“Yes, Apple’s newest A7 Cyclone SoC is a beast — but it’s a long, long way from being in any way competitive with Intel’s Core chips in terms of performance,” Joel Hruska writes for ExtremeTech. “That aside, there are three good reasons why Apple won’t adopt ARM across its entire product line anytime soon — and one good reason why Intel should be worried about the long-term roadmap.”
• Software compatibility. [Intel-based Macs run the world’s largest software library. Period. – MDN Ed.]
• Hardware competitiveness. This is an issue people gloss over by claiming that Intel’s weak mobile performance is equivalent to the G5′s problems in the mid-2000s. That’s a vast oversimplification…
• Timing and stability. Every time Apple considers a major product transition, it considers the entire ecosystem today and several years into the future — and it makes a move only when it’s convinced that a winning horse has emerged from the pack…
“The one reason why Apple might seriously build its own ARM core for high-end computing is that such an approach would give it more control over its product family, its roadmap, and its future,” Hruska writes. “But as much as Apple prefers controlling its own destiny, it’s also not willing to compromise the performance and high-quality brand its built just because it can build its own chips. If Apple ever makes the jump from x86 to a future ARM-based processor, history tells us it will do so across the entire product line at once, and only when it can deliver a significant performance improvement.”
Much more in the full article here.
Related articles:
Intel-powered Macs: The end is nigh – August 4, 2014
Intel’s Broadwell chips further delayed; not shipping for most Macs until early-mid 2015 – July 9, 2014
Apple will inevitably drop Intel for their own A-series processors in the Mac – June 26, 2014
How long before Apple dumps Intel from MacBook Air? – June 26, 2013
At last an article that makes sense on this subject
Software compatibility
All a Mac NEEDS to run is OSX and OSX apps. If you need other OS’s then you’re not the core market 🙂
• Hardware competitiveness
I don’t get this one. If I’m only running OSX on it, and OSX runs well, then you don’t need to be competitive.
• Timing and stability
Ummm, this is a stretch. Apple’s hit their stride in their processor releases for Mobile, why do we think it’s be any different for anything else they do?
In the end, knowing that this will be online forever, they hedged their bets to basically say “If Apple can ever produce a processor that runs OSX and OSX Applications at a performance that’s better then Intel, though, they’ll do it”
Which is pretty much what everyone on the “Mac on A(n)” has been saying. So, good article, but perhaps not for the reason you think 🙂
“Software compatibility
“All a Mac NEEDS to run is OSX and OSX apps. If you need other OS’s then you’re not the core market :)”
Macs currently run Mac OSX, Window, UNIX, Linux, and a large number of specialty, niche OSes (e.g., some real-time OSes for experimental data collection). If Apple moves to A series chips all but Mac OSX and iOS go away. They will have to be run under emulators. You remember how slow Windows ran under the 68000 and PowerPC series chips don’t you? Even a 10% speed hit can keep lots of people from buying Macs. And then there are emulation compatibility issues on top of that.
“• Hardware competitiveness
“I don’t get this one. If I’m only running OSX on it, and OSX runs well, then you don’t need to be competitive.”
Maybe *you* don’t get this one, but Apple does.
The chipsets that go into the current motherboards are commodity items. This is because Apple just uses a slight variation on the standard theme of Intel systems (except for the Mac Pro’s physical layout — and even then the electronics design is just a variation on the standard Intel based personal computer). This makes the components much less expensive.
As just one example, think back to the 68000 based Mac days. NUBUS was MUCH better than anything the DOS/Windos side had. But, Apple was the only major player supporting it. This, among other things, made the Mac motherboards much more expensive to build than the run of the mill DOS/Windows motherboard.
“• Timing and stability
“Ummm, this is a stretch. Apple’s hit their stride in their processor releases for Mobile, why do we think it’s be any different for anything else they do?”
Timing is just a “do it when it makes sense”, and I believe we all agree on this. We don’t seem to agree on the “when” part.
