New IBM Plant Will Mass Produce .1 Micron Chips

Ruger writes "AP News is carrying this story about IBM opening a new plant in upstate New York. What's most interesting about the story is that IBM will be producing .1 Micron Chips rather than the usual .25 or .18 produced by Intel and other chip makers, or .13 Micron chips they currently make for their PowerPC chips."

Comments

Returning to the fold.

[GodInHell]

'Bout time IBM got back into upstate NY.

I remember when I was just leaving the area, the last of the local plants finally scaled back to just a matinance group, the whole area died. IBM was the heart and soul of quite a few towns in New York, and they didn't do very well when it left.

-GiH

Intel at .13?

[timwhit]

Isn't the northwood P4 produced at .13 micron? And the AMD Throughbred is also at .13? The header says that other chip manufacturers produce chips at .25 or .18 when this simply isn't true.

Re:Intel at .13?

[megalomang]

No, it is neither. The "micron" dimension associated with a particular fabrication technology is the average width of a transistor. The smaller transistor, coupled with better design resulting in redundant circuitry and better fabrication processes resulting in fewer faults, allows more transistors to fit on a chip. causing Moore's law to continue to tick forward.

There is another dimension generally provided, which is the wafer size. Recently, Intel became the first to start high-volume production of a 300mm (aka 12") wafer size, versus the previous 200mm (8") wafers that most of the industry still uses. Combined with the .13 micron process that most new P4s are fabricated with, this results in an extremely high die count (number of chips that can be masked onto a single wafer) which is of course offset by the enormous Northwood die size!

The entire digital chip is generally masked using the same process, including the core, L1 memories, L2 memories, and sometimes (e.g. Itanium) even the enormous L3 memories.

Re:Supertiny G4's

[ivan256]

IBM does not make G4s. They don't have a license for Altivec. They already make quite speedy G3s, but you don't see them in consumer products that are marketed based on Mhz.

Re:To the naysayers...

[jmv]

What are the current guesses on how much smaller we can get?

Usually, the current guesses are about twice smaller than current technilogy :-)

Seriously, there are two (in fact more) limits: there's the smallest transistor possible that works correctly and there's the smallest features size we can mass-produce with reasonnable (well, it's already unreasonnable...) cost.

Right now, the most limiting factor is the second. The visible light is already much too big (wavelength) for lithography so they're using (AFAIK) ultra-violet, but one of the problems is that the smaller the wavelength, the harder it is to find a transparent material at that wavelength (glass doesn't work past a certain wavelength).

I'll be damned...

[powerlinekid]

"Look ma, I'm on slashdot"... well not exactly but I actually work there. I program the testing systems so that the engineers can run test on the wafers. The ribbon cutting was pretty cool, CEO Sam was here and so was George Pataki. Nothing like sitting in the conourse for lunch and seeing a massive black helicopter fly overhead. Got a free hat out of it... to be entirely honest this is a big deal but business here really isn't going to change. We've been porting our testing system from the old design to the 300mm for awhile now and theres been alot of restructuring of the departments such as moving people to the new 300mm ones etc.

I was wrong, Intel will be the first .09

[megalomang]

It says here that Intel's Fab 24 is now slated to support a .09um/300mm process by end of 2003. Although no dates were indicated for IBM, they may indeed beat Intel to 0.1um. So why is IBM going for .1um when Intel is going to .09um?

http://www.siliconstrategies.com/story/OEG20020118 S0081

Re:Not only that, but Intel will be the first at .

[max cohen]

Hah, competitors hardly need to "catch up." Seimconductor companies almost never want to be the first to build a fab supporting the largest wafer size, unless your design a chip that no one is buying and have to dedicate 420+ mm2 per die just to get decent performance. ;) Being first sounds good on paper, but it also means you get to debug all of the new tools from vendors. If you thought beta software builds were costly, try running your expensive wafers though a $4M+ Endura from Applied Materials and having the robot shatter them. Not only have you lost your test vehicles, you wasted expensive chemicals and have to clean up the vacuum chamber. Not fun or cheap by any means.

The running joke in the biz is that every company wants to be in second place in the race.

Additional Coverage

[leibnizme]

If you want to have more details about this fab, check out:

IBM's news

Yahoo Story

NY Times (free reg, blah)

IBM is an icon but ....

[Knightfall]

AMD and Intel are right there. Consider ...

http://www.amd.com/us-en/Processors/ProductInforma tion/0,,30_118_608,00.html

and

http://www.intel.com/ebusiness/products/roadmap.ht m

Both of these show .09 in the next year. I'm all for giving a company its due, but lets not leave the other players out. Maybe even, *gasp* go for a complete story.

"New" IBM Plant

[wilsonjd]

Just for the record, I work at IBM in East Fishkill, NY (but not in Chip manufacturing.) I do NOT speak for IBM.

IBM has been in East Fishkill since the early 60s, manufacturing chips and packaging (MCMs, etc.) for mainframes. IBM employs over 10,000 people in Poughkeepsie (10 miles away) and East Fishkill. The "New" plant is a new chip fabrication line in an old building (building 323) that used to be used to make bipolar chips for mainframes. They have been working on this new plant for over two years, and it is already producing sample chips. Normal production is scheduled to begin in the fourth quarter (btw, the current production from this plant is 0.13 micron, but in the future, it will move to sub 0.1 micron processes.)

IBM is using this plant as a high-end foundry. In other words, customers will design high performance chips that will be manufactured here. They are already working with some high-volume customers (Nintendo, Sony, etc.) Customers will also include IBM chip designers (mostly IBM servers.)

Oh, and on the whole upstate, downstate issue: People who live in upstate New York consider us downstate. People who live in downstate New York consider us upstate.

And, as Gov. Pataki said yesterday, the Hudson Valley is much nicer than Silicon Valley. We have trees.

Re:To the naysayers...

[imnoteddy]

There is a reasonably well researched report published each year called "International Technology Roadmap for Semiconductors", homepage is here.

The executive summary of the 2001 edition predicts that in 2016 the drawn gate length for microprocessors will be 13 nanometers (0.013 microns).

Now that we're on the verge of 0.1 micron transistors it is time to dump the microns unit and start using nanometers. The tables in the "International Technology Roadmap for Semiconductors" all use nanometers.