IBM Fishkill Plant with Problems?

[MacRumors]

This Reuters report (http://story.news.yahoo.com/news?tmpl=story&cid=581&ncid=581&e=3&u=/nm/20030717/tc_nm/tech_ibm_chips_dc) indicates that IBM's new 3 billion microchip factory is having production problems.

Demonstrating the risk of huge investments in advanced chip-making tools, International Business Machines Corp. on Wednesday cited production problems at the gleaming chip plant in East Fishkill, New York, as a reason why its microchip business would lose money this year.

It's unclear if this affects PowerPC 970 production.

[mac15]

ouch, this is one thing apple doesn't need. But to make money you need to spend money

[arn]

Does anyone know for a fact that the Fishkill plant is where 970s are being produced?

as far as I recall - this has all been speculation and rumor.

arn

[ZildjianKX]

And people were saying IBM was going to kick Intel's ass... sigh...

[Magicite]

. . . but I believe the article was referring to economic problems rather than production problems. i.e., not making a profit.

Anyone care to correct me? This may not even be the article that I read on the matter earlier today, but I'm assuming it is.

edit -- erp, did I misread something? Time for sleep!

edit 2-- ok, I swear the article was different when I read it early this afternoon . . .

[eric_n_dfw]

Originally posted by arn
Does anyone know for a fact that the Fishkill plant is where 970s are being produced?

as far as I recall - this has all been speculation and rumor.

arn
From sidebar of http://www.apple.com/g5/ibmprocess.html
At IBM’s state-of-the-art 300-mm chip fabrication facility in East Fishkill, NY, front opening unified pods (FOUP) hold silicon wafers during various stages in a weeks-long process.

[MrMacMan]

Uh oh, not a good thing.

[arn]

Originally posted by eric_n_dfw
From sidebar of http://www.apple.com/g5/ibmprocess.html
At IBM’s state-of-the-art 300-mm chip fabrication facility in East Fishkill, NY, front opening unified pods (FOUP) hold silicon wafers during various stages in a weeks-long process.

ah - thanks :)

I guess it may impact apple then.

arn

[Abstract]

Also, Steve said that this is where the chips were made in his keynote. ;)

If the G5 towers aren't shipped in late August/September, heads will roll. :( I don't even want to imagine what would happen if Steve announced a delay until December or something.

*knock on wood*

[Flowbee]

Originally posted by Magicite
. . . but I believe the article was referring to economic problems rather than production problems. i.e., not making a profit.


The quote is: "...International Business Machines Corp. on Wednesday cited production problems at the gleaming chip plant in East Fishkill, New York..."

I think that means production problems. At the risk of stating the *way* too obvious... this could be a very bad thing for Apple. Not good to hype the fastest personal computer on the planet, and then not be able to deliver them in mass quatities. I hope this refers to past production problems that have since been ironed out.

[BostonMJH]

Originally posted by arn
Does anyone know for a fact that the Fishkill plant is where 970s are being produced?

as far as I recall - this has all been speculation and rumor.

arn


Yes when Steve Jobs and the exec from IBM get on stage at the WWDC ans say that the G5 is made at the state of the art facility and it cost 3 Billion....Its Fishkill.;)

[trog]

I think its worth mentioning that my order of a dual G5 was originally scheduled to be delivered "on or before" september 2nd. A few days ago that was quietly changed to August 29th instead of Sept 2nd.

There isn't much difference between those two dates, but it seems unlikely they would go to the trouble to change the shipping date to earlier if they were expecting production problems.

We'll see, but this is only one article and I'd bet this fab is used for more than just 970 production (POWER 4 AND 5s??).

[BostonMJH]

If this is true.

Apple :mad:

IBM :(

MOTO :D

Intel/Microsoft :)

[Gyroscope]

In an related announcment Apple informs customers who pre-ordered their G5 towers, about revised specs on the following:

Dual G5 2 Ghz-> Dual G5 1.8 Ghz

Single G5 1.8 Ghz -> Single G5 1.6 Ghz

Single G5 1.6 Ghz -> Single G5 1.4 Ghz


Shipment date late september-early october, no price changes.


:) :p :o :) :( :cool:

[nagromme]

One possible interpretation: the problems were solved, and were expensive to fix! Probably still a drop in the bucket to the plant's initial cost.

