@Scary,

That's the reason why I contacted POET to ask about the fact that in the past they have reported 2 frequencies. One has been for the always-reported shrinking n-channel which was always around 40Ghz and has been projected to be as fast as 350Ghz. The other is for the p-channel, which historically has been slow (GaAs is sometimes reported to have an even slower hole mobility than Si) and is only 3Ghz. AFAIK, POET has never attempted to shrink the p-channel and it has therefore remained at the same freq for a 1um size.

To be honest from the answer I got I'm still not sure which frequency is the one to watch. It's possible that the p-channel speed limits the complementary transistors to the same as the p-channel (currently 3Ghz), and it's also possible that the n-channel is used at different times than the p-channel so it's an average of some kind. It's even possible that the n-channel "leads" while the p-channel follows, meaning the n-channel is the one to watch. I'm not sure.

But based on other parts of my response I think the point is moot. What makes POET special is that Taylor has either found a way to make fast holes, or has negated the need for fast holes. In other words, the p-channel will be fast enough when reduced to 100nm to go where no transistor has gone before. That much has been confirmed.

Let's say for the sake of argument that the p-channel goes to 12GHz, and we learn that the p-channel is a limiting factor - meaning the complementary logic only goes to 12GHz, altogether. That's still over 3 times faster than Si and there's still room to shrink, and even room to work with GaAs (e.g., straining) in the same way as they have with Si over time to increase the hole mobility.

Personally, I don't worry about these things because POET appears to give device and chip makers lots of room to shrink for speed before they have to get creative.

And this ignores the optical side of POET which I think give POET almost infinite speed possibilities.