Star IBM Transitioning to 22nm Chip Fabrication
Linked by David Adams on Mon 29th Sep 2008 02:52 UTC, submitted by Punktyras
Benchmarks Intel has already shipped 45nm processors, while AMD is trying to get its 45nm to market. TI is claiming big 45nm performance too. Meanwhile, IBM believes it can lead the market to 32 and then 22nm chip production: "At 22nm, the existing techniques for microprocessor manufacturing useful at the 45nm and 32nm nodes will become obsolete. While currently lithography techniques cannot produce 22nm circuitry, IBM has developed a new approach called Computational Scaling, which will allow for this tiny-scale production. The new technique uses advanced mathematical computation to adjust the shape of the masks and illuminating source during etching."
E-mail Print r 3   6 Comment(s) http://osne.ws/fox
Thread beginning with comment 331856
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE: So?
by Nicholas Blachford on Mon 29th Sep 2008 16:28 UTC in reply to "So?"
Nicholas Blachford
Member since:
2005-07-06

So... IBM managed to stretch current lithography a little longer... in a lab.. and with RAM apparently...

Hmm.. show me a current CPU scaled to 22nm or shut up!


Starting with SRAM is quite normal for a new process.

They don't even claim to be ahead of anyone ( though the article eludes to such ) - because they are probably merely on par w/ Intel.

What would have been interesting:

1. I.B.M. doing the above and licensing it to AMD

AMD will almost certainly have this technology available to them.

2. I.B.M. had a breakthrough and skipped 32nm

Not even Intel could do that.

3. I.B.M. saying they would be releasing 22nm
....products to mass production first.

It's hard to predict who's first, e.g. the first 45nm part was a BluRay chip.

4. I.B.M. creating a 22nm at-home CPU generator,
....come up with a design and build it!

Speaking on point 4: Has anyone ever attempted to
make this possible? It seems it SHOULD be possible
to create a single integrated device which would build chips from input data and materials alone. It would certainly be a good way to get an early production chip made.

Sure it would cost a fortune ;)


With that much automation, the chips you produce are not likely to be very good.


Also, let us not forget that I.B.M. has a poor record when it comes time to deliver. The mythical SOI on which AMD spent plenty of money they didn't have, was a major PITA for AMD, costing them dearly. SOI was supposed to end leaking transistors,


Actually AMD already had SOI, except it didn't work very well, that's why they went to IBM. Who gave them technology that worked.

but AMD had more side-effects from all of the unknowns I.B.M. had never uncovered ( because they never seriously tested the process ) than they would have had just sticking to the same process for another generation and focusing on WHY transistors leak, rather than how to insulate them better.


If they didn't test their process how come it's used by so many big players in the IC industry?

Reply Parent Bookmark Score: 3