As previously noted in this thread Microsoft is creating a windows 10 for arm that will be x86 compatible.
https://www.howtogeek.com/309119/what-is-windows-10-on-arm-and-how-is-it-different-from-windows-rt/
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Interested in Developing ARM-Based Mac Processors and iPhone Modems in House
- Thread starter MacRumors
- Start date
- Sort by reaction score
As previously noted in this thread Microsoft is creating a windows 10 for arm that will be x86 compatible.
https://www.howtogeek.com/309119/what-is-windows-10-on-arm-and-how-is-it-different-from-windows-rt/
Doesn't matter to Apple. They likely figure out that some number of people with switch to buying an x86 PC, but even more PC users will switch to getting a fast ARM Mac for better battery life and portability. If not for style. Win for Apple.
They’ve already out executed intel. Compare volumes. Compare performance per watt. Compare performance per square mm. Compare performance per dollar. Intel wins only at the super high end where cost and power are no object, which is not where most macs have been in the intel-Mac era anyway.
[doublepost=1506750624][/doublepost]
Exactly. No fan, 24 battery life, running anything from the MAS out of the box and running office and CC within a year of release (or sooner via emulation). Running all of the iOS App Store out of the box too. My only doubt re emulation is intel might very well sue, though they like Apple as a customer still for baseband chips, and for all we know intel might actually end up being apple’s ARM fab.
I disagree completely. IBM/Motorola PPC alliance is my prime example (PPC is an offshoot of the Power arch) they were so geared for high power computing they couldn’t make a mobile G5 processor cool enough and power efficient enough to compete in the growing laptop market which was one of if not the main reason for Apple switching to Intel! We were at the point where dirt cheap PC laptops were wiping the floor with very expensive PowerBook G4s.
Additionally, Apple doesn’t really design its own chips from the ground up. It uses ARM designs and tweaks them, not unlike what happened with the PPC being downstream tweaked Power cores. What will happen when ARM engineers don’t care about desktops especially if only Apple is doing ARM desktop in volume and has a small desktop market share. Apple can tweak and hype all they want a mobile design and load are considerably different than a desktop load. It may be acceptable for office applications but it will be terrible for scientists and media people that have massive compute loads and are the Mac’s biggest acolytes.
If Apple could persuade Intel to play foundry for Apple ARM designs aimed at the 15-45W power envelope (which, at this point, are not publicly known to exist), it's not hard to imagine that Apple would seriously consider the option. But Intel taking that deal strikes me as highly unlikely.
So if Intel won't do it, who would (or could)? That would be the other foundries with leading-edge High power/performance manufacturing processes, of which there are three(-ish): TSMC, GloFo, and Samsung. The problem for Apple is that those foundries remain at least one process generation behind Intel for higher power envelopes.
The GeekBench benchmarks that you're seeing, where an iPhone X scores slightly under a MacBook Pro on single core (by core I assume they mean threaded) performance and slightly over on multicore performance, compares the A11 against an i5 7267U. That's a a dual-core, SMT chip, listed at 28W TDP, manufactured on Intel's 14nm+ process. In those systems (the system, including screen resolution, seems to have an unusually large impact on GeekBench scores), the A11 and i5 7267U score, respectively, 4197 and 4223 on single thread, and 9861 and 8947 on multithread, performance, according to Geekbench. For comparison, the i7 7700HQ (4 cores, 8 threads, 45W TDP, 14nm+) found in the 15in MacBook Pro scores 4356 and 14400. A stock Ryzen 7 1700 (8 cores, 16 threads, 65W TDP, 14nm GloFo) scores 3890 and 19935.
Apple's A11 is manufactured on TSMC's newly launched "10nm" process (which is slightly smaller than Intel's 14nm process in terms of feature size). One reason that Apple is able to ship product on TSMC's 10nm process is due to the A11's low power and relatively conservative (in terms of die size) design. Apple's recent A-series chips devote more die space, compared to high performance Intel parts, to cache memory, which is significantly more dense than logic. TSMC's 10nm process is not yet available for higher power (and lower density) designs, and the comparatively memory-dense A-series SOC designs mean smaller die sizes and higher yields. This is worth bearing in mind, given that we're talking about scaling Apple's designs up for increased performance and higher power envelopes (to feed additional power hungry logic, of the sort that would be competitive with something like the i7-7700HQ or comparable chips in the forthcoming 8000-series).
