I think it's more likely Apple moves to ARM for Macs than moving from ARM to Intel chips for iPhones and iPads, like it was rumored a few days ago. ARM chips will be powerful enough for Apple notebooks in a couple of years, so I could definitely see an ARM chip in a future Macbook Air sometime in the next 2 years.
it might also be that Intel will gain some momentum with the recent introduction of their new 3D transistor approach leaving ARM behind even in regard to power consumption.
this could discourage apple from making a major processor switch again, especially with the high overall costs (compatibility etc.) a switch would introduce.
Intel's big breakthrough with the 3D chips is that they can continue Moore's Law at a steady pace, even when they can't shrink the transistors anymore. But it's not a breakthrough in performance. So in regards to that they are still behind ARM as always. Intel is still moving at Moore's Law's pace, not faster. For now at least, ARM chips seem to be moving faster, doubling in performance every 12 months.
I think Apple is planning to move to ARM once they have their (most likely) quad core A6 chip.
The tri-gate transistors have lower gate delays (they can switch more quickly) and can operate with lower voltages. How is that not a breakthrough in performance?
It's mostly a performance improvement along Moore's Law. Are you saying that now with the tri-gate Intel's chips will double in power every 12 months or every 6 months?
And if this is just a one time only "big" jump in performance, then even if it's 100% increased performance over sandy bridge, it's still not that much. ARM chips double in performance every 12 months. That's my issue with this. Because most people seem to think Ivy Bridge puts Intel so much closer to ARM, but it doesn't.
Let's say that without 3D, the jump from 32nm to 22nm would give them 30% increased energy efficiency. Now with the 3D it gives them 50% increase, so an additional 20%. As I said,the jump in performance is nice but not a world changer.
Some competitive studies that a professional acquaintance of mine did indicated that if the ARM Cortex A15 meets it's performance goals, it would roughly have IPC [1] parity with AMD's Bobcat. Bobcat is now at 1.6 GHz and the Cortex A15 is being designed for 2.5 GHz. This means that the A15 might be up to 10-20% faster than Bobcat. Now, Bobcat is at approximately about 50-60% performance of a high-end notebook today, which leaves the A15 looking very decent performance wise.
I think the real threat to the Cortex comes from the 22nm Atom. If the 30% power savings due to the FinFETs are true [2], taking into account the fact that the current-generation Atoms are still on 45nm, it's conceivable that Atom power consumption might reduce by something like 70%. Also note that the A15 will still be on either 28nm or 32nm, so due to Intel's manufacturing muscle means the Atom part will be very competitive.
Things are beginning to get interesting in the microprocessor industry. :-)
[1] IPC: Instructions per clock cycle.
[2] I'm not a circuits guy but my semi-informed guess is that the numbers are not far from reality.
It will take a while for Intel to move Atom to 22nm. They are just moving it to 32nm next year. Normally, that means they would switch to 22nm in 2014, but if they are in a real hurry they'll do it in 2013. But I'm not seeing them doing it any faster than that. Two years is quite a lot of time in the mobile industry. It might be way too late for Intel, then.
Plus, they'll only be able to get close with something like a single core Atom, like they are trying this year. We'll have quad core Cortex A15 chips by then, which means even if Atom catches up in energy efficiency, the ARM chips will be significantly more powerful by then - all at the same price or cheaper than Intel's Atom. So I don't see much incentive for manufacturers to use it.
Don't forget the A15 is also at least a year away.
I'm not sure what you mean by this: "Plus, they'll only be able to get close with something like a single core Atom, like they are trying this year."
> We'll have quad core Cortex A15 chips by then.
I don't think this is a big deal. Intel already have dual-core Atoms which actually have 4 threads, which are going to be quite competitive out of the box with a quad core ARM chip.
> all at the same price or cheaper than Intel's Atom.
I don't buy this. Intel's fabs can blow the pants off everyone else in terms of yield and cost. The big pile of cash they're sitting will also help.
