ARM based MacBook Pro discussion [merged]

[leman]

That's what Windows RT - Windows running on ARM and that failed.

It wasn't the same Windows, thats the point. I don't know details as I never looked into WinRT in any depth, but I have read a lot of complains that APIS were missing or shifted.

Are you sure about that? Do you know that all source code is so easily compiled to a different platform?

If its only C/Objective-C/C++/Swift, without any binary code objection, CPU-specific intrinsics or inline assembly, then yes, I am very sure. As I said, the basic data aligments and sizes are the same for 64 bit ARM and 64 bit x86.

So you're saying ARM and Intel are essentially the same plarform

Sorry, poor choice of words. What I mean with 'platform' here is the OS contract. Same OS behaviour, same API.

 

[maflynn]

It wasn't the same Windows, thats the point. I don't know details as I never looked into WinRT in any depth, but I have read a lot of complains that APIS were missing or shifted.

You don't know the details, but you're saying its not the same thing. I think you need to know the details before saying its not the same thing

If its only C/Objective-C/C++/Swift, without any binary code objection, CPU-specific intrinsics or inline assembly, then yes, I am very sure. .

How many apps are there that are that meet that criteria?
Does MS office, or Adobe's CS suite or Carbon Copy Cloneer?

 

[leman]

You don't know the details, but you're saying its not the same thing. I think you need to know the details before saying its not the same thing

Already reading the wikipedia page about WinRT makes it very clear that it is a new API. You can't just take a standard windows app code and compile it for WinRT. You need to develop an app from ground up with WinRT in mind. Its like OS X and iOS, to put it a bit dramatically.

How many apps are there that are that meet that criteria?
Does MS office, or Adobe's CS suite or Carbon Copy Cloneer?

Why would any of those app use low-level CPU-specific stuff? Most problematic are going to be interpreters, JIT compilers and so on.

 

[mfram]

Having worked on a large software system professionally that went through an architecture switch, I can say that for the most part switching to ARM on the Mac would be a recompile for most apps. Drivers for hardware support would likely be a separate issue.
Interestingly from a developer standpoint one thing can occur. Bugs in your code can sometimes manifest themselves in different ways on different architectures. It's even possible for an app to "work" on one architecture but fail on another even though the same issue is present on both. It means more debugging work for the developers. Yes, the transition would take time. The biggest risk for users is that they are using a program where active development has stopped and therefore it won't ever be ported.
One big problem Apple is running into now with Macs is that they are held captive by Intel's schedules falling behind. Presumably if Apple was using ARM they would have more control over development schedules and we wouldn't have to wait 2 years for a new rMBP.

 

[Aquamite]

https://www.techtastic.nl/besturingssystemen/overstap-naar-arm-chips-voor-macbooks-komt-dichterbij/

It's coming!!!!

 

[keysofanxiety]

If you're like me, and your Dutch is a little rusty, here's the same article translated: https://translate.googleusercontent...erbij/&usg=ALkJrhgIQEHpNStdu3sYf86Ka0W5C2U5oA

 

[Aquamite]

What are the requirements to make this thread a stick thread?

 

[mabaker]

I'll be the first in line to buy this fanless beauty.

 

[Aquamite]

Why did they merge this? I just wanted to make another "waiting for" thread.

 

[Closingracer]

So with reports of the A10-Fusion matching the performance of the current Mac Pro (with caveats, of course), one can't help wonder if a desktop version running at a higher clock speed might match Intel's current performance. Whether it can today or not, it seems inevitable that it eventually will, which makes me wonder, what would the implications to the Mac platform be if Apple swapped out Intel for AX-chips?

I could see these advantages:
︎ Cost -- since Apple is already investing in the processor for the iPhone, it seems they would save money by not buying Intel chips. (And of course Apple wouldn't pocket the profit, they would pass it to us consumers.)

︎ Predictability -- the ability of Apple to release a new processor every year with the new iPhone on a predictable schedule I think helps the platform. Syncing Mac's update schedule with Intel's has not worked so well recently!

︎ Tuning -- the ability to tune the processor to the operating system would at least in theory lead to better performance or battery-life. Intel's chips have to satisfy a wide-variety of applications which must have tradeoffs for silicon space, heat, etc. Apple could focus on only what they needed.