[P-Worm]

Man this isn't boss at all...

P-Worm

[raschild]

Hope Steve can get clear of this!

[MrMacMan]

Originally posted by BostonMJH
If this is true.

Apple :mad:

IBM :(

MOTO :D

Intel/Microsoft :)

a) how is this anyones fault?
b) why smile at Moto, they STILL can't make the G4 a good chip, max at 1.42 overclocked where as G5 = max 2.0 not overclocked

And this is GOOD for Intel?

Wtf!

[Nermal]

Originally posted by MrMacman
a) how is this anyones fault?
b) why smile at Moto, they STILL can't make the G4 a good chip, max at 1.42 overclocked where as G5 = max 2.0 not overclocked

And this is GOOD for Intel?

Wtf!

I think he means that Apple will be mad, rather than him being mad at Apple, and so on. That's how I interpreted it anyway.

[projectParanoia]

The situation at the IBM Fishkill plant is that there are some manufacturing problems, but these will be fixed in the future.

This is also IBM's first attemt to manufacture chips for other companies, therefore there is a lot of room for improvement.

The economic state is contributing to the problems in the facility because the plant is currently being underused. When the economy picks up, the plant will gain more efficiency and that translates into profit.

IMO, this could mean that Apple would order more PowerPC 970's aka the G5 from IBM so that the Fishkill plant can be more efficient and make a profit. We mac users would then receive lots of G5-based systems. Imagine that, an iBook G5...:D I should stop before I druel to death. This is, of cource, just an opinion.

[rjwill246]

This doesn't mean anything yet. For all we know the problem is fixed. If not, we don't know the extent of it. If we did, we might be........ informed!
This rumour is getting is getting a life of its own... it's time for fish kill!

[nagromme]

The article cites "production problems" AND a lack or orders for as many chips as they'd like to sell. Sounds like a temporary problem for IBM's profits, nothing worse than that.

[hacurio1]

Originally posted by Nermal
I think he means that Apple will be mad, rather than him being mad at Apple, and so on. That's how I interpreted it anyway.

I'm sure that's what he ment. I agree with you

[projectParanoia]

This is not a rumor its in the news. Specifically, the Reuters link in the first post. link (http://story.news.yahoo.com/news?tmpl=story&cid=581&ncid=581&e=3&u=/nm/20030717/tc_nm/tech_ibm_chips_dc)

[GetSome681]

I think you guys are taking "production problems" in the wrong way. This is business talk that refers to the "entire" process of production, which includes the economics of that production as well. I think this is what they are referring to, as in they aren't meeting their scheduled ROI, which is a result of the economy not giving them a chance to fully utilize the plant. IMO, this isn't anything to get your panties all uptight about.

[tex210]

I believe this means IBM will take a loss on these first chips out the door. They have some chips process as bad, that's wasted material....which cuts their profit. They've already sold a certain number to Apple at a certain price. In Apples broadcast about financial statements, this comes up. Apples response is, they will produce, and as for apple, there is still a profit, just not as much as they would have liked. (The next batch was a higher price?)
They will get better and more efficient with each yield though. Once IBM goes .09 the flaws become more likely, and the process of learning begins again.

[Freg3000]

Does anyone know how many products IBM makes at this place? Of course it is not solely for the 970, but does IBM produce many different chips there, and this problem is with one specific chip?

[acj]

In related news, Prescott (newest Pentium 4 @ 3.4-5GHz) is said to be possibly having higher yields than any microprocessor ever made.

[OKComputer]

sounds to me like IBMs is suffering from a common business world flu called the "wayyy too much overhead costs" bug. It happens. Its their first run at a new processor. of course there are going to be flaws in the production process. if i recall the same thing is rumored to have happened with the first G4 towers. It seems to me that this reporter is just sensationalizing water under the bridge. I am sure these production problems are long fixed. Count on your G5 in august those of you lucky enough to order one.

This doesnt look good for apple investers though.

its a twist of IBMs words to get a story out of nothin.

[nagromme]

Someone here (http://forums.maccentral.com/wwwthreads/showflat.php?Cat=&Board=general&Number=495647&page=0) seems to think that the problems could not affect the G5, because they were with "low-k dialectric based silicon."

