LOL, another victim of corporate marketing!
ARMs are cheaper because they're simply cheaper to produce. Given the same process ARMs take less space because they are a simpler and more efficient design. That means you can put more of them on the same wafer and a greater percentage will work.
If you were so inclined you could also adapt the ARM to have a similar performance profile (basically improving single thread performance at the cost of multicore performance) and the ARM would still be cheaper while also performing better in terms of power and computing speed for the same price.
Intel slavishly sticks to the x86 ISA although it's clearly a disadvantage because it gives them an effective monopoly.
Thanks for calling me a corporate marketing victim, but unfortunately I am not. I guess the
underdog's fanboi is a better title for me. The thing is you didn't seem making a sound point either. What I was taking about is that
today's ARM chips being cheap and power efficient is because they are intended to go along this way. There is nothing wrong about this, as it is a super generic statement. Microcontrollers are cheap because they are intended to be super low performance (yet adequate for its job) and super cheap with thin margins. Anything really wrong?
What I think people went wrong is they placed the same assumption on these chips (and microarchitectures sometimes) for a lower performance, lower price point processor market on a broad series of market segments above, which isn't really the same living habitat after all.
Let's face it. The more complex the design is, the more insignificant the "x86 overhead" is, or in other words "negligible" as the cost proportional to the entire scale is significantly smaller. It is just the cost of decoding the instructions and supporting corner-cased instructions after all if we are speaking of the core itself. For the rest, an add is an add, and a multiply is a multiply.
Let's also not forget CPU nowadays attributes to only a small portion of die layout in a complex SOC, while the performance of the cache hierarchy, the platform power management and sometimes your sets of I/O and accelerators matter more. Yes, ARM is a more elegant instruction set and can be implemented in a cleaner way undoubtedly, but does it matter the most when you weight other stuff? Lots of people may disagree.
Yeah, Intel sticks to it probably because it wants to hold
monopoly in its land. It is the nature of business. Who would leave an advantage on the table unused? The PC and the server markets are reluctant to x86, this is the truth, just like how the mobile market is reluctant to ARM. This is a historical factor, and less a technical factor. If you have time, please also help ask AMD why they would still develop a new x86 core from clean sheet, targeting 2016 launch, if x86 really sucks and has no business value.
I would not deny that Apple may have the ability to take on Intel, given the great design of its Cyclone micro architecture. If they are able to push it into higher clocks and pioneer the invasion into the heterogeneous era, it would probably be something nice after all. But Mac is a product stack with premium price points, and it is competing with a sea of Intel-inside (and rarely AMD-inside today...) PCs, unlike the mobile market that the major competitors are all on the same ground (as Intel is nearly inexistent). It IMO needs something that justifies the price points, while splitting the ecosystem shall be avoided if possible. Unless Apple intends to make breakthrough in areas other than the CPU business (which AMD is attempting via HSA...), making use of its vertical integration model, I wouldn't hold my breathe tight on a Mac transition to ARM. So just don't let the hype train lose its brake, please.
