A very big deal was made out of notched gates and what they made possible for Intel on .18. I posted a link to an article discussing that it was different from their existing, well established, .18 process.
I've read the articles on it. I've also got an old semiconductor international article which talks about the notched poly process. Given these articles come from the end of 1999 (one from November, one from December) and given that the coppermine was released in September of that year... it looks like the 100nm notched gate went in fairly early, rather than to "produce the 1GHz P3 and 1.7GHz P4" as you conclude. There is one interesting piece of data, though. The only cross-section SEM images I've seen of Intel's 0.18 micron process have notched poly gates, which still leads me to believe that they'd been playing with the notch all along. I would actually be very surprised if any coppermines were produced without the notch... of course, we'd need someone to go strip-back one of the initial release P3E's to find that out... I also don't think any P4's were produced sans notch.
From the .25 generation to the .18 generation took gates from .20 to .10, while moving from the .18 generation to the .13 generation will only take gates from .10 to .07 - a 30% decrease instead of a 50% decrease.
Regardless of whether or not Intel moved to 100nm out of the gate with 0.18 micron or made a brief stop at 130nm, your point is well taken. It will be interesting to see if Intel can make up for the somewhat reduced gate scaling through better transistor design or not... From a process design side, Intel has always had some of the best transistors (in terms of drive current, Ioff, etc.), if not the best transistors in the industry (certainly at release, Intel's 0.18 micron notched transistor was the tops in the industry).
As far as the article with the nice chart you posted... it's a nice article, but as Intel isn't the source, I must take it with a grain of salt rather than as the gospel truth. If Intel actually released microprocessors with a 130nm gate length at the 0.18 micron node, I'm willing to bet that they were the mobile Pentium II's which were released 3-4 months before the coppermine. |