I don't know what that means, I just like saying it :)

[catalystx]

It's really entertaining to see the range of pessimists and optimists react to news like this. Can you say snap judgements?

Relax, IBM is a huge company with lots of funds to throw at "production problems" which are most likely just too high operating costs. This isn't another situation like Motorola not being able to fab above 450MHz, their facilities are puny compared to Fishkill.

[shadowself]

I would guess Apple planned on announcing (and very, very shortly there after shipping) the G5s at WWDC when they moved WWDC. The move was to accomodate the announcement of the G5 machines as originally projected when the move was made.

So... IBM informs Apple of production problems in late calendar Q1 and more info on problems in Q2. However, by the WWDC IBM assures Steve and company that they have the problems solved.

Thus... Steve goes on stage and declares the machines will be available "in August" -- not "tomorrow" or "next week" as originally planned.

Therefore I would expect there is nothing to worry about. Apple clearly knew what was going on at Fishkill before this week's financial announcements conference call with Apple's senior management. During that call Apple rather emphatically still stuck to the "shipping in August" date. Thus I would get from that statement that Apple believes rather firmly Fishkill's problems are behind it.

Thus I don't see this as a problem going forward as far as G5 availability.

The real problem will be perception. If this story gets widely distributed and loudly hawked by the pro Wintel crowd then this could hurt Apple's stock price. It could also possibly even impact people planning on waiting to purchase a G5. If they perceive, because of the Wintel media machine, that the wait might be months longer than originally anticipated, then they just may buy Wintel machines instead. Such is the power of FUD.

[wilco]

Originally posted by projectParanoia
I should stop before I druel to death. This is, of cource, just an opinion.

Just an opinion? Thanks for clearing that up!

[Multimedia]

Originally posted by Gyroscope
In an related announcment Apple informs customers who pre-ordered their G5 towers, about revised specs on the following:

Dual G5 2 Ghz-> Dual G5 1.8 Ghz

Single G5 1.8 Ghz -> Single G5 1.6 Ghz

Single G5 1.6 Ghz -> Single G5 1.4 Ghz

Shipment date late september-early october, no price changes.

:) :p :o :) :( :cool:
This is your idea of a bad joke right?

[fpnc]

During yesterday's financial teleconference an analyst asked Fred Anderson (Apple's CFO) about yield problems at IBM's East Fishkill plant. Mr. Anderson said that with any new product like the G5 there would always be some schedule risk but that they had complete confidence in IBM's manufacturing capabilities and that Apple was still expecting to ship the first G5 PowerMacs in August. He then added that he could not tell the analyst on this day whether such production problems would affect Apple's ability to meet PowerMac G5 demand in the coming months.

[tychay]

Some doomsayers are reading too much into this report.

IBM microprocessor is trying to explain away last quarter losses by saying that they had both lower yields and less demands (http://www.siliconstrategies.com/article/showArticle.jhtml?articleId=12800791). Since IBM closed a lot of its smaller fabs in order to "bet the farm" on this one, it sounds to me like they got lucky.

Imagine if they had higher yields and less demand? Excess capacity. Or worse, imagine if they had poorer yields and more demand? Expect delays on your pretty dual-G5.

Imagine if Fishkill had just come online and they were just finding out now how bad their yields are instead of last quarter? You'd see some royally pissed-off Mac fanatics then! Remember four years ago when some poor slobs paid more $ for less megahertz for their PowerMac G4? ;)

You're talking about an area (12") where only DRAM manufacturers (and Sony) dare to tread and plan on matching DRAM in process shrink (130nm to 90nm). What do you expect? Or are all these DIMM boards that I have to return as defective aberrations?

IBM produces chips for many companies at this plant (the PPC 970/G5 and Apple's ASIC are just two of them) including themselves. When they commissioned this plant three years ago, it was after the boom and many questioned whether IBM, a company known for their services and intellectual property, should be moving into the cutthroat world of manufacturing.

Now that demand is soft and yield is low and all the naysayers are coming out of the woodwork and saying crap like
In contrast to IBM, advanced chip equipment is easing the way for Intel Corp.

Hello? All Intel chips come out of 8" fabs and they're having serious yield issues with their 12" conversion in Arizona. What's so "advanced" about that? But I guess its not Reuters job to do real analysis.