> Two years is quite a lot of time in the mobile industry
This might be true, but my points were specifically made for laptops and not mobile phone chips.
Intel's dual core chips have a 10W power draw. If they moved it today to the 22nm Ivy Bridge. it would still use like 5W.
My point was that while ARM only has to double in performance every year and keep the same level of energy efficiency (which seems every ARM chip maker can do), Intel will have to double the performance of Atom every year, while also making the power draw of the chip much smaller than the previous generation. So Intel has to do twice as much in the same time as ARM chip makers.
We can't be more than a few years away from the iPad or iPhone becoming a full computer. Slide it next to a keyboard, mouse, and monitor and it transforms into a desktop.
It seems unlikely for high end laptops and desktops. Intel continues to dominate that performance category and doesn't show any sign of letting up. When it comes time to crunch hours upon hours of video and audio you are going to want a high end Intel processor. The lower end portables in Apple's lineup I can definitely see incorporating ARM processors.
I agree - a Macbook Air and an iMac running iOS makes some sense, but not things like the Macbook Pro and the Mac Pro - tools like Photoshop (and even things like iMovie) would be crippled by the reduced CPU power.
That's exactly what I think, too. Obviously they won't start putting ARM chips in iMacs and Macbook Pro's anytime soon. But a quad core ARM chip in a Macbook Air would be good enough for most people. It's not like Macbook Air has the latest Intel chips in it or have a very high clock frequency anyway.
.. original post on semiaccurate is filed under the category "humor" and backed by the "fact" that previous rumours mentioned on sa turned out to be "true".
It seems very unlikely to me that they'd switch the MacBook line to ARM and rely on an emulation layer to run current applications. ARM processors are getting faster all of time, but catching up to full Intel CPUs is going to be a very difficult task.
Meanwhile, Apple would have to support ARM-based MacBooks and Intel-based workstations until ARM became fast enough to replace workstation-level processors. While software makers caught up to the new architecture, MacBooks would lose much of their 'desktop replacement' status unless they truly could emulate x86 and x64 code with impressive speed.
I really doubt that Apple would split the architectures of their laptop and desktop lines.
If Microsoft can move to ARM, so can Apple. ARM chips don't have to beat Intel's most powerful chips from the start. They just need to get good enough for a Macbook Air, and then transition them to the other notebooks once they become more powerful over the coming years.
A key selling point of the Air is that it is a full-fledged Mac computer. If it has to run normal Mac applications in an emulation layer, which will undoubtedly be slow because the Air's processor is not powerful, this will not be true.
Also, this is the whole endgame of the A* series of chips that are currently in the phone and iPad. Whose to say that the A5 or A6 won't be multicore laptop/desktop chips? Apple would be fools to tie their future Intel's roadmap if they can possibly avoid it. It's not like Apple buys in sufficient quantities to force Intel to meet their specific requirements in chip designs, and even if they did, there would still be a year turnaround to get something from Intel to market. Imagine if the iPad was dependent on the Intel Atom? We never would have seen the iPad.
(btw. an iPad based on the Atom was an actual early prototype according to a friend of a friend that I have no reason to disbelieve).
At first look, there would appear to be two substantial roadblocks to this happening. The first is simply performance. For computers (as opposed to iPads) performance is king, and Intel is the undisputed champion of performance, at least at the moment. We would need to see substantial advances in performance by ARM manufacturers that we just aren't seeing yet - maybe NVidia has the capacity to make it happen?
The second problem is that trying to take an application that runs on a powerful Intel chip, and then emulate the Intel instruction set on an ARM chip would be just diabolically slow. Apple have already changed the processor family twice for the Mac (68000->PowerPC, PowerPC->Intel), but in both cases the new architecture was substantially more powerful than the predecessor. That is not the case with a transition from Intel to ARM.
