Your entire reply is based on the fallacy that a high-end ARM would operate in the same way a low-end ARM does. Most of the energy and die space, etc. etc. in a CPU is dedicated to the caches. L1, L2, and sometimes L3. And they keep getting bigger even as the rest of the CPU gets smaller.

A low-end ARM doesn't worry about heat because it's "underpowered" (loosely speaking). But if you want to approach the performance of an i5, the natural side effects will be tons of heat, much, much worse battery life, and even then, only mediocre performance.

There's a lot more to a CPU than RISC vs CISC. Intel has put decades of research into chip design and various instruction-level optimizations that take place within the CPU. You're making a big mistake assuming ARM can just directly scale their smartphone/tablet CPUs to the multi-core behemoths that are today's Sandy Bridge and tomorrow's Ivy Bridge. It just doesn't work that way - look at AMD.

At the end of the day and at this point in time, ARM entering the high-performance desktop sector would be as poorly executed as Intel's foray into the low-energy CPU segment with Atom... except it would suck more.