Then again, they're no different than Tom's Hardware (http://www.tomshardware.com/cpu/20030623/p4_3200-02.html) which claims that somehow that Intel manufacturing with a 90nm process on 300mm wafers in a fab that isn't even up and running yet on a chip that doesn't exist yet is somehow magically come out like gangbusters on yield.

The G5 uses a 9-layer copper damascene process (http://www.apple.com/g5/ibmprocess.html). Intel's P4 uses 6 layers (AMD Athlon XP uses 9). BTW, yield is inversely related to the number of layers. How do you think Intel turns a profit and AMD consistently turns in a loss (http://www.pcpro.co.uk/?[url=http://www.pcpro.co.uk/news/news_story.php?id=44812)? This chip needs a state-of-the-art facility 3 year in the making if we expect this thing in quantity at a reasonable price. What is Motorola fabing at?
MOTO :D
Yeah right next you'll be telling me,`` IBM's been punked!'' That'll be worth the laugh.

IBM produced an expensive fab that went operational when demand for silicon is really low. Hopefully companies like Apple (who needs 3 chips for every 2x2 G5 from that plant) and AMD (Dresden is starting to look a little long in the tooth, rumors are that AMD's R&D is already in Fishkill) will change the demand equation. Time allows IBM and others to deal with yield issues. 300 mm is the future, and, unlike their competition, IBM has a plant today.

If demand is so low, what incentive do others have of finishing refits for a 12" fab (http://siliconvalley.internet.com/news/article.php/2236141) now? Oh wait, it's too late, they've already sunk in the money and their future designs on it. Or are we going to claim, like Reuters implied, that the fall of 2003 looks better economically than that of 2000 and IBM Fishkill was a big mistake.

Hah!

[Gyroscope]

Originally posted by Multimedia
This is your idea of a bad joke right?


Yes man it is.

I was jokingly refering to G4 fiasco in 1999, that didn't seem much of an joke at that time.

[hvfsl]

Originally posted by ZildjianKX
And people were saying IBM was going to kick Intel's ass... sigh...

Intel has loads of problems of its own, there was a recent problem with netware software not working on Pentium M notebooks.

[panphage]

Originally posted by Freg3000
Does anyone know how many products IBM makes at this place? Of course it is not solely for the 970, but does IBM produce many different chips there, and this problem is with one specific chip?

Reading the Reuters article will answer your question. Fishkill is also supposed to make chips for Nvidia, Analog Devices, and Xilinx Inc (whatever the hell that is...)

The same activity will calm all the "sky is falling" people. This is IBMs excuse for a loss from Fishkill rather than a profit. I get the sense (mostly from lowballed rumors of a top speed of 1.6Ghz or maybe a 1.8, and from the Stevenote) that apple and IBM are both quite pleased (perhaps giggly) over the yields on the 970. The article makes the "production problems" seem more like an economic issue than a physical problem with making chips.

[macnews]

Boy some of you really need to learn how to read financial calls and reports.

Don't read too much in to this as the only reason it is "shocking" news is that the division lost money and wasn't expected to. This was in part due to "production problems and plant underutilization [which] took away $45 million from IBM's overall profit." The article goes on to talk about how companies weren't ordering as many chips due to economic down turns.

Realize this plant was made (acording to the article and other related news stories) to design customer SPECIFIC chips. Thus they have some learning curve.

My take is they threw in the production problems as an excuse to balance out the reason why they lost $$$. This is a common tactic to toss in a small (but true) reason why you lost money to balance out the bigger reason. Why do this? IBM can fix, control, set dates when "production problems" will be fixed. IBM can't fix, control, set dates on sales or a bad economy. Investors can stand production problems if you have a product in demand. Investors start to bail when you have a product no one wants (or wants on the scale you need it to be profitable).

I do see an upside in this report - underutilization. If other companies can't fill there bills maybe IBM will offer cheaper prices for Apple to increase their orders - or (as mentioned) offer other special chips ealier since they now have room to make them (G5 laptops?).

[tam]

Originally posted by panphage
Reading the Reuters article will answer your question. Fishkill is also supposed to make chips for Nvidia, Analog Devices, and Xilinx Inc (whatever the hell that is...)

Xilinx are one of the (if not the) biggest player in the FPGA market. It's programmable hardware, very useful when testing an ASIC design, and often in finished products as well because of the lower cost. Perhaps not widely known to public, but hardware designers are more than familiar with them.