How do you know Apple has “hit their stride” in the processor releases? Do *you* know the reason why Apple didn’t release an A7x chip with a higher resolution iPad Air? Is it because Apple couldn’t make that chip in time, a high enough yield, or with a necessary performance/power ratio? No one outside of Apple’s senior staff knows this for sure. But to flatly state that Apple *has* hit the mark every time is ignoring the fact that you don’t know they have.
“If Apple moves to A series chips all but Mac OSX and iOS go away”
And that should be the case. Yes, there are users that DO those things, but most users don’t. We don’t expect iOS devices to run any other OS, if this becomes the future, the same would be true for OSX.
“This, among other things, made the Mac motherboards much more expensive to build than the run of the mill DOS/Windows motherboard.”
Apple has reached economies of scale with their iOS hardware, I’m confident they could do the same with any potential desktop class replacement. Plus, how much of a motherboard’s expense now is just because Intel knows they can charge for it.
“But to flatly state that Apple *has* hit the mark every time”
It was semantics, sorry. Apple has “hit their stride” in that, along their current processor maturity path, they have produced a 64-bit dual core low power processor that, even at a raw level outperforms the competition. So, even if they are doing horrible by their internal standards (they meant to have quad core, 128 bit at 4 GHz), what escapes into the world is beating companies that have been doing this for awhile. They could stumble and be stuck here for two years, but it looks like they’re going to widen the gap.
I think a lot of people agree that Apple WILL do it when it makes sense (even the article makes that point contrary to the subject). My disagreement is with those that think that it will never make sense.
Mac and OS X is *not* the core market, in general. The Mac/OS X is stronger in the consumer/home market and small business market, particularly in the U.S. The proliferation of the iPhone and iPad (iOS) has also improved the penetration of the Mac in the corporate sector. So your smiley face doesn’t mean much.
With respect to hardware competitiveness, the issue is Mac competitiveness in relation to Windows PCs. The PowerPC architecture started out strong, but gradually dropped behind the intel Pentium, especially for mobile applications (laptops). Apple suffered a great deal from this fact. When Apple switched to intel CPUs, this issue largely disappeared. Macs and Windows PCs have access to the same CPU components, and performance differences come down to the operating system, drivers, and application software. If Apple were to split away from intel CPUs, then it would not only sacrifice native Windows execution, but it would also put itself into a position in which CPUs became a discriminator between Macs and Windows PCs. If that worked in Apple’s favor, then all is good. If not (like the stagnant G4 and G5 PowerPC development of the late 1990s and early 2000s), then Apple is placed at a disadvantage.
With respect to timing and stability, Apple has the advantage in the smartphone and tablet space because ARM was very power efficient and intel was weak in that area of CPU development, having focused primarily on higher power applications – servers, desktops, and laptops. But intel is a very large and capable company with a massive microprocessor fabrication infrastructure. Apple will be taking a risk if it goes up against intel in the CPU space.
Rewritten just for you 🙂
If you need other OS’s running on your Mac other than OSX, then you’re not the core Mac market.
“If that worked in Apple’s favor, then all is good.”
Apple will do it once they can show, “Hey, look at this speed demo comparing FCPX running on Intel and A11. It matches and slightly beats the intel chip at a lower clock rate.”
“Apple will be taking a risk if it goes up against intel in the CPU space”
They’re not trying to unseat Intel as the powerhouse purveyor of performing processors, they just need a CPU that will run current OSX software at acceptable speed. If they succeed, they would never come close to selling as many chips as Intel makes in a year.
Were you a Mac user during the PowerPC era? Apple did everything you wrote (including Photoshop bake offs between the G4 and the then-current Intel CPUs). NOTHING Apple did could convince the tech press that the G4’s performance was at least the equal of the Intel GPUs. And the Apple-haters had a field day with all their FUD. As bad as the FUD is now, it was much worse then.
