Intel Expecting Apple to Transition to Custom ARM-Based Chips Starting in 2020

[pankajdoharey]

Very Positive move in my view, to address some issues :

For Bootcamp, there is already a Windows ARM 64 that Microsoft gave a preview of some yrs back.
For Mac Apps - Dont worry, Apple has a good strategy they will utilize the same technology they utilize in watch apps the Watch apps are essentially llvm bitcode that gets JIT'ed when you run them, this will make your apps highly portable it wont be a heart burn like Rosetta days, though i actually liked the rosetta apps aswell they were very fast compared to standard emulation. Both the strategies would work.

 

[MyopicPaideia]

Interesting. When taken together with the 2020 TSMC 5nm architecture article, it makes sense.

What Apple have been able to do here is achieve an app library in advance of the migration gradually through iOS. All the UI SDKs and stuff, single/fat binaries...I mean, all the compatibility is already there, all of Microsoft’s apps are already on iOS, Adobe is putting their bread and butter apps on as well...it is a matter of giving them parity with macOS/Windows versions...meaning simple feature updates. There is no need for a next generation Rosetta because the app library on iOS is already enormous. I’m sure one of the VM companies like virtual box, Parallels, or VMWare will be quick with an ARM product allowing x86 native OS’es like windows or Linux to run very well.

 

[Thomas2006]

Looking at the beast that is the A12X in the iPad Pro, I can see why.

As a customer looking to buy an Apple desktop this year, I'm really torn on what to do now. Is the lifespan of my Intel-based Mac going to be shortened by this? Seems likely. I also wonder if the transition will start with portables or desktops? You would think with a major modular redesign of the Mac Pro this year that it will be supported for a long time, but now I'm not sure.

I would say the MacBook will be the portable getting the A-series processor and the Mac mini getting both processors. Obviously, the internals will be a bit different but the case and ports will be the same. If Apple puts the A-series processor into a smaller case then I see the Mac mini getting renamed to 'Mac' with the new model getting the 'Mac mini' name.

 

[MyopicPaideia]

Having lived through the Motorola to Intel migration I have some observations. First off, a *lot* of the software I owned became abandoned shortly afterwards as (especially small) developers couldn't/wouldn't justify the burden of rewriting all their software. For those that could I ended up having the privilege of rebuying, at full price, every software package I owned and for those that couldn't I ended up having the privilege of seeking out and buying, at full price, alternative software (with the accompanying aggravation of converting over to it). It was a monumental disruption and from memory all the software upgrades/purchases alone cost more than the new machine. The one silver lining was that the (virtualized) PC software that I needed to run could now be done in a much faster and straightforward manner because the CPU was Intel.

So here we go again, only this time my virtualized capabilities will suffer...

Nope, because virtually all the software has iOS/ARM counterparts already. This is the difference this time. A developer simply will have to do feature updates to give parity between the iOS/ARM version and the legacy x86 version of the app.

 

[pankajdoharey]

For this to be acceptable, Microsoft would have to offer Windows 10 ARM retail and only support pure .NET even on x86. Oracle would have to provide ARM Java.

Both Java and .Net Core on Arm Already exists.
https://www.oracle.com/technetwork/java/javase/downloads/index.html
https://github.com/dotnet/announcements/issues/29
[doublepost=1550931736][/doublepost]

The A12 is already as fast as an i5. By 2020 it will make an i7 look like an 8086.

Also note that the current line of Apple chips are optimized for iPhones and iPads. A notebook or desktop version would probably smoke a base i7 today.

A12X nearly matches i7 performance already
https://www.extremetech.com/mobile/...-nearly-matches-top-end-x86-cpus-in-geekbench

 

[cube]

Both Java and .Net Core on Arm Already exists.

.NET Core is not enough, and you can still run native programs on x86.

 

[LordVic]

Both Java and .Net Core on Arm Already exists.
https://www.oracle.com/technetwork/java/javase/downloads/index.html
https://github.com/dotnet/announcements/issues/29
[doublepost=1550931736][/doublepost]
A12X nearly matches i7 performance already
https://www.extremetech.com/mobile/...-nearly-matches-top-end-x86-cpus-in-geekbench

right from the very top of the article you linked...

When you actually look at specific test results, there are areas where the Intel CPU is dramatically faster than the ARM equivalent.

So nothing new in the ARM world. even the A12x. it excells at some specific tasks, but for the majority of tasks, no, the A12x is NOT as fast as an i7.