[Brother Mugga]

In a related theme...

Is it just me, or did anyone else notice that when Steve Jobs was talking about the G5 in his WWDC keynote, the sound 'dropped out' the first time he mentioned where they were made?

"And [IBM] also happen to have the world's most advanced chip fabrication in [---------------] on the planet."

[Of course he might not have mentioned the production site here; it's tricky to work in linguistically (if you see what I mean). But if not, what DID he say? Die size? Process? What?]


I'm not saying that G5s aren't being made at Fishkill [Steve actually mentioned it explicitly a few minutes later and IBM's Ned Flanders (the III) clearly confirmed this in his comments*]. I'm just curious as to what Steve said and why they seem to have edited it out.

Of course, there's always the possibility that Steve simply made a boo-boo and mentioned a Motorola factory (or maybe even the AMD fab that they'll be using in 2005 ;) ...).

Just a thought anyway.


Brother Mugga

* Also - is it just me, or did 'Mr. Adobe' look like a scrubbed Barney ["...and at these prices (BUUUUURP!!), you're likely to get one..."] and that bloke from Luxology come over as a weedy-voiced Troy McClure...?

Anyone...anyone...?

Oh, just me then.

[Lyle]

Originally posted by Flowbee
"...International Business Machines Corp. on Wednesday cited production problems at the gleaming chip plant in East Fishkill, New York..."So Apple really only has a problem if they were counting on an adequate supply of the gleaming chips, right? Is it an extra cost item to get one of the new PowerMacs with the gleaming chips?

[AngryAngel]

Originally posted by projectParanoia

This is also IBM's first attemt to manufacture chips for other companies, therefore there is a lot of room for improvement.


Ummm... you think IBM hasn't made CPU chips for Apple (and loads of others) before?

And why would the fact that the chips are being shipped to a third party mean that their process would go to pieces?

[LinuxGigolo]

Originally posted by AngryAngel
Ummm... you think IBM hasn't made CPU chips for Apple (and loads of others) before?

And why would the fact that the chips are being shipped to a third party mean that their process would go to pieces?

I was thinking the same thing. Not to mention the fact that IBM makes the G3 that Apple uses in iBook and has for quite some time now.

[zac4mac]

Some people have mighty short memories.
I've got a PPC601/75 card in a drawer that has "IBM" all over it. Made in 1995.
Also, I had a robot system at Fishkill in 89-91 that transported green/sintered ceramics in and out of a Hydrogen fired oven - substrate for BIG chips.

Z

[Rustus Maximus]

Originally posted by trog
I think its worth mentioning that my order of a dual G5 was originally scheduled to be delivered "on or before" september 2nd. A few days ago that was quietly changed to August 29th instead of Sept 2nd.

There isn't much difference between those two dates, but it seems unlikely they would go to the trouble to change the shipping date to earlier if they were expecting production problems.

We'll see, but this is only one article and I'd bet this fab is used for more than just 970 production (POWER 4 AND 5s??).

Originally posted by GetSome681
I think you guys are taking "production problems" in the wrong way. This is business talk that refers to the "entire" process of production, which includes the economics of that production as well. I think this is what they are referring to, as in they aren't meeting their scheduled ROI, which is a result of the economy not giving them a chance to fully utilize the plant. IMO, this isn't anything to get your panties all uptight about.

I think both these quotes nicely sum this article up...it's a non-issue for Apple apparently, otherwise they would be frantically changing their shipping dates in the OTHER direction.

(begin paranoid rant)

I know I'm a conspiracist (is that a word?), but this Wintel beast we are up against is desperate for any piece of news that might indicate trouble with a chip that will possibly change the face of computing. The beast isn't that concerned about AMD's 64 bit offering because it runs Windows...but Microsoft is afraid of OSX, hence the necessity to constantly downplay Apple's hardware and usefulness in the real computing world, blah blah blah...

(end paranoid rant)

IBM's plant is having birthing pains, if they really are, and they will get past this. The ride will only improve from here. :)

Ahhhh but these little news bits sure do bring out the trolls don't they?

...the trolls come out at night (bonk ee boowwww)....the trolls come out at niiiigghhhhtttt....(bonk booww)...the trolls come out at night...

