Apple Expected to Announce Arm-Based Mac Plans at WWDC, Transition Away From Intel to Begin in 2021

[Sydde]

How is this situation different?

Intel has been seen as the de rigeur architecture for development, primarily because of its prevalence in the MS systems. When Apple moved to Intel, it seemed like a lot of people were saying "well, finally", and assuming that this was the endpoint, that Apple would be foolish to move to some other architecture, of which nothing appeared to be competitive, at least at the time. The landscape has changed somewhat since then, but Intel still holds onto a lot of hearts and minds, so moving away from it seems a bit risky, as one would be sacrificing BootCamp, Parallels, et al.

But ARM is quite strong in the mobile sector, which has grown significantly more than the computer sector, and ARM keeps getting better while Intel and AMD have to work pretty hard to keep their Moores-millennia-old architecture competitive. If MS can get their stuff properly ported to ARMv8 (64-bit), Intel could be in a world of hurt, and Apple would, at least for the decade, be seen as pulling the leading edge forward.

 

[1221320]

Funnily enough, as a life-long Mac user who shifted to Windows a couple of years ago after my MBP died and I couldn't use the butterfly keyboard in their new models, just as they fix that keyboard problem and I was considering going back, they're going to change the CPU! I use Windows a lot, and was quite happy with Parallels. But if that isn't going to work with ARM, there's no way I can transition back to Apple.

I suspect there is a sizeable slice of the userbase who will be in the same position.

 

[Unregistered 4U]

They are doing it because they will control everything and make more money that's it.

Let’s not forget the fact that Intel has been WOEFULLY bad at it for the last few years.

 

[1221320]

Let’s not forget the fact that Intel has been WOEFULLY bad at it for the last few years.

But AMD are going strong. Lots of x86 manufacturers are shifting from Intel to AMD, or at least offering AMD options (Lenovo, Asus, etc).

 

[cmaier]

Intel has been seen as the de rigeur architecture for development, primarily because of its prevalence in the MS systems. When Apple moved to Intel, it seemed like a lot of people were saying "well, finally", and assuming that this was the endpoint, that Apple would be foolish to move to some other architecture, of which nothing appeared to be competitive, at least at the time. The landscape has changed somewhat since then, but Intel still holds onto a lot of hearts and minds, so moving away from it seems a bit risky, as one would be sacrificing BootCamp, Parallels, et al.

But ARM is quite strong in the mobile sector, which has grown significantly more than the computer sector, and ARM keeps getting better while Intel and AMD have to work pretty hard to keep their Moores-millennia-old architecture competitive. If MS can get their stuff properly ported to ARMv8 (64-bit), Intel could be in a world of hurt, and Apple would, at least for the decade, be seen as pulling the leading edge forward.

None of this supports the hypothesis that Apple should offer any sort of board swap.
[automerge]1592400769[/automerge]

But AMD are going strong. Lots of x86 manufacturers are shifting from Intel to AMD, or at least offering AMD options (Lenovo, Asus, etc).

They did that in 1996, too. How long did that last?
[automerge]1592400865[/automerge]

Funnily enough, as a life-long Mac user who shifted to Windows a couple of years ago after my MBP died and I couldn't use the butterfly keyboard in their new models, just as they fix that keyboard problem and I was considering going back, they're going to change the CPU! I use Windows a lot, and was quite happy with Parallels. But if that isn't going to work with ARM, there's no way I can transition back to Apple.

I suspect there is a sizeable slice of the userbase who will be in the same position.

Not very sizable. Most people who need windows use PCs. Most people who use Macs don’t use windows. 98 percent don’t install boot camp (according to published statistics). Some may use parallels or VMware, but it’s not going to be a sizable percentage.

 

[Richdmoore]

They could also implement some sort of logic board exchange program that would allow Intel Mac owners to switch their hardware to ARM. They have not done that in the past, but this situation seems somewhat different from previous architecture transitions.

There was an iie to iigs upgrade kit way back in the day. It was so expensive, and you still had to replace the monitor, it mostly wasn’t worth it. (Those original converted systems are sought be collectors nowadays.)

 

[theorist9]

Raspberry Pi 4 uses a Cortex-A72 processor at 1.5Ghz (just over half the speed of your i7); the current iteration, 4 years later, is Cortex A78, which, at a minimum, is the architecture that Apple would be putting into ARM Macs. So, the comparison is probably a little flawed. ARM architecture appears to be advancing somewhat faster than Intel's.

Sure, and not only that, but Apple's implementation of ARM has many custom enhancements over "stock" ARM chips. Given this, it would certainly be better if I had a Mathematica benchmark on, for instance, the current iPad. I don't have that, so the best I could do was to say, essentially, "here's the closest I can come with the data that's publicly available thus far".