I think the A series "fast enough" for a very large swath of users. But It needs to be clear that for some, no, these CPU's are not ready to just swap into our desktops and we'd be happy with their performance.

First, those benchmarks are for the "U" variant of the i7. which is a low wattage part. So while possibly comparable for lower end notebooks / ultrabooks, anything that uses a higher watt part, the larger MBPro's, or the desktops, would still outperform the CPU in the iPad by a large margin.

For the benchmarks attached in your very link, the following tasks performed dramatically faster on the i7 MBPro (13") than the iPad's A12x. These are specific things I believe normal users will be impacted by and not just "niche" things.

So far the i7 still significantly dominates in:
JPEG, LZMA, PDF Rendering, HDR, Gaussian blu, speech recognition, memory latency. there are others, but are more complex and not really user focused.

either way, don't confuse weighted combined scores that favour certain tasks in a benchmark as accurate representation of performance based on usage. this is one of those places where "on paper" something appears one way, but in actual reality, it's a completely different story.


Don't take this "rant" out of context though. What Apple has currently with their ARM designs is the best in the industry. There could be a logical and good way to do this migration. But to make claims that today their CPU's out perform those that they're slated to replace is unfortunately not quite true yet.

 

[Lankyman]

Funny: Because from what I have heard/seen, outside of Surface commercials on TV, VERY few people actually USE their Touchscreens for anything more than the occasional "tap" (and that, mostly because the trackpads on ALL Windows laptops SUUUUCK!).

I think your last sentence should read ALL trackpads suck - period. I tried Apples offering which was quickly returned for a refund.

Give me a Logitech mouse any day of the week.

 

[Freida]

Why will it be a mess for a few years? Because the past TWO Architecture changes were?

Oh, wait...

Not quite sure why you are so 'aggressive' but maybe its just misleading in the text.
Yes, those changes were a pain and so will be the next one. You can't switch architecture or system without hiccups. We are not in a perfect world.
As I said before, most developers will be fine and will catch up but some will decide not to port their app due to cost, time etc. and will be out. We can see something similar - as an example - Zbrush is perfect app for iPad and yet its not available for iPad. The developer just simply chose not to support that platform even though the iPad screams for it.
And Pixologic has money, imagine small developers. Simply put, transition will have its victims and won't be super smooth. That was my point.

And as you brought up the previous ones - you don't remember how slow rosetta was?

 

[Lankyman]

iMac is poorly engineered piece, it has always been. CPU options they give you are just a marketing trick cause all of them go down throttled and are their sustainable performance is crippled by tight chassis, improper cooling, PCB shortcomings, etc. Same spec CPU and GPU in some normal Windows laptop running resource hungry Windows 10 would give you better average performance.

Is that the fault of the iMac or is it a case of the user buying the wrong tool for the job?

 

[cmaier]

Linus Torvalds seems to understand this (and the article even uses Steve Jobs to back him up):

https://www.theregister.co.uk/2019/02/23/linus_torvalds_arm_x86_servers/

The reference to Steve Jobs was stupid. Linus is saying developers want x86 cloud servers because they have x86 desktops.

Steve was saying people should code native to iPhoneOS and should use flash because native is better.

These are completely unrelated principals.

And neither suggests x86 is better - if people had ARM desktops Linus would demand ARM cloud servers.

 

[Unregistered 4U]

I can't imagine Apple not taking this into account when the switch is made.

I’m sure it’s been taken into account, and Apple has a pretty good idea of how many systems out there are booting into Windows. Of all the computers sold every day, it’s likely a percent of a percent nowadays.

BootCamp was more of a byproduct of running on an Intel standard CPU and motherboard than anything else. There are people that need it, and those people just won’t buy the new ARM systems (many are likely already running years old systems, so they’re already not really the cutting edge upgraders).

 

[Glockworkorange]

I’m sure it’s been taken into account, and Apple has a pretty good idea of how many systems out there are booting into Windows. Of all the computers sold every day, it’s likely a percent of a percent nowadays.

BootCamp was more of a byproduct of running on an Intel standard CPU and motherboard than anything else. There are people that need it, and those people just won’t buy the new ARM systems (many are likely already running years old systems, so they’re already not really the cutting edge upgraders).

You're probably right. I wonder how they can tell which users also fire up Windows? Survey? Analysis of Parallels data? I can't imagine they can determine which machines have run Boot Camp. They're clearly measuring this, I'm jut not sure how.

 

[theluggage]

Boot Camp is simply a set of Drivers to bridge Apple Hardware and Windows' I/O APIs.