>If I’m only running OSX on it, and OSX runs well, then you don’t need to be competitive.
I don’t know anybody who is only running OSX. Most of us are running applications, from Apple and other vendors, that run under OSX. That is the “core audience,” with those running other operating systems a subset. (You may regard that subset as irrelevant, but many users do not.) I’m sure you do agree that a “Mac” that can’t run OSX apps is… well, about as useful as a Windows RT machine that can’t run Intel Windows apps. We’ve seen how well that worked out for Microsoft.
Every existing OSX application is compiled to run on an Intel processor. Running them on any sort of ARM chip would require emulation software. Inevitably, the overhead from that layer significantly slows down the software (anybody who had to run Windows software on a PowerPC Mac can testify to that).
The last two CPU transitions (68000->PowerPC->Intel) handled the problem only partly by emulation. It worked because the new processor families were so much faster than the old ones that the speed hit was acceptable to most users. In each case, the new chip families had evolved specifically for speed, not low power consumption like the ARM family and its A-series descendants. It will be a long time before we have an A-chip that can run complex Intel apps under emulation as fast as Intel can run them natively.
The preferred solutions during each transition were “fat binaries” or “universal applications” that contained separate executables for each processor family. The resulting files were huge, and I would guess that an ARM executable would be much bigger than one that can take advantage of the Intel complex instruction set. A lot of software producers didn’t bother, which is why so many programs died when Classic was discontinued and so many more with Rosetta. Even Apple didn’t get all the 68000 code out of the Mac OS for years; it partly relied on emulation long after every current Mac was a PPC. Again, that was only possible because the new processors were so much faster than the old ones.
“I don’t know anybody who is only running OSX”
Understood. That should read if you’re only running OSX and all of the applications that are compiled to run under OSX, then being competitive with a Windows standard doesn’t matter much. Microsoft has a LOT of stake in their Intel friendship. It behooves them to make sure that everyone understands that “Intel is still the best guys!” Apple has no such ties.
“Running them on any sort of ARM chip would require emulation software.”
OR, you just force everyone currently on the app store to recompile. Apple has been pushing developers off of Rosetta, Carbon and into their frameworks for years. Those that have moved see that while it was a pain, you can usually take advantage of new OSX API’s with a recompile. The benefit lasting through each iteration of the OS, the cost being a one time sloughing of old frameworks.
“The last two CPU transitions (68000->PowerPC->Intel) handled the problem only partly by emulation.”
Correct! Primarily, that was due to the requirement to continue to support old frameworks/api’s that didn’t fit into Apple’s plan. That was when you could create applications with whatever you liked. Today, most developers use Xcode. Controlling the development environment means that you can make it create whatever back end code you’d like. In addition, Apple’s in no hurry, they just wait until ARM performance is where it needs to be, release the latest Xcode, tell everyone to recompile, done.
“The resulting files were huge”
I’d have to look into this. I don’t recall the files being huge, but with 1TB drives being standard now, “big files” is not the ominous thing it used to be. I do remember some companies selling products that would allow you to remove the binaries you didn’t need on your CPU.
“A lot of software producers didn’t bother, which is why so many programs died when Classic was discontinued and so many more with Rosetta”
And each one of these was an example of a bunch of API’s that needed to go away because they weren’t re-entrant or suited to the environment Apple has created with OSX. Just recently Apple deprecated most of QTKit in favor of AVFoundation, removing yet another problematic framework that wasn’t scaling well.
I agree with your viewpoint. A couple points:
* Releasing an A-series MacBook won’t stop Apple from continuing to ship higher end Intel MacBook’s for those who need that.
* By contenting to supply Intel MacBooks they open up lots of time to make the transition to A-series at the mid-range laptop market (where Apple has oodles of profits and marketshare to gain).
* Fat binaries are not necessary now that there is an app store that knows what device it is downloading too. Even if transitioning to a new machine with TimeMachine, the app store could detect the wrong binaries on the new Mac and offer to install the right ones.