Of course, this time around there is a potential game-changer, the Mac App Store. Applications built for the App Store are at least in theory very easy to rebuild for another CPU architecture, and then push to users. You have an app bought for your Intel Mac, you "upgrade" to an ARM Mac, and the App Store automatically grabs the ARM version of your software for you.
Still, the two problems are showstoppers, each in its own right. Power has to improve substantially, and the Mac App Store needs to cement itself as the way to distribute software on the Mac before any such transition could be envisaged. Which is not to say that Apple doesn't already have a build of Mac OSX running on an ARM motherboard - indeed I would think it likely that they do have such a beast in the lab - but its purpose would be more as a bargaining chip with Intel, and as an interesting future direction, rather than as an actual product.
Not so much for laptops. Given the existence and usage of the Atom, there are plenty of people that would be fine with battery life, heat, and weight changes that could be brought by the ARM. If the Macbook / Macbook Air switched to ARM (I would guess an A-15 Cortex), then they would fit in fine and app developers would get their software recompiled. Heck, cross platform developers need to be ready for W8 on ARM.
I don't think that Apple see the world that way. If I'm reading the Steve Jobs tealeaves correctly, from Apple's perspective anyone that is happy with an Atom-based laptop today is tomorrow's iPad user, and Macs are going to be reserved for those cases where you really need some power under the hood.
I get the point, but am not sure I 100% agree. I think your dead on about the "Pro" line, but I can see the non-pro laptops going ARM. The customization ability might be the key.
Apple was on Motorola 68k for some 10 years, then on PowerPC for some 10 years and now they're five years on Intel. If this is a pattern, they may be working on the "next big thing" in a lab somewhere.
We know that OSX is very welcoming to architectures, with it's NeXT roots supporting four architectures simultaneously.
A situation of a low-power ARM and high-performance Intel similarly to the dual-GPU architecture seems more likely.
I hope this stays a rumor. While I certainly prefer Mac OS to Windows, there's still some development left to be done under Windows (Windows CE based barcode scanners) and there are still more games available for Windows than for the Mac.
Right now I can do the windows development in VMWare at acceptable speeds and the gaming in bootcamp at excellent speeds.
If they move to ARM, that means no more gaming with Bootcamp and very, very slow development if VMWare should decide to even begin emulating Intel processors.
This would possibly force me to move away from Macs again which I would really, really hate to have to do. Please, Apple, think this through before rushing it.
I think this is just a far fetched rumor. But, didn't Microsoft just demo Windows running on ARM and supposedly Windows 8 will be native on both architectures?
"Now you realize why Apple is desperately searching for fab capacity from Samsung, Global Foundries, and TSMC. Intel doesn’t know about this particular change of heart yet, which is why they are dropping all the hints about wanting Apple as a foundry customer. Once they realize that Apple will be fabbing ARM chips at the expense of x86 parts, they may not be so eager to provide them wafers on advanced processes."
A more plausible direction, which could have been twisted into this rumor, would be an iPad with a keyboard. It could invade the laptop niche, without the support horrors of introducing an incompatible Mac.
ARM's only advantage over x86 is power usage and licensing costs. In every other factor, x86 absolutely blows it away. Alot of the newer tech coming from intel is very, very sexy as well.
ARM has its place and is making great strides. But I for one wont be buying a ARM apple anytime soon.
Another advantage of ARM is a simpler and more straightforward instruction set, unburdened by level after level of legacy. This makes both compilers and chips simpler, more efficient, easier to design and less bug-prone.
There is nothing superior about x86, the Intel advantage is not because of the instruction set but because of Intel's design/production processes. If Intel produced ARM chips they'd blow everything away.
Part of this is probably Apple trolling Intel, but this one is pretty believable because Apple generally don't care about protecting their developers' investments, and because they can design ARM chips themselves.
Consider that Apple has neglected the non-Pro MacBook for quite a while: it still uses a Core 2 Duo processor, and the MacBook Pro line includes a small-form-factor 13" machine that uses the i5 and i7 chips.
