TREZA (no.delete@this.ema.il) on September 27, 2013 1:50 pm wrote:
> > Decades ago (MIPS R2000 vs 386 say) RISC had a distinct advantage in terms
> > of results for design effort. For any market that needs more than a low frequency
> > in-order scalar processor the difference drops down to noise levels today as
> > performance requirements rise to even just current smart phone levels.
> >
> > Saying an A7 core was vastly easier for Apple to design than the Bay Trail core
> > was for Intel to design is just sheer nonsense. High performance x86 is a long
> > solved problem (at Intel if nowhere else) whose solutions continuously improve
> > based on accumulated knowledge. The portion of a Baytrail SoC dedicated to
> > x86 decoding that the A7 SoC lacks is ridiculously infinitesimal in terms of the
> > over all device with multiprocessor cache coherency, OOOE execution cores,
> > dynamic power management, and all of the non-CPU integrated functionality.
>
> I think that doing the A7 was vastly simpler for Apple than doing a Bay Trail or a Jaguar.
>
> Starting an x86 from scratch is certainly far more complex than an ARM. Iterations and refinements of the
> design are probably also a bit more complex. Even if the decoder is a small part of a chip, keeping compatibilty
> with DOS and thoroughly validating it has a cost. There are also many invisible tricks (like predecoding
> in the cache) which add up complexity and must be adapted and opimised for each iteration .(unlike, for
> example, the x87 cosinus microcode which was probably not changed much for a decade).

High performance RISC processor have used pre-decode information and it's certainly possible that Apple uses it for their processor. In comparison all of Intels processor doesn't use that "trick" while AMD currently uses a simplified form only in the Bulldozer derivatives while the lower power Bobcat/Jaguar doesn't use it. In short: this isn't a complexity difference.

The most complexity in x86 isn't really the implementation in itself where the complex parts have been managed through the years but in the verification that all parts together conforms to the x86 ISA. But is it so much worse than for ARM where a modern core have to support ARM, Thumb, AARCH64 encodings plus other cruft added through the years?

> In a sense, maybe ARM is too easy. There are many smallish teams doing
> competing simple designs which cannot fight the Intel behemoth.
>
> IMHO, the ARM side will need some "consolidation" if they want to seriously challenge
> Intel. The role of ARM itself may evolve as they cannot be a "standardisation
> organisation" deferring the work of creating high end CPUs to others.

Like that'll happen. ARM is a for profit corporation.