Quote Originally Posted by savantu View Post
The whole problem started with GF 32nm process. GF and the IBM club use a gate first approach which is more problematic to control than the gate-last approach used by Intel. Gate-last requires more restrictive design rules, but if you can live with that, it is doable as Intel proved and offers similar performance if not better than gate-first.
Again, that's basically investor board fud as usual.

Globalfoundries provides measurement data showing that "Gate first"
is working fine, with low Vt variation, low Idsat variation and is showing
healthy Weibull slopes under accelerated degradation tests comparable
to Poly-SION.

http://issuu.com/globalfoundries/doc...owFlipBtn=true


Regards, Hans