If this rumor has any legs, I can definitely imagine Apple moving to ARM for the MacBook Air and MacBook, to which will greatly benefit in terms of battery life and size along with its iPhone, iPad, and Apple TV bretheren. Most of these customers will be using the App Store, so the "fat binary" problem is probably much less of an issue, since Apple can mandate that submitted apps must be bundled for both platforms, which shouldn't be much of a problem if XCode supports it. They will remain with Intel for the MacBook Pro and desktop lines, for customers who need the CPU power, and are less sensitive to power usage.
44 comments
[ 3.9 ms ] story [ 98.5 ms ] threadthis could discourage apple from making a major processor switch again, especially with the high overall costs (compatibility etc.) a switch would introduce.
I think Apple is planning to move to ARM once they have their (most likely) quad core A6 chip.
The tri-gate transistors have lower gate delays (they can switch more quickly) and can operate with lower voltages. How is that not a breakthrough in performance?
And if this is just a one time only "big" jump in performance, then even if it's 100% increased performance over sandy bridge, it's still not that much. ARM chips double in performance every 12 months. That's my issue with this. Because most people seem to think Ivy Bridge puts Intel so much closer to ARM, but it doesn't.
Let's say that without 3D, the jump from 32nm to 22nm would give them 30% increased energy efficiency. Now with the 3D it gives them 50% increase, so an additional 20%. As I said,the jump in performance is nice but not a world changer.
Some competitive studies that a professional acquaintance of mine did indicated that if the ARM Cortex A15 meets it's performance goals, it would roughly have IPC [1] parity with AMD's Bobcat. Bobcat is now at 1.6 GHz and the Cortex A15 is being designed for 2.5 GHz. This means that the A15 might be up to 10-20% faster than Bobcat. Now, Bobcat is at approximately about 50-60% performance of a high-end notebook today, which leaves the A15 looking very decent performance wise.
I think the real threat to the Cortex comes from the 22nm Atom. If the 30% power savings due to the FinFETs are true [2], taking into account the fact that the current-generation Atoms are still on 45nm, it's conceivable that Atom power consumption might reduce by something like 70%. Also note that the A15 will still be on either 28nm or 32nm, so due to Intel's manufacturing muscle means the Atom part will be very competitive.
Things are beginning to get interesting in the microprocessor industry. :-)
[1] IPC: Instructions per clock cycle. [2] I'm not a circuits guy but my semi-informed guess is that the numbers are not far from reality.
Plus, they'll only be able to get close with something like a single core Atom, like they are trying this year. We'll have quad core Cortex A15 chips by then, which means even if Atom catches up in energy efficiency, the ARM chips will be significantly more powerful by then - all at the same price or cheaper than Intel's Atom. So I don't see much incentive for manufacturers to use it.
I'm not sure what you mean by this: "Plus, they'll only be able to get close with something like a single core Atom, like they are trying this year."
> We'll have quad core Cortex A15 chips by then.
I don't think this is a big deal. Intel already have dual-core Atoms which actually have 4 threads, which are going to be quite competitive out of the box with a quad core ARM chip.
> all at the same price or cheaper than Intel's Atom.
I don't buy this. Intel's fabs can blow the pants off everyone else in terms of yield and cost. The big pile of cash they're sitting will also help.
> Two years is quite a lot of time in the mobile industry
This might be true, but my points were specifically made for laptops and not mobile phone chips.
My point was that while ARM only has to double in performance every year and keep the same level of energy efficiency (which seems every ARM chip maker can do), Intel will have to double the performance of Atom every year, while also making the power draw of the chip much smaller than the previous generation. So Intel has to do twice as much in the same time as ARM chip makers.
If not, it isn't really a real computer.
Meanwhile, Apple would have to support ARM-based MacBooks and Intel-based workstations until ARM became fast enough to replace workstation-level processors. While software makers caught up to the new architecture, MacBooks would lose much of their 'desktop replacement' status unless they truly could emulate x86 and x64 code with impressive speed.
