Gabriele Svelto (gabriele.svelto.delete@this.gmail.com) on October 26, 2016 2:23 am wrote:
> David Hess (davidwhess.delete@this.gmail.com) on October 25, 2016 9:59 pm wrote:
> > The alternative is that Apple designed a 4 GHz processor
> > but only operates it at 2 GHz do to power constraints.
> > They could do that but a dedicated 2 GHz design would be higher performance or lower power or both.
>
> It would also be simpler to design. Targeting ~4GHz right from the start means that your
> timing constraints are very tight and your critical paths significantly shorter.
>
> Now, considering that Apple does not bin its CPUs we might assume that they have some
> headroom so that lower quality dies can still achieve their desired clock-speed; they
> might actually be targeting something above 2.35GHz but probably not by much.


The A9 also required headroom for the higher clock speed A9X. Granted that's a separate die, but it is hard to imagine they did a complete re-layout with different FO4 targets in the A9X's cores (which would presumably show up in benchmarking somewhere as slightly different latencies for an instruction or two) so the A9 had to be capable of being clocked to the A9X rate (plus headroom) which suggests it was designed for something closer to 3 GHz.

If rumors are true that the iPad refresh is next spring to allow using a 10nm A10X, then they wouldn't need that extra iPad headroom in the A10, but it is clocked faster than A9X despite being on essentially the same process so probably had a similar target.

But what is Intel's target? Just because their highest bin is 4 GHz doesn't mean that's their target. They need headroom above their max turbo frequency as well, and air-based overclocking results suggest their FO4 target is closer to 5 GHz.