You should be careful not to oversimplify when throwing in technology nodes, these values only describe the ability of the lithography technique used and only have an indirect connection to the actual circuit elements that they form - and a rather small influence on the overall power efficiency by themselves. There are tons of other aspects aside from the basic architectural design that influence such things - even the manufacturing process itself has a large impact on the efficiency, because it has a huge influence on the electrical properties of the transistors. We have long reached the point where whole transistors can't shrink any further, nowadays people are looking for ways to pack and construct them better instead.
I think we must simplify so all common people can understand (not all people knows about this questions). And we must simplify so we don't write too much big text.
¿Small influence? Not only this Atom x5 has 4 times smaller transistors, so even it uses <strong>Tri-gate process</strong> -Intel is very good in silicon manufacturing process/techniques- . Obviously I don't expect those refinements in much old TI OMAP5. Of course this SOC can have much more transistors (for example a big cache has a lot of transistors).
For me it is clear: although I don't like x86 architecture this Atom x5 is a much powerful/modern/advanced product than TI OMAP5.
About heat: under the same source code, same SO, same cores, I bet Atom will be more efficient. Under full load it may be other question as Atom can be bigger (smaller and more advanced transistors, but much bigger transistor count for 4 cores and, in comparison, big cache).
Of course tests in real hard are needed to know how they perform in reality.
About "the point where whole transistors can't shrink any further": we are near, but they have at least 2 more size reductions.
PD: When I say I don't like x86, and I like ARM, I say it from a assembly programmer perspective, as I suffered x86
(when assembly was still used). It was a really bad design in its origin and even Intel didn't plan to continue with it, bit its other project (an OO processor was a market flop, and IBM saved Intel selecting 8088 for IBM PC).