* All Apple has to do is give a date when the Mac App store will no longer accept apps that are not both Intel and A-series compatible and the transition to A-support will happen very fast.
Find Steve Jobs truck analogy and read it. This guy knows squat.
No, they should stick with intel chips but they should take one of them like the MacBook Air 11 inch and put one their own, the A7, in there and see what happens. But at the same time use a intel chip as well on the 11 inch as a optional processor. I feel that if they make their own chip and put it in a small package like the 11 inch MB Air and optimize OS X they will have a good product at a great price.
I am pretty sure they are always doing things like that in house. Surely they look at many strategies. If and when they find improvements they will adopt them.
EXACTLY.
Everyone seems to forget (or maybe they’re not old enough to remember) the Apple internal Star Trek project in which Apple started playing with using Intel processors in Macs well over 10 years before Apple switched the Mac line from PowerPC to x86.
I would be shocked if there isn’t an A7 and A8 MacBook (of some kind) running in an Apple lab.
Yet, as I’ve said on this site several times, I agree with those that say Apple won’t be moving from Intel processors for Macs any time soon. I’d say 2017 at the very earliest and likely not until 2020 or later.
I would agree with those numbers. They may have a contender before then, but they’re in no hurry as EVERYone would be all lagged together if Intel continues to miss marks.
Despite what many think, Apple is functionality first then decide if it can be done with the hardware or what hardware is needed to achieve it.
For example, finger print sensor: Apple already had decided a multi-year plan for improve security/authentication; it was only when they found the right hardware did they pull the trigger.
Meaning, Apple will not just add another CPU into a box and what what might happen.
Re the actual Article, we should remember that many people use Macbook variations with a VM running Windows. Macs are better PCs than PC. You have the best OS, and have access to aging corporate software.
“Meaning, Apple will not just add another CPU into a box and what what might happen.”
But they will, according to your analogy, create all the required motherboard integrated circuits (which they’re already good at) and just wait until an Aseries chip makes the performance cut.
I find it more feasible that Apple computers may use both. Intel plus A-series chips would allow both OSes to run on one device. And while I don’t think a forced convergence works (see Microsoft Surface), the ability to run iOS apps along side OS X could have it’s benefits.
Macs don’t actually need an A-series chip to run iOS software. Given the current performance gap between the x86 and ARM platforms, an intel chip can emulate an A chip reasonably fast.
Anyone remembers Rosetta?
Just like the iPad showed most people get by on apps that don’t need Intel horsepower, a low cost ARM notebook would have a market for people that need more than an iPad, but don’t need to run Office, Boot Camp, Parallels, vast databases, etc. This would probably include more than 50% of Mac users now. They would be perfectly happy with a lower cost 11 – 13 inch MacBook Air ARM computer. If people like Chromebooks, I really think the Mac-ARM has an even bigger market. PS- Steve would be very pissed about this Broadwell delay, and we know what he would be doing in some black ops room somewhere on campus.
So are you suggesting Apple make a notebook? I think they were smart to avoid that disaster.
There’d have to be an extremely compelling reason to get the ARM Macs over Intel ones. Lower price alone won’t cut it, since a whole slew of existing apps would no longer run on it without a recompile (or if they can run as-is, it’ll be emulated and slower than native Intel).
If users “only” need basic apps like Safari and maybe some iLife apps, they might as well just get an iPad with keyboard.
Surface RT failed precisely because there was no compelling reason or killer app to get it over a more expensive, capable and backwards-compatible Intel-based Surface Pro.
“since a whole slew of existing apps would no longer run on it without a recompile”
Apple controls Xcode. All they’d have to do is include a choice to compile for x86 or A10 or both in a fat binary.
That much is obvious. But they do not control 3rd party apps, so the point was if it’s not open source then you are entirely dependent on them re-compiling it on their schedule, if ever.