...which only works because the Mac is basically a PC clone with UEFI firmware that is partly compatible with the firmware in modern PCs (the original BootCamp included the BIOS emulation layer for pre-U EFI). There's a lot more to it than just having the same processor instruction set - which is all we can currently say about a hypothetical ARM-based Mac.

Not saying that dual-booting or virtualisation is impossible on ARM - the issue is having a compatible OS that you can run. You're not going to be able to stick Windows/ARM, Ubuntu/ARM, Android or RISC-OS on your ARMintosh until/unless someone finds it worth their while to build a compatible distribution. Of course, the fact that those operating systems already exist for ARM makes that far more feasible, but its not a certainty. Ultimately, it all comes down to whether Apple thinks that dual-booting/virtualization is a priority.

Not sure its a big issue unless Apple make the transition too quickly, though: Easy virtualisation was one of the reasons I switched to Mac in 2006 and its been invaluable over the years, but I find I'm using it less and less. Circa 2006 I was still worrying about testing websites on Internet Explorer 6 and testing apps on various versions of Windows, so Parallels was invaluable. Now, IE pre-about-version 10 is dead (ding! dong! hallelujah!), the newer versions are far more standards-compliant and the majority of Windows users have switched to Chrome anyway (even MS is switching to chromium). Even with an x86 Mac, if I get any more paid development work I'll probably get a Surface Pro for testing, because the issues arising now are things like supporting Windows touch screens, docking/undocking tablet sections and mixed-resolution multiple displays. Increasingly, testing under Parallels has thrown up VM-related issues (electron - a runtime for node.js apps and used by things like VS Code doesn't/didn't work, for example).

I've also been tinkering with some stuff to generate waveforms/pulses/control-voltages on a USB audio adapter that is intended to run on an ARM-based Raspberry Pi. So, a bit more than 'Hello World' - but the code (in C) is 99.9% identical between Intel Mac and ARM RasPi because it uses an existing cross-platform audio library and POSIX multithreading - and the few differences are down to BSD vs Linux differences, not processor architecture. The real 'compatibility' issues are because the Pi was built down to a price using an existing off-the-shelf SOC that funnels all of the I/O through a single USB 2 port - if Apple is planning on selling an Armintosh for $35 then we may have a problem

 

[Unregistered 4U]

Yeah, I mean if they have it running FCP X and most of the Adobe apps out of the gate, I could see the Mac Pro being the transition Mac.

I agree. It’s a small group of total users that, if they can push the ARM to deal with their heavy workloads, then they can absolutely handle the web surfing and iMovies of the majority. Here’s another thought, Apple offers a range of systems with different processors because Intel produces a range of chips. What if Apple has highly performance ARM processors, and they put the same professional powered chips in everything? That would actually help to define why they always mention “modular” when talking about it. The real difference between the iMac and the MacPro will be how much modularity is possible... the raw compute power could be the same. Even the laptop form factor could end up with a level of performance you don’t normally see because Intel isn’t really driven to deliver the BEST performance across all lines, they have to deliver certain tiers so that they can charge a lot for the high end and the low end goes cheap. Does “good” “better” “best” go away?

I don’t think the industry will need much time. Apple’s already got most all the important developers using Xcode, so they know what code is being written, how it’s being used and how best to compile for Intel and ARM. I think after developers get their hand on the new Xcode (well, we KNOW some have it now, they’re just not talking), they’re going to find the transition too easy to NOT do.

 

[cmaier]

You're probably right. I wonder how they can tell which users also fire up Windows? Survey? Analysis of Parallels data? I can't imagine they can determine which machines have run Boot Camp. They're clearly measuring this, I'm jut not sure how.

Because they never see anyone with bootcamp at the Genius Bar?

 

[AppleScruff1]

You're probably right. I wonder how they can tell which users also fire up Windows? Survey? Analysis of Parallels data? I can't imagine they can determine which machines have run Boot Camp. They're clearly measuring this, I'm jut not sure how.

They can probably see everything people are doing if they so choose.

 

[Unregistered 4U]

You're probably right. I wonder how they can tell which users also fire up Windows? Survey? Analysis of Parallels data? I can't imagine they can determine which machines have run Boot Camp. They're clearly measuring this, I'm jut not sure how.

For Parallels, it’s going to be limited to whether or not the user shares data with Apple (many do). For BootCamp, once you boot into Windows, you have to install the custom drivers. Whenever Apple updates those drivers, all these BootCamp installations have to call home to get the updated drivers. If out of all the macOS systems in the world, 2 million are getting the updated Windows drivers, that would be a pretty good measure.