I really doubt that Apple would split the architectures of their laptop and desktop lines.
Also, this is the whole endgame of the A* series of chips that are currently in the phone and iPad. Whose to say that the A5 or A6 won't be multicore laptop/desktop chips? Apple would be fools to tie their future Intel's roadmap if they can possibly avoid it. It's not like Apple buys in sufficient quantities to force Intel to meet their specific requirements in chip designs, and even if they did, there would still be a year turnaround to get something from Intel to market. Imagine if the iPad was dependent on the Intel Atom? We never would have seen the iPad.
(btw. an iPad based on the Atom was an actual early prototype according to a friend of a friend that I have no reason to disbelieve).
The second problem is that trying to take an application that runs on a powerful Intel chip, and then emulate the Intel instruction set on an ARM chip would be just diabolically slow. Apple have already changed the processor family twice for the Mac (68000->PowerPC, PowerPC->Intel), but in both cases the new architecture was substantially more powerful than the predecessor. That is not the case with a transition from Intel to ARM.
Of course, this time around there is a potential game-changer, the Mac App Store. Applications built for the App Store are at least in theory very easy to rebuild for another CPU architecture, and then push to users. You have an app bought for your Intel Mac, you "upgrade" to an ARM Mac, and the App Store automatically grabs the ARM version of your software for you.
Still, the two problems are showstoppers, each in its own right. Power has to improve substantially, and the Mac App Store needs to cement itself as the way to distribute software on the Mac before any such transition could be envisaged. Which is not to say that Apple doesn't already have a build of Mac OSX running on an ARM motherboard - indeed I would think it likely that they do have such a beast in the lab - but its purpose would be more as a bargaining chip with Intel, and as an interesting future direction, rather than as an actual product.
Not so much for laptops. Given the existence and usage of the Atom, there are plenty of people that would be fine with battery life, heat, and weight changes that could be brought by the ARM. If the Macbook / Macbook Air switched to ARM (I would guess an A-15 Cortex), then they would fit in fine and app developers would get their software recompiled. Heck, cross platform developers need to be ready for W8 on ARM.
We know that OSX is very welcoming to architectures, with it's NeXT roots supporting four architectures simultaneously.
A situation of a low-power ARM and high-performance Intel similarly to the dual-GPU architecture seems more likely.
Right now I can do the windows development in VMWare at acceptable speeds and the gaming in bootcamp at excellent speeds.
If they move to ARM, that means no more gaming with Bootcamp and very, very slow development if VMWare should decide to even begin emulating Intel processors.
This would possibly force me to move away from Macs again which I would really, really hate to have to do. Please, Apple, think this through before rushing it.
"Now you realize why Apple is desperately searching for fab capacity from Samsung, Global Foundries, and TSMC. Intel doesn’t know about this particular change of heart yet, which is why they are dropping all the hints about wanting Apple as a foundry customer. Once they realize that Apple will be fabbing ARM chips at the expense of x86 parts, they may not be so eager to provide them wafers on advanced processes."
ARM has its place and is making great strides. But I for one wont be buying a ARM apple anytime soon.
There is nothing superior about x86, the Intel advantage is not because of the instruction set but because of Intel's design/production processes. If Intel produced ARM chips they'd blow everything away.
If this rumor has any legs, I can definitely imagine Apple moving to ARM for the MacBook Air and MacBook, to which will greatly benefit in terms of battery life and size along with its iPhone, iPad, and Apple TV bretheren. Most of these customers will be using the App Store, so the "fat binary" problem is probably much less of an issue, since Apple can mandate that submitted apps must be bundled for both platforms, which shouldn't be much of a problem if XCode supports it. They will remain with Intel for the MacBook Pro and desktop lines, for customers who need the CPU power, and are less sensitive to power usage.