A big benefit of a centralized app store is that Apple can force developers to recode at any time. They’ve been doing this for years. After some WWDC they say, “Before your app can be on the app store (after x date), you have to provide a new version that’s been built with the latest version of Xcode.”
They’ll usually try their best to make sure that the effort for devs really is no harder than a “recompile” button (that’s the way it was for the switch to Intel), but, by the time Intel Macs rolled out, i.e., Apple’s schedule, all developers had at least recompiled their apps.
“If users “only” need basic apps like Safari and maybe some iLife apps, they might as well just get an iPad with keyboard.”
You make a very interesting point. Maybe the BIGGEST reason why Apple won’t switch from Intel to ARM for Macs is because, before that would have a chance to happen, Apple would have already stopped making Macs!
Without a general-purpose computer, that’d leave Apple without any workhorse “trucks”. Steve Jobs himself acknowledged that the world cannot get by on cars alone, and metaphorically neither can Apple.
So either iOS’s capabilities gets beefed up a lot or current restrictions are relaxed (you can’t currently develop for iOS on iOS itself, there’s no compiler), or Macs have to stay.
There’s no reason to think that the common set of desktop and tablet capabilities won’t continue to increase. At this point, I still see desktops being needed, but that could be just because I can’t envision all the pieces required to be in place to do otherwise.
I think “iPad with keyboard and external storage” but that’s just the modified desktop paradigm. There’s likely projects Apple will release in the future that will seem obvious once they’re out. 🙂
I thought the whole point of running winblows on your Mac is to gradually wean yourself off of it as you get more used to OSX(for people switching from winblows to OSX) oh wait, some people (enterprise) just have to have winblows no matter what. My bad. Apple still does enterprise? Could have fooled me. Apple is a consumer based company now, and no matter how much winblows users wish it supported enterprise, get over it, it doesn’t. Apples bread and butter is in the consumer market now. If it wasn’t, we would still have xservers. And while Mac Pro is a professional machine, that is not really aimed at enterprise, but specialty power users.
We have AirPlay. I expect we’ll soon have an Apple tech that lets devices share processing work wirelessly and share a visualization of the work (the “screen”).
In this world the boundaries between devices would essentially blur into nothing. Every device will be able to “run” every bit of software and interact with it.
If I can play a game on my iPhone but have my MacPro in the next room do all the processor work and just send my phone a streaming image my iPhone virtually has the same power as my MacPro.
So also for my iWatch, etc.
Apropos of this article, it soon won’t matter what’s running on what. It MAY matter whether you have all the bases covered in your home — do you own a computer with an Intel chip? do you own a beast of a computing powerhouse — but that may not actually matter, either.
Consider the world in which wi-fi is insanely fast and all “programs” operate on servers in the cloud. Your device seems to be doing the work, but in fact you’re just getting a streaming image off the net. Yes, I know it’s not new, but it will come with new energy. First it will come to your home, as I describe above. Then it will come to you wherever you are, but not from the cloud, rather from your own computers at home through cellular/wi-fi. Then, as necessary, it will come from the cloud.
Incidentally, this will be driven at Apple by the iWatch and wearables which won’t have the capacity to do a lot of processing onboard, but will be able to display data crunched on your iPhone (and eventually your Mac, if necessary), with aplomb.
Anyone remembers xGrid?
>Consider the world in which wi-fi is insanely fast and all “programs” operate on servers in the cloud.
What world is that? It’s not the one I live in. A family of four, each with his or her own 5-megapixel screen drawing down 60 refreshes every second from a remote computer/server, on top of a couple of television sets streaming HD movies, is going to swamp most normal home WiFi networks.
A fair number of us have “broadband” service that uploads data to the cloud at rates of less than 1 megabyte/second and downloads at 5 MBS or less. Because most ISPs are effectively monopolies or duopolies, and their industries have been deregulated, they have no reason to improve their service. As for the cellular providers, they have monthly caps on their 4G data that the hypothetical family would run through in much less than a day.
