anon (anon.delete@this.anon.com) on August 6, 2014 7:54 pm wrote:
> It's also not really disputed that at the very small scale, x86
> designs can't compete with simple ARM based microarchitectures.

This is *very* true. It was also evident in the 1980s, when the 386 was like 3X slower than the MIPS R3K despite having similar complexity (as measured by transistor count). The x86 ISO does increase the minimum gatecount for a pipelined mostly-hardware implementation (as opposed to a mostly microcode one), and if you happen to be looking for a core sized just below that gatecount then it will be utterly noncompetitive. Cores like A9 and R4 are basically right in that sweet spot, which is wny ARM has never faced significant x86 competition down there.

The most authoritative estimate that I've seen for the x86 penalty was from Bob Colwell (chief architect of P6 and Pentium-4). He put it at ~1.5 million transistors for a superscalar OoO machine like P6.

> I think they would find a way to make it happen, if the economics made sense. I think they
> would certainly be considering their options, and it's going to depend on how their mobile
> designs progress, design team capability, iOS and OSX markets, etc. My prediction is that
> anybody who says they certainly will or will not (without qualifying a short time frame)
> is wrong, because Apple probably have not ruled out either option at this point.

True, but meaningless in the sense that you could apply this statement with minimal rewording to any rational actor in any enterprise.