Maybe Apple will release a Mathematica benchmark for their future ARM-based Mac at WWDC. Of course, even that needs to be taken with a grain of salt:

When Apple introduced the dual-2.0 GHz G5 PowerPC at WWDC 2003, Steve Jobs claimed that Mathematica ("Wolfram Research") was 2.3 times faster on the PPC than on the "fastest PC [chip] money can buy—a dual 3.06 Xeon". [see:

]

However, when I later ran Mathematica timing tests myself, I found that that Mathematica was typically 20% - 40% slower on my G5 than on my plain-vanilla PC (2.8 GHz Pentium IV). When I spoke to a contact at Wolfram about this, he said the performance comparison was based on "a pretty specific function, large integer multiplication", where the G5 is faster.

IIRC, generally speaking, the PPC was competitive with Intel CPU's for integer calculations, but not for floating point calcs.

I.e., Apple cherry-picked the data. That's why you can rarely trust benchmarks run by someone with an economic interest in the product.

 

[jdb8167]

I hated the first Intel Macbooks (Pro), so maybe its the same with the first ARM Macbook.

My first one was DOA. The second one had the worst coil whine I had ever heard. I think battery life was on the order of 2 hours. It was fast for a laptop though.

 

[pankajdoharey]

Many problems: First in regards to VMs, if you have to emulate actual CPU you need to run the client OS you can just throw any/all hopes of decent performance out the window.

Your point would be valid if one has to actually emulate the x86 CPU on ARM, but AArch64 too has Hypervisor extensions which means you can run a ARM Linux or ARM Windows at full Speed. As far as Applications are concerned, Microsoft has already shown the Universal Windows Platform (UWP) apps are excellent and run very fast through wow64 binary translation layer. The only thing it cannot use is x86 drivers which is fine. If you thinking about Docker images then Docker has already announced support for Multiarch images. Which would run on both x86 and ARM. And if you are thinking about x86 compiled apps for Macs then I think a similar program to Mac Universal image would be launched. When Apple Transitioned from PowerPC to Intel they used Rosetta Stone technology for binary translation of powerpc apps for x86. And if you have used the universal apps you would know those apps run fast even games, so if they were being translated was hard to determine. Same thing can happen again in this transition period.

 

[PostApple2016BrainWash]

It is not just an announcement of ARM support for macOS but Apple needs to demonstrate the performance and battery life of A-Series SoC that can offer genuine improvements over Intel CPU.

 

[Falhófnir]

Once the transition begins, anyone want to predict how long it will take to:
1. Transition the entire Mac line to A processors?
2. Stop issuing new MacOS version updates (other than security/minor updates) for Intel machines?

I could see a longer transition than the PPC to Intel era, due to apple updating the Mac Pro last year, but honestly support for non-consumer focused models probably isn’t the best benchmark.

I guess Apple could also just do a “clean break” as well, like they did with the Apple II to Macintosh transition (and continue to sell both for awhile.). That would be way riskier, and more confusing to the consumers.

Think the entire lineup will have an Arm replacement within 18 months of the announcement, but that a select few Intel machines might stay on sale quietly for a little bit longer, maybe up to ~36 months from now?

I'm also inclined to think all Intel machines will get 5-7 years of support in one form or another, there's too many insanely expensive Intel machines for Apple to just turn around and drop them after 2-3 years. The question is more, will they get full macOS updates for years to come, or will they get up to say 10.18 and all get left there, but Apple keeps supporting 10.18 with security patches and minor updates for years afterwards. I think the latter option actually has benefits in itself - if most developer support has gone to Arm, then not having continuous yearly updates which often break things will help keep x86 programmes working after the developers pull the plug or lose interest.

 

[JPNMac]

Here's a quick timeline of Apple's archtecture transitions along with the first Macs with each new architecture:

* 1984: Macintosh 128K, 68000-based
* 1994: Power Macintosh 6100, PowerPC 601-based
* 2006: First Intel iMac, Intel Core Duo-based
* ~2021: TBD, ARM-based

As you can see, the time is getting longer and longer before Apple transitions to a new processor. It took 10 years to transition to PowerPC, then after that it took 12 years to transition to Intel, then after that it will take ~15 years for Apple to transition from Intel to ARM.

 

[theorist9]

Not very sizable. Most people who need windows use PCs. Most people who use Macs don’t use windows. 98 percent don’t install boot camp (according to published statistics). Some may use parallels or VMware, but it’s not going to be a sizable percentage.

