Roughly ~1.5 process gaps, since AMD just released Llano on TSMC’s 40nm, and is about to release Bulldozer on GF’s 32nm. (If we take the major nodes at 45nm, 32nm, 22nm, from Intel’s tick-tock cadence. Although note that the foundries have their own proper definition of major- and half-node processes.) TSMC and GF are already working on their 20nm/22nm process, but they haven’t any chips to show for it yet, judging by the lack of news on that front. No doubt each fab is already hard at work researching the next few nodes after that, but I won’t speculate further without updated official announcements.
The new gate transistor isn't the
actual exciting news though; after all, who really cared about the hi-k process Intel used on 45nm? What’s scary is that Intel already has
working chips with these new transistors, as demonstrated in
this video (assuming the video description is accurate).
That’s right; Intel pretty much has LGA2011 and Ivy Bridge ready for release (or close to it), and they’re just delaying them to avoid cannibalising their Lynnfield/Sandy-Bridge offerings. Of course, it’s not like they can snap their fingers and say “let’s roll Ivy Bridge out now,” but I doubt it’ll take a year for them to carry out validation, tape-out and mass production (Ivy Bridge is tentatively slated for Q2 2012).
Even with Bulldozer’s release imminent, it seems AMD is still on shaky ground; they certainly won’t have much leeway to rest on the laurels of their Bulldozer microarch, even if it manages to match Sandy Bridge performance. Not to mention that AMD is entirely reliant on GF/TSMC’s progress with the 22nm process. Next year might be another interesting year as well, when Komodo goes up against Ivy Bridge (probably not as interesting as Sandy Bridge vs Bulldozer, since next year will be the year of node-shrinks and not new architectures).
For a layman explanation of this tri-gate transistor, take a look at
Anandtech’s article.
