Originally Posted by shadowself
Originally Posted by rcfa
The 64-bit ARM chips are not significantly slower than a variety of x86 chips, but they use an order of magnitude less power.
Not significantly slower? You're kidding, right?
Yes, the best ARM chips are equivalent to (and in many cases better than) Intel's Atom line. (But that's NOT what runs the current Macs.) Yes, the best ARM chips are roughly equivalent to the bottom of the line i3 chips for most (NOT ALL) functions. The best ARM chips are almost equivalent to some i5 chips, which Apple uses for its slowest/least expensive Macs, for most (NOT ALL) functions. But no current ARM chip (not even the 64-bit variants like Apple's) is in the class of the i7 chips which Apple DOES USE in the MacBook Air, rMBP, and iMac.
For one, Apple wouldn't use last year's chips.
For two, they wouldn't use the clock speeds they are forced to use to deal with the small batteries in iPhones and iPods.
For three, they would use a different memory architecture
So plenty of room to make up that i5 to i7 gap.
By going massively parallel, Apple can make devices that are more powerful and use less energy at the same time.
Part of switching to x86 was improved performance per Watt.
The reason for going ARM on iOS devices was the lack of comparable x86 choices, and after many years, Intel is still lagging behind, while ARM's 64-bit cores have been rapidly catching up.
The reason for Apple going to an ARM architecture for the iPhone was power requirements, plain and simple. The ARM architecture was just fast enough. Apple has made great strides in its implementation of the ARM architecture, but it's not equivalent to Intel's mainstream chips and won't be for a few years. The ARM chips are not the same as Intel's main stream chips.
Intel is making great strides in its ultra low power chips. Intel's reduction in power requirements and processing capabilities in these chips. BayTrail is a significant step forward. Intel's motion forward in this type of chip is much, much faster than it is moving with the other end of the spectrum -- the Xeon chips or the Itanium chips.
Additionally, the I/O architecture of the ARM chips is not the same as Intel's mainstream chips with their support chips. Apple's iOS devices effectively have just four inputs: screen/buttons, lightning (one only), sound in, sound out. The typical Mac has many more -- even a MacBook Air. This will be an additional architecture change above and beyond that of massively parallel processing you mention above.
Apple used to make their own chip sets, and they have the know how to do it again.
Apple is an ARM architecture licensee, which means they are among hue he few who can alter the platform and develop their own implementations of CPU cores and a given ARM instruction set.
Apple was busy buying up chip design talent for the last few years, and they are no going to waste that investment. Macs at this point are such a small part of apple's business, that some have speculated Apple might stop making Macs altogether, so them lowering their costs and condensing development by a more unified platform for Mac and mobile devices makes perfectly sense.
Intel's CPU performance is stagnant, because the complex instruction set starts hitting the barriers of Physics when it comes to making smaller and more power efficient implementations. The ARM instruction set, particular the 64-bit instruction set, is essentially legacy free, streamlined and can be scaled in a variety of ways that Intel would have trouble doing with the x86 chips,
While maybe ARM may need a few more years to compete with the XEON chips in a MacPro, I see no issues powering the entire MBA and MacMini range with ARM with a desktop variant of several A8 chips.
Parallels can license/develop Rosetta-type technology and Windows compatibility won't be an issue, and who knows, maybe Apple and M$ have a deal for a full desktop-class ARM based windows, which would also help M$' tablet/laptop hybrids.
Again, you're kidding, right? It's not an issue if you can live with a system that runs at 75% of the original, native speed (and in some cases as slow as 25% of the original speed). Most people can't take that hit. Most can't. I tried it back in the day. It was acceptable for some tasks but not for most. That was why VirtualPC was a viable industry, but most can't.
Most people using Windows on a Mac are using it for occasional business apps: accounting software and stuff like that. The few people who use a Mac to run performance hungry Windows software are such a minority that Apple can and will ignore them in their decision making.
And, Apple's ARM variants are NOT code compatible with the rest of the industry's ARM chips. It is unlikely that Microsoft (or any other mainstream OS developer) will write a variant of their OS to specifically support Apple's version of the ARM chip.
Apples Chips are perfectly compatible, just because they are he first to jump on the 64-bit ARM architecture doesn't mean others won't do the same, particularly M$ with their iPad-Zune
Apple has the luxury of issuing a new version of its OSes (whether Mac OS or iOS) when it issues a new chip such as the A7. Apple has the ability to work on that new OS variant many months in advance so that the new variant of iOS is ready when the chip ships -- and still keep the capabilities quiet. Is Apple going to be able to get other OS vendors to work on their OSes in advance -- AND still keep them under wraps -- so that those OSes are ready when the chip ships? That is extremely unlikely.
The technology is there, so it's just a matter of marketing.
It would also be easy to use ARM in the consumer space and for server products, and use x86 or an x86/ARM combo for pro laptops and desktops. Apple already has multiple CPUs in the system so this is totally doable and allows Apple to go places where that competition can't easily follow.
The technology ISN'T there YET
. It will be in a few years (say 5-10), but not today. It isn't just a matter of marketing.
You may not have followed the performance increases in each subsequent A-series chip, and that was before Apple was on the much more scalable ARM 64-bit architecture. Apple clearly doesn't need 5-10 years to catch up with intel who has trouble bringing out a meaningful improvement every two years, that's maybe 20% faster than the last generation...
Apple does NOT have "multiple CPUs in the system" with regard to any single OS or general system architecture. There's Apple's own ARM based chips for iOS and Intel based chips for Mac OS. That's it. You don't even see x86 compatible chips (e.g., AMD's chips) in Apple's Macs. Why? because it would add more complexity to the OS to support other x86 based chips. Supporting ARM with Mac OS effectively would require two distinct and different versions of Mac OS. Not only building it, but supporting it, would be a huge increase on Apple's costs.
Apple has multiple CPUs: PMU, GPU, CPU, and probably others. The user generally doesn't consider them or write code for them (other than where Apple provides tools such as GCD, shaders, etc.) but that doesn't mean the CPUs aren't there or Mach kernel isn't capable of handling a mixed CPU-type, multi-CPU system.
Has all been done before.