Perhaps this wasn't your intent, but it seems you're portraying Parallels/VMware as an afterthought — small even relative to Bootcamp.

In fact, it's worth noting that Parallels represents many more of the Win-on-Mac users than Bootcamp. Parallells says >6M of their customers are using Windows on the Mac (https://www.parallels.com/blogs/parallels-desktop-users/). VMWare says it has 500k customers (though those might not be all VMFusion), and then there's Virtual Box, which is free, and thus could have a decent no. of users for that reason. Adding in the 2M bootcamp users (2% of the ~100M Macs currently running), while at the same time recognizing that some people use both Parallels and Bootcamp, that gives roughly 10M people using Win-on-Mac, so call it roughly 10%. Still small in absolute terms, but much larger than 2%.

But two caveats: Maybe Parallels will run on Arm, thus obviating the problem. According to https://appleinsider.com/articles/2...-arm-mac-is-going-to-be-a-bumpy-road-for-some , "It's not known yet whether a version of Parallels will run on ARM."

Second, the % that *need* to run Windows on their Macs (i.e., those for whom not being able to run Win software is a deal-breaker) is going to be some fraction of that 10%. I have no idea what that fraction would be.

 

[makman]

Does it make financial sense for Apple to go it alone with their own high-performance silicon (CPU and GPU) to supply a desktop/laptop market share that sits at 10% of the PC market, and may go lower when people who require Windows compatibility are forced to switch? Intel and AMD's R&D costs are paid for by much larger customer bases that include the server market as well, and the architectures are going to be getting even more complex going forward as Moore's Law limits are reached and additional performance must be obtained through architecture, not process node. Yes, Apple has a ton of money but from a business unit perspective, I wonder if the numbers add up.

 

[k27]

The life expectancy of Macs could be shortened by an ARM switch. At a certain point Apple will no longer support certain Mac models with macOS (depending on the model, this may happen sooner or later). These are mostly business decisions and not technical decisions. (See dosdude or Hackintosh bootloader on Apple hardware, like old Mac Pro).
The support of older macOS versions will be discontinued. The hardware (Mac) itself may still be fully functional and fast enough. But you will not receive security updates for macOS anymore. The workaround is to install Linux and Windows. Both are supported much longer. The life of the Mac hardware will be extended by Windows or Linux. Whether this will still work with ARM Macs is unclear:
- Boot loader / Security
- Driver for Apple's own SoCs
I myself have never installed Windows or Linux natively on a Mac (I have always had Mac models that were supported for quite a long time or I sold them in time before the support ended). But I would do it in such situations before using a macOS without security updates.
If and how strong these concerns are, we will see.

 

[Atlantico]

The papers i cited have ”cmaier” in the list of authors, dude. Again, whatever.

You can be who you want to be, I don't care. However, don't try to push any appeal to authority on me again, when pushing your opinions.

(Ps: i am no longer a CPU designer. I retired in 2007)

So you're conceding you're hopelessly out of the loop. But you would know that there's no such thing as a free lunch in electric engineering and ARM is not a magic device. Instead you shove your opinion down our throats based on an imaginary CV.

You consistently use the wrong technical terms, you demonstrate only the most superficial knowledge on the topic at hand and you can't make any argument that isn't based on wishful thinking or fabulous optimism.

Clearly, out of the loop. So stuff your appeal to authority. Thanks.

Interesting. What we have to believe is one of two things:

1) 13 years ago, someone decided to create an account using the name of a recently retired AMD chip architect in order to one day be able to pull rank on you.

2) Someone who uses a pseudonym on here is actually Jim Keller, but cannot really provide any details that are not public information.

Wonder which one of these two to believe.

You'd be a fool to believe either one, tbh.

 

[cmaier]

Does it make financial sense for Apple to go it alone with their own high-performance silicon (CPU and GPU) to supply a desktop/laptop market share that sits at 10% of the PC market, and may go lower when people who require Windows compatibility are forced to switch? Intel and AMD's R&D costs are paid for by much larger customer bases that include the server market as well, and the architectures are going to be getting even more complex going forward as Moore's Law limits are reached and additional performance must be obtained through architecture, not process node. Yes, Apple has a ton of money but from a business unit perspective, I wonder if the numbers add up.

1) the same cores they design for macs can also be used on ipads and probably iphones. So they essentially get the mac chip design on-the-cheap. Every R&D hour spent on design of most of these chips can be spread across the massive numbers of chips that will be installed in every iphone, ipad and mac, which is a very large number indeed.

2) Arm will increase Mac market share, not decrease it. It will enable better Macs, better interoperability with the massive iOS/iPadOS software base, new form factors and capabilities that differentiate from cheap windows machines.