Maybe someday, but not this year and probably not this decade.
Thank you Joel Hruska who writes for ExtremeTech for being thoughtful and intelligent. Let’s kill THE STUPID RUMOR FROM HELL.
Some of my own relevant Q&A:
Q: Why did Apple every leave RISC PPC chips behind and go CISC Intel?
A: Because IBM lied to Apple for years that they were going to make PowerPC G5 chips efficient and cool running enough to be used in laptops. Apple got sick of the deceit and told IBM to stuff it. Apple leapt into Intel chips because they were the best option at the time. (IBM never did make any mobile G5 chips).
Q: Now that ALL Mac software is based on Intel CISC chip sets, how psyched is Apple to jump the Intel ship and go RISC again via the A-Series ARM chips?
A: Not At All. The recoding of ALL Mac software back to RISC chips again would be an outright nightmare with all the worst consequences imaginable, including outright riots by Mac developers.
Q: Why did Apple go to RISC chips for iOS devices?
A: Because they could. There didn’t need to be ANY code parity between iOS and OS X, despite their identical core kernel. RISC ARM CPUs offered Apple the opportunity to design their own CPUs with no dependence on any particular third party for anything apart from licensing ARM’s core design. RISC also has inherently far less overhead allowing for far more efficient computing for specific services and tasks. Dumping all the vast legacy Intel API baggage was an opportunity Apple could not pass up.
Why THE STUPID RUMOR FROM HELL has any legs is beyond comprehension. But limp along it does, yet another litmus test for stupidity. (Sorry Mr. Gassée).
From the article:
“If Apple ever makes the jump from x86 to a future ARM-based processor, history tells us it will do so across the entire product line at once, and only when it can deliver a significant performance improvement.”
Which is pretty much what folks have been saying all along. Even though his title says otherwise, he agrees that as soon as the performance power envelope matches what Apple’s looking for, they’ll make that move from Intel.
… But that’s not Apple’s goal, catching up to CISC Intel and going RISC ARM on Mac. Not gonna happen. I already pointed out one reason why. Not ‘wrong again’. Watch.
Re: riots by developers
—> What if Swift did all the work for the developers in question? Click this button to compile for Intel. Click this one to compile for iOS.
I’m going to sit out going into the details of the software development business and explaining yet again the difference between CISC and RISC and how a simple recompilation can solve the transition, blahblahblah. What’s going on with this stupid rumor from hell is lots and lots of ignorance that I have zero hope of curing. I’m just pointing out that it is indeed a stupid rumor from hell. The future will continue to point this out without my aiding it further.
Precisely 🙂 they have a compiler which will allow them to compile to “target architecture”. Add to that a programming language that follows their specific rules that makes it easier. For example, using Swift I can create a simple “hello world” app for the Mac. Then I can, without changing any of the code that does “the work”, compile to iOS. That’s the same code effectively being one recompile from running on a different CPU. From CISC to RISC and back again. Can it be done? Absolutely. Will it be done? If Apple can create a processor with the performance envelope they’re looking for
Apple WOULD be better off using the SAME chip for ALL its stuff.
How can they do that? When?
Well maybe IBM will help.
Maybe IBM will help like it did with the PowerPC Alliance and the Taligent OS. I was talking last weekend to an IBMer who worked on both those projects. She was not hopeful that the new alliance is going to be any more successful in getting past the problem of big egos nurtured by very different corporate cultures.
“Apple WOULD be better off using the SAME chip for ALL its stuff.”
Why? Is Apple better off using the same UI for all its stuff? iOS on desktop, or Mac OSX on phone/tablet?
Obviously they decided they wouldn’t be, so despite some superficial similarities iOS and Mac are still distinctly different after 7 years.
Use the right tool for the right job. Microsoft is learning this the hard way right now.