Integration is really the key challenge and it's unclear that Apple can or wants to do that. Intel's recent Digital RF on CMOS Atom SOC ("Rosepoint") is the first of its kind; no one else has demonstrated that level of integration on a modern process tech node.
The node advantage is expected to reduce going forward if not go away. Smaller technology shrinks (22nm-14nm) are not providing the dramatic reduction in power or increase in frequency anymore. So if the Apple design team can get close to intel performance with better integration of other functions, the management would take that and get a huge improvement in margins.
Remember that Apple is already pretty much vertically integrated already.
The "architecture" (x86 vs ARM) is irrelevant. x86 is actually a disadvantage, but a minor one, as all Intel chips since the Pentium immediately translate x86 into an ARM like RISC ISA.
What's left is microarchitecture and circuit technology. Of these, Intel currently leads in absolute performance per core, while ARM leads in performance per watt, which translates into throughput performance.
Blog Doing Math in FPGAs Tom Burke 24 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...