[iconmaster]

Originally posted by Brother Mugga
Is it just me, or did anyone else notice that when Steve Jobs was talking about the G5 in his WWDC keynote, the sound 'dropped out' the first time he mentioned where they were made?

"And [IBM] also happen to have the world's most advanced chip fabrication in [---------------] on the planet."
Watch it again; Steve just got a little tongue-tied. He was about to say, "world's most advanced chip fabrication in the world," realized that would be utterly redundant, and used the power of synonymy to quickly come up with something only mostly redundant. :)

[macphisto]

Originally posted by shadowself
I would guess Apple planned on announcing (and very, very shortly there after shipping) the G5s at WWDC when they moved WWDC. The move was to accomodate the announcement of the G5 machines as originally projected when the move was made.

So... IBM informs Apple of production problems in late calendar Q1 and more info on problems in Q2. However, by the WWDC IBM assures Steve and company that they have the problems solved.

Thus... Steve goes on stage and declares the machines will be available "in August" -- not "tomorrow" or "next week" as originally planned.

Therefore I would expect there is nothing to worry about. Apple clearly knew what was going on at Fishkill before this week's financial announcements conference call with Apple's senior management. During that call Apple rather emphatically still stuck to the "shipping in August" date. Thus I would get from that statement that Apple believes rather firmly Fishkill's problems are behind it.

Thus I don't see this as a problem going forward as far as G5 availability.

The real problem will be perception. If this story gets widely distributed and loudly hawked by the pro Wintel crowd then this could hurt Apple's stock price. It could also possibly even impact people planning on waiting to purchase a G5. If they perceive, because of the Wintel media machine, that the wait might be months longer than originally anticipated, then they just may buy Wintel machines instead. Such is the power of FUD.

My sentiments exactly. Steve knew about this well in advance, and is anticipating as such. Don't you think that this is why the G5s have been delayed till August/Sept? Seems reason enough for me.

[Brother Mugga]

Originally posted by iconmaster
Watch it again; Steve just got a little tongue-tied. He was about to say, "world's most advanced chip fabrication in the world," realized that would be utterly redundant, and used the power of synonymy to quickly come up with something only mostly redundant. :)

Hmmmmmm. I dunno. He clearly says something (lips moving) that they've edited. Anybody a lip-reader? It doesn't look to me as though he just says "the worlds most advanced chip fabrication [in the world (edited)] on the planet" (which would actually be a *double* tautology, if you think about it).

And besides, how egotistical IS he if actually told the engineers to go back and edit out the second "in the world"?

Unless I've misunderstood what you're saying?

Maybe he said "in the world in [oh my god! - it's just occurred to me I'm not yet the richest man] on the planet!"?

Yeah...time for another raise, methinks.

Brother Mugga

[yosoyjay]

Originally posted by tychay

300 mm is the future, and, unlike their competition, IBM has a plant today.


Actually, Intel has had a 300mm fab up (Fab 11X) in Rio Rancho, New Mexico since November of last year.

They then donated all the old manufacturing stuff (6") to the University of New Mexico.

[Xapplimatic]

Originally posted by Macrumors
...It's unclear if this affects PowerPC 970 production.

We wouldn't actually be so ignorant as to think that the G5 is the ONLY chip IBM is going to make at their flagship chip plant in Fishkill would we?? ;) Didn't say what KIND of production problems.. Might be contamination.. might be suppliers.. could be lots of things that don't affect G5s in specific.

[tychay]

Originally posted by yosoyjay
Actually, Intel has had a 300mm fab up (Fab 11X) in Rio Rancho, New Mexico since November of last year.

They then donated all the old manufacturing stuff (6") to the University of New Mexico.

Well if you to be technical, Intel has a 300mm (12") fab up in the Pacific Northwest long before New Mexico opened. It is a research fab, I believe which is designing what 11x is supposed to be producing. And 11X launched in October, not November. ;)

I guess I wasn't clear. You are technically right but it doesn't make me wrong.