3) When I started designing CPUs in 1992 we were told we were just about at the limit of Moore’s Law - maybe 3 years away. I’ve heard that every year since. (And Moore’s Law isn’t so much the concern - we were worried we were at the physical limit of lithography and there are only a few hundred atoms now, etc. etc. ). We always found a way through it.

 

[firewood]

In fact, it's worth noting that Parallels represents many more of the Win-on-Mac users than Bootcamp. Parallells says >6M of their customers are using Windows on the Mac (https://www.parallels.com/blogs/parallels-desktop-users/). VMWare says it has 500k customers (though those might not be all VMFusion), and then there's Virtual Box, which is free, and thus could have a decent no. of users for that reason. Adding in the 2M bootcamp users (2% of the ~100M Macs currently running), while at the same time recognizing that some people use both Parallels and Bootcamp, that gives roughly 10M people using Win-on-Mac, so call it roughly 10%. Still small in absolute terms, but much larger than 2%.

Add to that, the current fad in Machine Learning, where tons of students are using Docker images with x86 TensorFlow or PyTorch, etc.

To compete with that, I wonder if Apple will be adding some ML training processors to their A chips to complement the ML inference Neural Engine in the A12?

 

[Falhófnir]

Does it make financial sense for Apple to go it alone with their own high-performance silicon (CPU and GPU) to supply a desktop/laptop market share that sits at 10% of the PC market, and may go lower when people who require Windows compatibility are forced to switch? Intel and AMD's R&D costs are paid for by much larger customer bases that include the server market as well, and the architectures are going to be getting even more complex going forward as Moore's Law limits are reached and additional performance must be obtained through architecture, not process node. Yes, Apple has a ton of money but from a business unit perspective, I wonder if the numbers add up.

Evidently it does or they wouldn't be doing it

 

[cmaier]

Add to that, the current fad in Machine Learning, where tons of students are using Docker images with x86 TensorFlow or PyTorch, etc.

To compete with that, I wonder if Apple will be adding some ML training processors to their A chips to complement the ML inference Neural Engine in the A12?

I think they will. I think they will market the mac as the best place to create AI apps and datasets.

 

[09872738]

I think they will. I think they will market the mac as the best place to create AI apps and datasets.

Maybe, but that‘d mean they’d compete with all the cloud computing behemoths. Not an easy or quick thing for sure

 

[cmaier]

Maybe, but that‘d mean they’d compete with all the cloud computing behemoths. Not an easy or quick thing for sure

Apple’s not always good at competing (particularly where they really don’t understand a market), but they’re never afraid of competing.

And in this case I think they do probably understand the market, because they do so much with ML themselves. And they have a very excellent silicon design team, plus the market is comparatively young so it’s a good time to get in there.

 

[09872738]

Apple’s not always good at competing (particularly where they really don’t understand a market), but they’re never afraid of competing.

And in this case I think they do probably understand the market, because they do so much with ML themselves. And they have a very excellent silicon design team, plus the market is comparatively young so it’s a good time to get in there.

Not that I disagree; however for now they totally screwed it up. ML/AI is an almost CUDA exclusive, mostly cloud based business right now. Abandoning Nvidia put them pretty much out of AI business (apart from iOS based apps, which are for now little more than PoC toys), and I cannot see how they‘d come back anytime soon.
Of course, Apple‘s got the money and stamina if required.
Still, they compete with established frameworks like TensorFlow or PyTorch, all practically Nvidia exclusive. I also doubt AWS, Azure or GAE are interested in supporting Apple, quite the opposite. Hence Apple is fighting an uphill battle right now in that regard

 

[firewood]

Maybe, but that‘d mean they’d compete with all the cloud computing behemoths. Not an easy or quick thing for sure

Because of Nvidia's cloud provider pricing, a lot of smaller ML researchers (e.g. not Google or Apple) are finding it cheaper to build "gamer" PCs with a couple 1080 Super GPUs (et.al.) than to buy cycles from AWS (et.al.) for training. It would certainly be interesting if Apple decided to compete with GPU cards and TPUs in this ML area. Apple certainly has work to do to help Siri catch up with the competition.

 

[theorist9]

In June 2005, to facilitate the PPC->Intel transition, Apple offered (for $999) an Intel-equipped Developer Transition System to select developers. It was contained inside a PowerMac G5 case, and ran OS X 10.4.

[Seven months later, in Jan 2006, when the first Intel Macs were released—consisting of the first 15" MBP, and a new iMac line—they offered to swap these systems for iMacs.]

I wonder if they'll be doing the same this time, except offering developers, say, Arm-equipped Mac minis.