That has nothing to do with the architecture -- Intel's superior process scaling just means that Intel is great at process scaling, but they could do equally well shrinking ARM chips if they still made them. Not that this matters, in the end -- Intel doesn't make ARMs, and TI doesn't make Atom SoCs, so the question is how their products actually compare, not how they ended up that way.

And ARM already wins big in performance/watt, regardless of whether it's "a step behind". (Can you imagine what a 20nm ARM would do? )

Except that no corporation wants desktop software running on mobiles, because the UX is horrible with unoptimized software on touchscreens, QWERTY thumbpads, and the like. They just don't see the added functionality as a selling point. So as nice as it might be to power users, it doesn't enter into any of the decision-making that matters (barring some Pandora-like team building a hobbyist mobile based on x86).

Moreover, I've only ever had occasion to use a handful of x86 binary-only software on Linux -- Skype, Flash player, and Google Earth are the only apps that come to mind, along with a few closed hardware drivers (which are platform-dependent anyway). Right now my main desktop is a pure x86_64 system, with no 32-bit compatibility layer, and I'm doing fine with none of those. So I don't really see the big advantage for cross-platform binaries -- source-compatibility is enough.

Not to answer for sjgadsby, but his statements are objectively based. Everyone who knows anything about CPU architecture (including CISC fans -- x86 is quite possibly the ugliest CISC in the world) agrees that x86 is a huge bag of fail, but since backwards compatibility rules the market, it keeps going.

No, that's the question to argue endlessly about. We will always have RISC and CISC chips, they will always be reasonably competitive, and the argument will never be settled.

Certainly, if the RISC/CISC war were to be won (by RISC, obviously (me? biased? )), that would settle the x86/ARM battle for high-powered mobiles, and indeed everywhere else. But since that won't happen, x86/ARM will happen on its own -- since nobody codes assembler, and compilers optimize code beyond recognition anyway, ideology will prove subservient to practical effects (which is faster? which uses less power? which is compatible with relevant binaries?)

Then your reading is woefully inadequate. x86 is one specific CISC family, and it has a head start for certain markets (starting with desktops (i.e. the IBM PC) and scaling upwards to servers), sure. It has almost no share in the handheld market -- an incursion here or there (HP LX100 anyone?), including Intel's current push, but as yet no long-term presence. Their success on the desktop was directly tied to the success of the IBM PC architecture, which in turn succeeded not due to any particular technical win, but because of the PC clones -- other systems using different CPUs never got that level of commoditization, so they got squeezed out.

But that says nothing to the overall CISC/RISC war, and RISC (and more specifically and relevantly, ARM) has an equally big head start in the handheld market. Which just might say more about which one will win the handheld market.

The last bit is just plain wrong -- RISC are simpler, therefore (all else being equal) cheaper to manufacture.