︎ Security -- having a propriety chip would probably only help increase Mac security

Downsides:
︎ Compatibility -- they would probably need to emulate an Intel chip for a time to make software compatible. Although this seemed to go fairly well when Apple did it before with Rosetta.

︎ Bootcamp/Virtualization -- would we still be able to run Windows?


Interested in what other's think...

The A10 fusion doesn't match Intel cpu's.... Ghz is not a matching point. You would need to have Mac OS to compare performance.

 

[Omega Mac]

there is no doubt that controlling the stack makes for a very fast efficient system on the likes of iOS, but i don't see that translating across to a desktop OS yet. To me, iOS is just not there for desktop performance and the Ax is too far away from running MacOS.

No matter how they put across the iPad Pro abilities, compared to desktop/laptop tasks, its just not there for multiple windows etc etc, and I don't see it being there for a while yet. So the efficiencies of Ax chips and iOS won't transfer to desktop either for a while yet.

On the comments of intel chip cost ... is it really that much more expensive all in?

For example, how much would apple add to the cost of a desktop Ax chip as a result of development? and would that additional cost make the price similar to an intel chip. Would it be easier to then just buy an intel chip complete where intel has spent all the money on development?( i honestly don't know )

I don't understand this point. I have Mac's running on Intel core 2 duos, they run El Capitan fine (these a SL gen machines) and yet the Geekbench score of my iPhone 6s in far in excess of these machines in terms of raw processing power I am assuming. My point is that power is not an issue since Mac can run fine on much older and far less powerful Intel CPU's. (Maybe there is a technical aspect I am missing that means rather than things been apples vs apples we really are dealing with apples vs oranges)

The Ax Chips in a Mac as the previous poster pointed out will come in far cheaper for Mac, with greater control, if power is a problem they can beef it up by including for e.g. 4 A10 chips in a MacBook (16 cores!!! 8 high power and 8 low power). The format facto is so huge compared to a iPhone, they may be so liberate dint eh way they can design their internal configuration you might see totally new thinking in how machine are put together. I imagine that would be a huge attraction because then the build cost come down even more not just the cup inclusion.

The main issue at this point seems to be the Intel backward compatibility at this point.

'Course Intel will be fabbing these Mac Ax chips either way!
[doublepost=1475355117][/doublepost]

Unsure on switching costs, but no way does the A10 come anywhere near the cost that Intel charges for their CPU's.

There is absolutely no doubt Apple is running Macs in their product labs with A chips in them. Will they ever make it to market? Perhaps, if Intel continues to remain complacent due to little to no real competition.

Exactly. This is a huge motivation for Apple, money!

If you can reduce the amount of 3rd party components that makes your final package (because Apple are all about the whole final package) you can reduce your cost, increase your margin and maybe even reduce your retail price points and win on every level.

AMR'd Mac are not far away. Closer than many care to believe.

 

[Aquamite]

Just because you add more cores doesn't mean the performance is going to be better. Most applications only make use of one single core and most of the ones that are ready to use several cores normally have such latencies in core usage that there is just no gain in performance.

 

[294307]

http://www.iclarified.com/57138/apple-adds-arm-support-to-macos-sierra-kernel

An interesting read. It might also explain why developers cannot submit final binaries to the Mac App Store but instead intermediary bit code which is then compiled into binary code by Apple for the specific CPU architecture. This might address apps within the Mac App Store, but what about apps not on the Mac App Store?

I'll be very surprised if Apple announces something about this before the end of the year, but there is certainly evidence to suggest Apple might be moving forward with its plans for an ARM-based Mac of some kind in the near future.

With the kind of benchmarks Apple's new ARM processors get, I think a device like the MacBook could perform well with a high-end Apple A-series processor inside.

 

[Blue Sun]

http://www.iclarified.com/57138/apple-adds-arm-support-to-macos-sierra-kernel

An interesting read. It might also explain why developers cannot submit final binaries to the Mac App Store but instead intermediary bit code which is then compiled into binary code by Apple for the specific CPU architecture. This might address apps within the Mac App Store, but what about apps not on the Mac App Store?

I'll be very surprised if Apple announces something about this before the end of the year, but there is certainly evidence to suggest Apple might be moving forward with its plans for an ARM-based Mac of some kind in the near future.

