MHz is truly a poor metric for performance.
Very true when comparing different architectures. However in this case it is pretty obvious that Intel lowered the clock rate to hit a wattage figure. Current testing of existing hardware indicate very bad performance for this chip. Especially if you expect laptop performance to move forwards.
Now granted this indication isn't the result of a broad collection systems so we don't know how specific implementations can change the performance profile. However we do know that 4.5 watts is the low end on power draw and that the chip can run up to 12 watts.
It is very hard to do 12 watts in a passively cooled laptop so the indications are that the chip will be throttled a lot in a fanless design. In the end cutting back performance to 800 MHZ is a problem no matter how improved your architecture.
As I said earlier, this can be shown by the fact that an Intel CPU completely destroys an ARM-based CPU running at the same frequency. CPU architecture is why 800 MHz is not the bottleneck you'd think it would be.
Architecture is important but so is the rate that the architecture can execute. Intels architectural advancements haven't been all that great of late so we may be lucky to get 5% from Broadwell for the CPU. On a machine that constantly throttles the CPU below what Apples Haswell machines ran at that will be a problem.
As for your slagging of ARM I'd love to see head to head comparisons of this chip and Apples A8X. Sure the Intel hardware has the potential for bursts of higher performance but I'm not convinced that it could keep up with A8X when implemented in the same form factor.
Turbo-boost makes a tremendous difference for 99% of users, and for the rest (those who run CPU-heavy tasks), turbo-boost typically kicks-in anyway, assuming the CPU is operating below its TDP.
Actually turbo boost will kick in when it can run the chip hotter than TDP. Beyond that you have to remember that these machines are reputed to be driving retina screens which means the GPU is a more significant part of the power profile. 4.5 watts is trivial once the GPU is under any demand at all.
The MHz metric is absolutely irrelevant, based on how modern Intel CPUs work.
Actually you are absolutely wrong here. The ability to sustain clock rate is Very important for Intel hardware. Throttling sucks! If the platform can not allow the chip to run at a sustained high clock rate performance will suck. This is what we are seeing in current benchmarking.
Now one thing I'm confident about is that Apple can do better thermal engineering than just about anybody in the industry. The problem is this looks like more of a 12 watt chip than a 4.5 watt chip. That makes for a very difficult fanless implementation if you don't want to compromise performance by throttling.
So I can see Apple delivering a machine built on this chip but I can't see them exceeding MBA performance levels fanless. In fact I see far worst performance. This especially the case when you can update the MBA's to Broadwell with even higher performing chips.
I really think people need to be guarded in their expectations here. Apple can only work so much magic with what Intel gives them.
