So you said new process tech doesn't make it half the power, which is true, so do you think it's possible for Intel to reach almost half the power with at least 20% more frequency?
Intel is starting today with a process that's optimized for high performance PC processors, NOT one that's optimized for best perf/watt at smart phone and tablet TDP points. And when you stretch a process or a design that was aimed for higher power points into lower power points, it's obviously going to lag one that was specifically targeted at a lower power point.
Even Intel's 45nm LP process that was used for Lincroft is a stretched version of the 45nm HP process that they used for their PC product line. The 32nm SOC process is really the first one they optimized for SOC designs.
Therefore, I would expect Intel to have larger generational leaps as they "catch up" to what ARM has already been taking advantage of since the beginning. Intel's still in the space of low hanging fruit, while ARM has already wrought out all of the benefits of optimizing the design, the architecture, and the process.
There's no "magic" here. Both ARM and Intel have to architect their products to hit a given design point, and both are very good at highly optimized designs. However, ARM has a lead, since that's been their core business for many years, while Intel is coming from a very different baseline.
Lincroft is a good improvement over the previous generation, because Intel put all the effort into reducing the power of their uncore. The previous generation basically reused an existing chipset design, but Lincroft rebuilt the uncore from scratch, which is why you see the drastic improvement. The next step is to put it on an optimized 32nm SOC process with both core and uncore optimized for that process node.
I think we'll see another step function as Intel moves to 32nm, while existing ARM A9 designs will see less of a benefit when moving from 40nm to 28nm. I'm a bit more worried about the upcoming A15 core, but we have yet to see Intel's next Atom architecture tock, which hopefully arives around the same time.