not make CPU engineering break throughs and fix production yeild issues. A person cannot just say " Engineers, create a better processor than intel in 3 months or your fired " and suddenly get results. AMD is out classed by intel in an industry that is being propelled by multibillion dollar R&D investments creating mordern day technological miricles.

IBM does not even have a HKMG 45nm process to sell AMD right now. This means tooling of a fab cannot even begin yet. Without a fab tooled you can't even Qualify the tools. Without the tools you can't begin to dial in a tool on test chips. Without test chips you can't get sample transistors. Without test chip samples you can't create process specs. Without process specs there are no processes to develop off of to make lead lots to spend months testing/changing/developing to create a prouct you want that yeilds enough to start ramping the product that will take 7-8 months to get to market in number. It is a long difficult processes and as of yet IBM does not even have a HKMG 45nm process to sell AMD right now

AMD is real far behind. They may never get a 45nm process. 45nm may kick their azz and they will never make something yeild that is that small. Nobody even knows yet if 32nm production chips are possible, they have yet to be invented realisticly.

AMD is in a real bad position now. Just got 65nm, but seems to be not very solid for the K10's if they are releasing at 2.2-2.5 like rumored.

Major,

Totally agree. IMO, AMD is way behind, since their 65 nm is not performing up to what a 65 nm process should deliver...IBM is knows for fancy R&D and breakthroughs but not so nice to manufacture...
AMD is getting sucked into making 65 nm work and IBM just started runing 65 nm process in NY (or they say they started 65 nm production now from their earnings release 2 days ago). AMD in my opinion is 18 months behind, and yes their 45 nm may not work. I think the company has a critical decision to make right now. AMD cannot build another 300 mm fab for now period. Better get revenues higher than Texas Instruments to be able to build a fab

 

[xpresso]

Vern,

There was no actual demo, if this is such a breakthrough, why wait till 2009?? when their competitor Intel is bringing it to the masses in 2007? sounds odd don't you think? Where is IBM 65 nm? :?

 

[Major_Spittle]

You have a link validating IBM partners high k paper launch? :lol: I wasnt aware they hadnt built a chip with it for testing. Ill go back through my eetimes news releases to find a link.

I am not denying thats the case, I just missed the memo.

http://www-03.ibm.com/press/us/en/pressrelease/20980.wss

here are the important parts:

Working with AMD and its other development partners Sony and Toshiba, the company has found a way to construct a critical part of the transistor with a new material, clearing a path toward chip circuitry that is smaller, faster and more power-efficient than previously thought possible.

Incremental work leading up to this achievement had been published earlier by IBM in scientific journals and presented at chip technology conferences. IBM plans to publish the summary of this final achievement in a similar forthcoming venue.

Can't touch it, see it, or feel it. But they developed a way to construct it according to their papers. 8O

 

[exit2dos]

Thata's very risky though in the position they are in...it is common though. still it screws their fab people.

They really wouldn't be screwed in this aspect, as they will still be AMD employees. It's just the tangibles (the buildings, equipment, etc.) that are sold then leased. From an employee standpoint - It's business as usual since nothing actually changes operations wise.

The employees will start being screwed as AMD starts cost-cutting. On top of layoffs, they might seek reductions in pay or benefits if things start getting really bad. On one hand, this is unavoidable as payroll/benefit expenses are huge for any company - so this is often where trimming begins. On the other hand, this creates turnover among essential people (who, naturally feel they may be next, so they leave at the first opportunity) - not to mention "intangibles" such as employee morale, company loyalty, etc.

 

[Major_Spittle]

Vern,

There was no actual demo, if this is such a breakthrough, why wait till 2009?? when their competitor Intel is bringing it to the masses in 2007? sounds odd don't you think? Where is IBM 65 nm? :?

Read the post again dewd :? i am agreeing with the major. mmmkay?

Holey Sheet, we agreed on something. * makes menal note of time and date, then buys a lottery ticket *

8O

 

[xpresso]

Vern,

There was no actual demo, if this is such a breakthrough, why wait till 2009?? when their competitor Intel is bringing it to the masses in 2007? sounds odd don't you think? Where is IBM 65 nm? :?

Read the post again dewd :? i am agreeing with the major. mmmkay?

my bad :lol: :lol:

 

[xpresso]

Vern,

There was no actual demo, if this is such a breakthrough, why wait till 2009?? when their competitor Intel is bringing it to the masses in 2007? sounds odd don't you think? Where is IBM 65 nm? :?

Read the post again dewd :? i am agreeing with the major. mmmkay?

my bad :lol: :lol:

welcome to the human condition.

now really is it dewd or dood. I know it's not dude, I think dood is more lie it