With that in mind, an interesting comparison point will be available next week, with the release of Intel's Coffee Lake architecture on its 14nm++ process. That process will actually have lower feature density than the 14nm+ process, but is expected to have significantly better performance/power characteristics (reportedly comparable to what Intel expects out of its 10nm process at launch). That is, it will give a sense of where Intel is currently at, from a performance/power standpoint, on a process optimized for higher power designs.
To circle back, the question is this: it's one thing to create a great design, bespoke for your needs, with all the hand layout and optimizations that requires, for a chip that targets a low power / high density manufacturing process. Though that's not to say that Apple couldn't create a higher power, more logic-heavy bespoke design that, on exactly the same process, that wouldn't perform better for Apple's needs than whatever Intel has designed for the same target market. In fact, at this point, I'd be surprised if Apple couldn't.
But even in the aggregate, architecture, design, and layout improvements over a prior design have a hard time matching the performance and power gains that would come from simply moving the old design, with no changes, to a new process. As a result, if Apple can't get someone to manufacture its latest and greatest designs, on the schedule it wants, on a process comparable to what Intel is using, it's hard to see the ultimate benefit (that couldn't otherwise be had for a much smaller investment in custom off-CPU hardware) of trying to use design resources to compete with or match Intel's process technology, even if that means using an Intel design and architecture that are inferior to what Apple could come up with starting from scratch.
So if Intel won't do it, who would (or could)? That would be the other foundries with leading-edge High power/performance manufacturing processes, of which there are three(-ish): TSMC, GloFo, and Samsung. The problem for Apple is that those foundries remain at least one process generation behind Intel for higher power envelopes.
The GeekBench benchmarks that you're seeing, where an iPhone X scores slightly under a MacBook Pro on single core (by core I assume they mean threaded) performance and slightly over on multicore performance, compares the A11 against an i5 7267U. That's a a dual-core, SMT chip, listed at 28W TDP, manufactured on Intel's 14nm+ process. In those systems (the system, including screen resolution, seems to have an unusually large impact on GeekBench scores), the A11 and i5 7267U score, respectively, 4197 and 4223 on single thread, and 9861 and 8947 on multithread, performance, according to Geekbench. For comparison, the i7 7700HQ (4 cores, 8 threads, 45W TDP, 14nm+) found in the 15in MacBook Pro scores 4356 and 14400. A stock Ryzen 7 1700 (8 cores, 16 threads, 65W TDP, 14nm GloFo) scores 3890 and 19935.
Apple's A11 is manufactured on TSMC's newly launched "10nm" process (which is slightly smaller than Intel's 14nm process in terms of feature size). One reason that Apple is able to ship product on TSMC's 10nm process is due to the A11's low power and relatively conservative (in terms of die size) design. Apple's recent A-series chips devote more die space, compared to high performance Intel parts, to cache memory, which is significantly more dense than logic. TSMC's 10nm process is not yet available for higher power (and lower density) designs, and the comparatively memory-dense A-series SOC designs mean smaller die sizes and higher yields. This is worth bearing in mind, given that we're talking about scaling Apple's designs up for increased performance and higher power envelopes (to feed additional power hungry logic, of the sort that would be competitive with something like the i7-7700HQ or comparable chips in the forthcoming 8000-series).
With that in mind, an interesting comparison point will be available next week, with the release of Intel's Coffee Lake architecture on its 14nm++ process. That process will actually have lower feature density than the 14nm+ process, but is expected to have significantly better performance/power characteristics (reportedly comparable to what Intel expects out of its 10nm process at launch). That is, it will give a sense of where Intel is currently at, from a performance/power standpoint, on a process optimized for higher power designs.
To circle back, the question is this: it's one thing to create a great design, bespoke for your needs, with all the hand layout and optimizations that requires, for a chip that targets a low power / high density manufacturing process. Though that's not to say that Apple couldn't create a higher power, more logic-heavy bespoke design that, on exactly the same process, that wouldn't perform better for Apple's needs than whatever Intel has designed for the same target market. In fact, at this point, I'd be surprised if Apple couldn't.