To bring everyone up to speed: There are three factors that affect chip price:
number of layers: more layers = less yield = more costly
size of the platter: larger platter = more chips per platter = less costly
size of process:smaller process = smaller chip = more yield = less costly. One issue is that the chips need to be resdesigned to take full advantage of a smaller process and that there are a lot of bugs in the latest technologies. Currently the "latest" is 12" platters (300mm) as opposed to 8" and 90nm process (.09micron) as opposed to 130nm (.13micron). We are comparing IBM Fishkill (which recently stole nVidia's business away from a Taiwanese fab) which is at 130nm (and 90nm) on 12" platters to Motorola (160nm on 8"), AMD (130nm on 8" when they aren't using Fishkill) and Intel (130nm on 8").

As mentioned by the poster, Intel's Fab11X in New Mexico is indeed 130nm on 300mm (12") to transition into 90nm sometime this year. It was commissioned expansion in 2000 for a $2 billion (not including the transition). (I had no idea about the donation to the University of Arizona... it seems a little strange because 11X is an expansion and 11 was too new for 6" and still operational--the combination makes it the largest factory in the world. Are you sure this equipment didn't come from some other fab they closed down?)

Intel's Chandler, Arizona is going to be upgraded also starting this year and will probably open in 2005. There was another 12" fab on the table in Ireland but it was cancelled.

Unlike others, Intel has a lot of fabs so their 8" are still cranking out while their they work out kinks and upgrade others to 12". Read the IBM article again and you see why Intel is so behind schedule on 12"--there is an overcapacity problem--why worry about 12" when your 8" meets demands and bring profit?

However, despite the fact that it has launched last October, 11X is the fab that has been unsuccessful at getting any chips out the door. As evidence:
the glaring lack of news of anything about this fab since it opened
all Pentium 4 chips are currently manufactured on 8" wafers when the intended use of the 11X is to produce the Pentium 4--look it up in any New Mexico press when in October.


Note: I could be wrong here because I found an article saying that the 3.06 GHz Pentium 4 is produced at 11X, but it was written by a reporter who knows nothing about the industry and was probably confused--no trade publication has claimed 11X is actually producing anything... you read a lot of carefully worded "will someday produce the Pentium 4" or carefully mention Fab11 and not Fab11X. I am open to anyone proving me wrong here! In my previous post I implied that the Pentium 4 is a simple chip relative to the G5 or Athlon. Yield is already "good enough" out of Intels 8" fabs.

And yet, as I mentioned we have many PC enthusiast sites claiming that Prescott (Intel's 3.2 GHz Pentium "5") to debut this fall at 90nm on 300mm wafers (and no doubt 9+1 layers) on the cheap.

Common sense says, if this is really the case, Intel is seriously behind schedule. Odds are Prescott will debut at 130nm on 8" wafers and later will get shrunk down. Why bother doing anything else? Consider, if you are Intel:
You have an overcapacity problem, you currently make more than enough to match demand.
You have yield issues at your 12" fab.
You have a serious lack of competition: AMD Opteron? That competes with the Pentium 4 XEON at best and has no tier 1 supplier; AMD Athlon 64? Puh-leez, maybe when Microsoft releases an OS for it, but not until then; IBM PPC970/G5? I love Macs but what is their market share again? When was the last Windows for PPC... NT4?
Seems to me like Intel is resting on their laurels as they always have when they've taken won (like just before the Athlon went 1Ghz, like just before DDR overtook RDRAM, like just before Transmeta Crusoe came out).

My original point is that IBM's situation is not unusual or unexpected. Relative to Intel, AMD, and Motorola, Fishkill is ahead of all the others--at worst, their problem is the same problem the rest of the industry faces: underutilization.

Any other conclusion is just a misreading of the reports or the current silicon reality. Any PC enthusiast who tells you differently is a troll.

The suggestion that Motorola is sitting pretty makes me wonder how their supposed upgrade to 12" is going and why the G4 is still at 160nm.

The fact that some of AMD's R&D is in Fishkill and AMD enthusiasts sites hope that IBM buys out AMD out speaks volumes as to their position.

As for Intel? Chipzilla seems to do best when they're back is against the wall, so I expect this situation to change if the 64-bit PC thing (G5 drives sales of Athlon 64) pans out. They didn't get that moniker for being a bunch of pushovers. ;-)

[myrdred23]

The quote from steve is:

"IBM offers the most advanced chip design and manufacturing expertise on the planet"

I do not know why it has been censored, but even IBM themselves quoted him saying this on their webpage at:

http://www-3.ibm.com/chips/splash/ppc970/foundry.html

A little google search can do wonders!