Re: "You expect a 933Mhz based upon process tweaks only. You are ignoring the potential impact of design tweaks. The Coppermine we see, was obviously rushed into manufacturing for market reasons.It was a complete re-layout.It stands to reason that speed path tweaks generate an extra 10% speed. So breaking the 1GHz barrier with Coppermine on 0.18micron lithography begins to sound realistic."
I don't consider Intel's .18um technology a process "tweak" Their .25um technology was very ordinary. This .18um
process is EXTRAordinary. My opinion is based on what I can see. Historically, what Intel publishes at IEDM shows up in product some time down the road. I don't know what Intel has done from a design point of view which will significantly improve Fmax. I seriously doubt (at this stage of the design) there are speed limiting paths which when "tweaked" will yield 10% improvements. In fact, I believe (when all is said and done), if Coppermine exceeds 1GHz, it will almost ENTIRELY based on .18um process enhancements and NOT design. Maybe, Intel has made extensive use of low Vt devices for critical paths in this .18um design that was not possible in .25um because of power constraints. This could add another 6%-7% which could get them to 1GHz. However, even this I would consider more of a process enhancement since it requires two additional device designs along with accompanying models. So, maybe they do reach 1GHz with Coppermine. (by 1GHz, I mean 1GHz excluding appropriate guard band and in sufficient quantities for a typical Intel release) If they do, that just makes my prediction wrong. As an outsider with no access to any inside Intel information, I can live with that. Actually, as an Intel long, I would be quite pleased.
THE WATSONYOUTH |