With the kind of benchmarks Apple's new ARM processors get, I think a device like the MacBook could perform well with a high-end Apple A-series processor inside.

As long as we don't see a hit on performance, bring it on!

 

[h9826790]

May be a touch screen OSX on the iPad, not necessary bring the ARM to Mac.

 

[bmac89]

I don't know much about these things, but assuming this happened, how big of an impact would this have on developing applications for Mac and running older applications? Would it be similar to the transiston from PPC - Intel with Rosetta?

 

[maflynn]

Just because you add more cores doesn't mean the performance is going to be better.

Nope people easily fall into that mistake, its like using 9 woman to make a baby in one month - it just not going to happen

(16 cores!!! 8 high power and 8 low power).

That sounds great on paper, but for real world apps, that won't really impact performance. The program needs to be written to use multiple threads, and even then, its not always easy or logical, and the performance increase is negligible.
[doublepost=1475404612][/doublepost]

Why did they merge this? I just wanted to make another "waiting for" thread.

There was no need for a waiting thread, as there was an existing ARM/A10 discussion thread.

 

[Aquamite]

There was no need for a waiting thread, as there was an existing ARM/A10 discussion thread.

But waiting threads are tradition.

 

[SoyCapitanSoyCapitan]

For the MacBook it would be ideal because the A series is better than anything Intel or AMD can offer in that form factor.

 

[294307]

I don't know much about these things, but assuming this happened, how big of an impact would this have on developing applications for Mac and running older applications? Would it be similar to the transiston from PPC - Intel with Rosetta?

That's the question I have as well. It appears Mac App Store apps would "just work" as Apple can compile them for multiple CPU architectures without requiring the app developer to intervene but I do wonder how this will affect the rest of the apps that aren't in the Mac App Store. About half of the apps that I use on my Mac are not on the Mac App Store.

It certainly won't be a complete transition away from Intel to ARM, and may never be. An ARM-based Mac might work for a thin and light MacBook, but no current ARM chip could deliver the performance needed for an iMac, MacBook Pro or Mac Pro.

This article is another good read on this topic. Intel x86 architecture needs to contain legacy instruction sets for backwards compatibility which requires it to have more transistors than an ARM-based processor, which increases the amount of power and heat it generates. This might also be another reason why Apple is pursuing the possibility of ARM-based Macs - aside from controlling the primary technology, which they've long known to be in favour of.

 

[dogslobber]

http://www.iclarified.com/57138/apple-adds-arm-support-to-macos-sierra-kernel

This pretty much nails it, folks. Quad core ARM in an Apple TV firm factor running Sierra.

Note that you CAN run Windows 10 on ARM so boot camp support is still there.

 

[294307]

I think it will be a MacBook or some other thin and light laptop that will have ARM based processors first. If you are going all in with ARM processors, what's the point in putting it into a desktop first, when the bigger benefits to consumers will be seen in something that needs to be portable and last a full day?

 

[dogslobber]

I think it will be a MacBook or some other thin and light laptop that will have ARM based processors first. If you are going all in with ARM processors, what's the point in putting it into a desktop first, when the bigger benefits to consumers will be seen in something that needs to be portable and last a full day?

Why does the Mini use laptop parts? Exactly the same reason it's gonna be ARM based.

 

[Shirasaki]

Even though arm processor dominates computing industry, switching away from x86 is just as costly as building up the entire ecosystem. Switch cost is too high. Countless apps need to be rewritten and Windows should be ready for arm processor. Oh, don't tell me Windows RT.
in such a closed system like Apple, switching to arm may be possible but fully transition is still a quite costly process.
(Sounds a bit less logical)

 

[poorcody]

Just because you add more cores doesn't mean the performance is going to be better.

True, in fact at some point the overhead of cores can actually slow things down. Although, in practical terms, you could probably make decent use of say an 8-core processor at least on a desktop. Remember you're not just providing threads to one running application, but to the many processes the OS is running. (The speed of shuffling data in and out of the processor becomes the main issue quickly, although separate caches for each core can help with that, sometimes significantly.)

There's something called "Amdahl's Law" that tries to quantify cores/processors versus performance. Here is a graph of what they came up with:

That's from: http://www.extremetech.com/computin...m-one-core-to-many-and-why-were-still-stuck/2