 

[DNichter]

Smart move by Apple. Their power per watt (the reason they switched to Intel in the first place) in the A series chips destroys where Intel is at. Give them a more space for heat and a larger power supply and they’ll easily outpace them for the next 10 years. Looking forward to some advanced form factors, but I’ll likely be sticking with iOS for everything.

 

[Unregistered 4U]

Interesting. When taken together with the 2020 TSMC 5nm architecture article, it makes sense.

What Apple have been able to do here is achieve an app library in advance of the migration gradually through iOS. All the UI SDKs and stuff, single/fat binaries...I mean, all the compatibility is already there, all of Microsoft’s apps are already on iOS, Adobe is putting their bread and butter apps on as well...it is a matter of giving them parity with macOS/Windows versions...meaning simple feature updates. There is no need for a next generation Rosetta because the app library on iOS is already enormous. I’m sure one of the VM companies like virtual box, Parallels, or VMWare will be quick with an ARM product allowing x86 native OS’es like windows or Linux to run very well.

I think developers are going to have a much easier time with this transition.

 

[theluggage]

The reference to Steve Jobs was stupid. Linus is saying developers want x86 cloud servers because they have x86 desktops.

...and even that is questionable, because people writing cloud apps are mostly writing in processor-independent Javascript/Typescript/Python/Ruby/PHP etc. and targeting server applications like Apache, Nginx, Node, MongoDB etc. that already run on ARM Linux and Mac and should therefore be trivial to build on MacOS/ARM.

Plus, cloud developers use this thing called the "cloud" that (for example) lets you spin up an x86 virtual server instance in Tokyo running the OS of your choice and test your code while sitting in a specialty vegan coffee shop in San Fransisco (and it will cost a lot less than your conflict free acorn latte).

 

[cmaier]

...and even that is questionable, because people writing cloud apps are mostly writing in processor-independent Javascript/Typescript/Python/Ruby/PHP etc. and targeting server applications like Apache, Nginx, Node, MongoDB etc. that already run on ARM Linux and Mac and should therefore be trivial to build on MacOS/ARM.

Plus, cloud developers use this thing called the "cloud" that (for example) lets you spin up an x86 virtual server instance in Tokyo running the OS of your choice and test your code while sitting in a specialty vegan coffee shop in San Fransisco (and it will cost a lot less than your conflict free acorn latte).

Linus isn’t exactly the most thoughtful guy...

 

[hans1972]

Performance per watt isn't what I'm basing any comparisons on, but I'm glad you feel that way. I like macOS, I don't like iOS-thats where I'll spend my money thank you. Best luck.

Just because Apple might switch to ARM, doesn't mean that it will come with iOS. macOS will just be recompiled for ARM and it till look exactly the same wether running on Intel or ARM.

 

[Mac Heretic]

Very Positive move in my view, to address some issues :
For Bootcamp, there is already a Windows ARM 64 that Microsoft gave a preview of some yrs back.
.

In that scenario, doesn't practically all existing Windows apps have to be rebuild for ARM by developers, because they won't run otherwise.

I recall that there were indeed some kind of ARM-based Microsoft Office applications, but there are already comparable Mac versions for those. Other than those, were there ever any important Windows ARM programs anyways?

Meaning, not so many will do that at least for free and one's old intel-based application purchases would become obsolete. If one would even think about doing the conversion, he would be better off to make an elegant ARM for Mac or Linux version instead than to target for random bootcamp or virtual os users.

I would suggest that a catalog of ARM based Android/Chromebook/Ubuntu/other Linux apps is a more likely scenario than a catalog of ARM based Windows apps. But even in that scenario, quality of applications may not be decent.

 

[theluggage]

Linus isn’t exactly the most thoughtful guy...

Well, he did kinda create the OS kernel that most ARM-based servers and phones rely on, so he's certainly worth listening to - and the irony is that its largely the success of Linux which makes the issue of what processor your Linux is running on less important.

However, he's a kernel developer so he's probably got a particular perspective on "development" that is rather closer to the bare metal than the much larger group of developers churning out CPU-independent C#, Javascript, Swift, Java, PHP etc. who don't really need to bother their heads with what processor they're using.

Also, he spent some time working at Transmeta on a new processor which really bet the farm on supporting x86 via binary translation. Not a huge success...

I'd say that he's right in that people who want x86 are gonna buy x86.