But even in the aggregate, architecture, design, and layout improvements over a prior design have a hard time matching the performance and power gains that would come from simply moving the old design, with no changes, to a new process. As a result, if Apple can't get someone to manufacture its latest and greatest designs, on the schedule it wants, on a process comparable to what Intel is using, it's hard to see the ultimate benefit (that couldn't otherwise be had for a much smaller investment in custom off-CPU hardware) of trying to use design resources to compete with or match Intel's process technology, even if that means using an Intel design and architecture that are inferior to what Apple could come up with starting from scratch.
Last edited:
^^That^^
https://boingboing.net/2012/01/10/lockdown.html
https://techcrunch.com/2015/04/18/on-the-war-on-general-purpose-computing/
https://www.eff.org/deeplinks/2017/08/war-piracy-turns-streaming-media-box-community
Maybe I'd better get in that order to Boxx...
I will be on the same boat as well.
ARM currently still no match for x86, and it's far away from close for its general usage.
Having a power efficient computer means nothing for power user. This is what MacBook "PRO" is meant for.
But recent update with Touch Bar and taking out all the useful ports, I honestly starting to think "PRO" is just a joke.
They are shifting the marketing to regular consumer than high end market.
Most tech companies buy bulk load of MacBook Pro for their employee will be shifting to Windows if it ever happen.
Many of them already starting to find problems with ports issue for development. Designer starting to freak out on awful color uniformity on the latest MacBook Pro as well.
This could be a good move..... If u develop in house, nobody to rely on when chips get delayed, no legal battles with agreements.
What could possibly go wrong.
Ya, so it will be more difficult when u move to ARM, but iOS developers don't seems to worried now. You'll only be worried be worried becuse you've enjoy x86, but ARM should be able to support that.
What could possibly go wrong.
Ya, so it will be more difficult when u move to ARM, but iOS developers don't seems to worried now. You'll only be worried be worried becuse you've enjoy x86, but ARM should be able to support that.
Last edited:
No apple doesn’t tweak arm designs. It designs its own chips to the arm architecture (since their second design) just like amd designs it’s own chips to the x86-64 architecture. They design their own microarchitectur, cores, floorplans, buses, and do their own physical design. Nothing in any A-series chips looks anything at all like what you can get from ARMs design team anymore.
Here is why I keep saying that Apple should acquire AMD. https://www.enterprisetech.com/2014/05/05/amd-unify-x86-arm-systems-skybridge/
https://www.anandtech.com/show/9232/amds-k12-arm-cpu-now-in-2017
https://www.anandtech.com/show/9232/amds-k12-arm-cpu-now-in-2017
Last edited:
So we see two reasonable arguments against:
1) They can't switch to ARM because they need Macs that perform at the high end, and we are seeing Apple move into higher performance with the iMac Pro, and the upcoming new Mac Pro.
2) They can't switch to ARM because they would lose those who use x86 VMs and/or Boot Camp to run Windows and Linux.
To address #1, I would say that the A11 isn't close to the highest-end Intel chips, but who says they would put a mobile chip in a high-end Mac? They have all kinds of head room to innovate in silicon when there are fewer thermal constraints. I agree they wouldn't do this without a powerful workstation-class chip that can match or best Intel, but I think they can get there in the reasonably near-term.
#2 is a more difficult nut. As for Linux, I suspect ARM would become a first-class platform for the major distributions. For the Mac users who need multi-platform support on one machine, Apple would need to consider carefully what they would lose by alienating these users. I personally don't think it would actually make a major dent in Mac sales and those who develop both for Apple platforms and Windows would suffer an inconvenience in needing two machines, but I doubt any significant number would abandon the Apple platform over this.
1) They can't switch to ARM because they need Macs that perform at the high end, and we are seeing Apple move into higher performance with the iMac Pro, and the upcoming new Mac Pro.
2) They can't switch to ARM because they would lose those who use x86 VMs and/or Boot Camp to run Windows and Linux.
To address #1, I would say that the A11 isn't close to the highest-end Intel chips, but who says they would put a mobile chip in a high-end Mac? They have all kinds of head room to innovate in silicon when there are fewer thermal constraints. I agree they wouldn't do this without a powerful workstation-class chip that can match or best Intel, but I think they can get there in the reasonably near-term.
#2 is a more difficult nut. As for Linux, I suspect ARM would become a first-class platform for the major distributions. For the Mac users who need multi-platform support on one machine, Apple would need to consider carefully what they would lose by alienating these users. I personally don't think it would actually make a major dent in Mac sales and those who develop both for Apple platforms and Windows would suffer an inconvenience in needing two machines, but I doubt any significant number would abandon the Apple platform over this.
There are lots of things for which touch with fingers is not fine enough nor practical enough such as cad, programming, gaming, spreadsheets, page layouts.
I have touch but wouldn't dream of using it for those.
But in someways you are right, if those growing up can learn to use touch to do the above things or interfaces change to allow them, then generation change will see PC/Macs disappear.
The cores are based on liscensed ARM v8 designs, hence being downstream. Yes Apple does chip work but it is not running its own spec free to define what it wants, it is downstream and bound by the ARM lisence. In fact if Apple did try to clone the core without licensing, I’d expect a massive lawsuit on par with Intel v AMD. If you really think Apple reinvented the wheel you need to put down the koolaid.
ARM v8 is not a core design. It is an isa. Did you even know the very first Arm v8 implementation was the Apple a7? It was even speculated at the time the isa was a collaboration when Apple and ARM given how fast the a7 came out. The Apple a7 SOC came out more than a full year (it was about 16 months) before any other implementation of arm v8.
Looking at the actual cores, the Arm’s implementations looks nothing like what Apple is doing. For one thing Apples design is 6 wide compared to 2-3 from arm’ own designs. Also per core performance of the a11 is more than double of the cortex a73 (arms latest core design).
It would be difficult to imagine reasoning any more specious than this. I, too, was not "down" with iMacs when I was 2. And?
I worked at AMD for ten years. Apple really should not.
[doublepost=1506779595][/doublepost]
Apple has an ISA license. This means they can do whatever they want. They do their own design. Many of my former colleagues work there, and that is what they do. These are not “make minor tweak” people.
+1000 likes if I could.
After all these years I find it odd that people still claim Apple just tweaks ARM designs. They apparently don’t know the difference between an instruction set architecture (ISA) like x86 or ARMv8 and a microarchitecture (Apples A Series cores or ARMs A73/A53 cores).
Apple designs their own fully custom microarchitecture (cores) to run the ARMv8 ISA. And they are obviously highly skilled given how much better they perform than ARM cores.
Another way to look at it, how can Apple simply “tweak” an ARM design and end up getting twice the performance? Are ARM cores so poorly designed that a few mods by Apple can extract that much additional performance? Why can’t ARM double the performance of their own cores with a few simple “tweaks”? For that matter, why can’t Samsung or Qualcomm “tweak” ARM cores and double the performance of their processors?
Apple doesn’t do everything perfect, but their processor design team is clearly world class.
after seeing the a11 bionic chip performance its not crazy to think this. Apple is winding up and getting ready to obliterate intel. Think about it....The Apple TV 4K has the iPad Pro A10X in it. The scores 9480's with geekbench and only consumes 6 watts (source: https://images.apple.com/th/environment/pdf/products/appletv/Apple_TV_4K_PER_sept2017.pdf) If apple sandwiched 3 of those chips together they could match intels latest 8700K chip and it would only consume 18 watts! vs intels 95w tdp. if they used the A11 chip they could blow past intels latest offering...i know its not that simple and geekbench isnt the best benchmark to go by. but its definitely a possibility
Recompiling code for ARM isn't that big a deal. Goodness I've written frameworks that compile for watch/phone/ipad and MAC OS.
The one thing that seems to keep getting lost in this discussion is Intels inability to provide adequate chip updates. That hs to be ticking off Apple constantly.
[doublepost=1506785313][/doublepost]
You sound like Ballmer talking about the original iPhone.
The one thing that seems to keep getting lost in this discussion is Intels inability to provide adequate chip updates. That hs to be ticking off Apple constantly.
[doublepost=1506785313][/doublepost]
You sound like Ballmer talking about the original iPhone.
If fact, the most recent Apple 64-bit processor cores look a lot more like a high-end Intel or AMD design (instruction dispatch width, & etc.) than any of the stock designs out of ARM/SoftBank.
It's quite possible that the Apple's processor team put together a high-performance micro-architecture to outright beat Intel's high-end, and then detuned their final design and layout for TSMC's low power process, in order to fit it in a thin mobile phone's power envelope.
x86 PC sales are down. If Intel's large very expensive fabs start running under capacity, enough to affect Intel's bottom line, there may be an interesting opportunity here for the two companies.
Last edited: