Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Everybody seems to be assuming this won't be for the high power systems like the Mac(Book)?Pro. I seem to recall somebody from ARM saying a while ago that in theory their designs, particularly the newer ones, could be good for servers. Presumably with some work somebody (like Apple) could make them suitable for high end workstations.

Could they have been thinking of this when they designed the new Mac Pro? I gather it can get pretty hot, using ARM might help mitigate that.
 
How would an ARM processor stack up to a Quad-Core i7 when doing something like compressing 1080p video or 4K video?
Thats a GPU task, generally.

To a certain extent yes. However you are assuming that Apple can't increase single core performance to address the current A7's performance. Remember the current cores only run at 1.3GHz or so.
You do the typical mistake of assuming that clock speed is the sole determinant of a processor's speed. Cyclone (and indeed all ARM CPUs) only attains its 'reasonable' speed if its workload is very narrow and very focused. Its branche prediction is vastly inadequate and inferior. This causes two problems:
1) It limits how well the CPU scales at higher frequencies. At higher frequencies, the prediction falls more and more behind - and the performance bottleneck gravitates towards its memory bandwidth. Intel devices are significantly better when working with larger datasets.
2) It limits how well the processor handles larger and more diverse workloads (which are most of your 'everyday' things). Geekbench is a very poor benchmark for a CPUs overall capacity because the benchmark itself is extremely small. A significant parts of the CPU is never actually tested in Geekbench.

There isn't any ARM CPU even close to currently matching one Intel's 'Big Core' x86. Forcing them to that performance results in dreadful power efficiency.
(Heck, even at the iPad speeds Haswell sports superior performance-per-watt, it just can't do extreme sleep levels as well)

For the price of a single intel xeon, you can add a substantial number of ARM cores.
You're making the assumption that the software actually scales with more cores. This is not the norm. Some workloads can be scaled quite easily, but a lot can't. And most workloads see an increasing amount of dependency collisions the more cores you shove in. Thats also before considering the increasing overhead cost on the CPU scheduler.

Indeed, one of Apple's advantages in their iOS devices is that they only have two cores. It benefits many types of workloads which cannot be parallelized (e.g. they can only run on 1/2 cores), and also provides a more consistent performance profile. It also makes it easier for the software developers, since they only need to keep two cores in mind - rather than 3 or 4. Its very easy to introduce non-obvious or downright weird bugs when parallelizing.

Its pretty much the reason AMD has completely failed in the x86 market: They over-emphasized many weak cores over few strong like Intel did.
 
Last edited:
Actually the semiconductor business was spun off as Freescale. Freescale actually has some nice products for the embedded world.

Not sure why everyone is ignoring the original quote - the guy said PowerPC was created by ibm and some small company. Maybe you're all born after 1995 or something, but I assure you that during the time when PowerPC was created, up through the time apple decided to adopt it, no one was calling moto a "small company."

His quote would be like saying "apple developed a digital camera with some small company" despite the fact that the company was kodak at the height of its glory, just because TODAY kodak barely exists.

Moto had 60,000 employees just a few years ago. It was by no means a small company.
 
There are over 1 million iOS and iPad apps that could easily get ported or boxed to run on an OS X system with a mouse and trackpad. It's far far easier to port a single/double touch app to a mouse/trackpad PC, than vice-versa. Most linux apps port to OS X, no matter what the CPU. Also, the vast majority of the apps in the Mac App store could be ported to ARM64 with just a few days work (mostly testing and re-optimization).

Not really. iOS devices have to largely fixed in screen sizes and dimensions. Those apps would have issues dealing with pragmatically arbitarily sized screens and multiple screens both of which are largely none existant in iOS.

It is far, far, far easity to port iOS n apps to iOS n+1 (or n+2).

It wasn't like apps could not be ported to Windows RT. They could. The Windows RT problem was far more so a chicken-and-egg situation. Are there going to be enough users to justify the port.

Additionally, the issue going to run into is similar to what RT faced. The legacy Windows ecosystem/API was much larger than RT. iOS is much larger than OS X. There is some movement to OS X from iOS but it isn't anywhere near 10,000's of apps let alone millions.


It also isn't just the ease of the initial port. The long term development costs are in running two stacks.

And X Windows on OS X ports aren't particularly a huge driver for OS X now. It is nice, but LInux apps with tightly mingle with specific widget sets aren't going to a huge driver.


Note that there may be more total developers these days who are currently developing personal apps for ARM systems, rather than x86 desktops/laptops. So Apple could jump to the future where most developers are going, leaving the dual GPU Mac Pro x86-64 for heavy lifting of legacy stuff.

There is little to no reason why future Macs from the rest of the line up can't target the current heavy lifting workload. Future Mac Pros would mov onto something else. A subset of the laptop and iMac line ups were dual GPUs long before the Mac Pro became so.

If Apple put XCode on certain ARM iOS system then the "need" for a large fraction of those "only interested in ARM" developers for an OS X system would drop significantly. What would be legacy for them at that point would be OS X. If Apple sold a far more affordable personal app development system a large segment of those folks would be stumbling over themselves to get it.
 
Since the rumor mentions a trackpad, I guess is that these devices would be running OS X with OS X apps, not iOS and iOS apps.

What could happen, is that it could run both.

Both re-compiled OS X app as well as iOS apps (modified for larger device displays, which may happen very soon :). Mouse-based apps present targets that are too small or hard to see on touch devices. But touch apps present easy to hit, well organized targets for a mouse or a trackpad, and are thus even easier to use by older people who need reading glasses (the biggest growth populace these days).

I run many of my own iOS apps in the Simulator on my MBA. Works great. Too bad developers don't make more of these builds available.
 
for the everyday soho user a reliable device with iCloud would be a win i think.
there are a bunch of people who buy dirt cheap pc-s and run (mostly) pirated versions of ms office/windows stuff, and in reality they only need:
- a web browser (which includes social media as well)
- an email client
- a way to create documents/spreadsheets
- instant messaging/facetime-like (video/audio) comms

and in the same time can't get around with tablets because of the lack of keyboard.

iCloud would be the best shot for them. i've been looking for a feasible solution to provide a small but usable home platform with just web access, and i also tried "chromeOS", but to have usable performance i really require a desktop class cpu, a proper OS to have a proper web browser. then why the fuss if i can run all them apps locally?

if apple can build a better chrome book with unlimited 3g/4g access(similar to amazons kindle) and wifi and icloud/imessage/facetime access only, it could be a win!

Apple already make the iPad for those people... the Mac is for other things. Introducing another processor would just make things complex and annoying.
 
Not to worry guys.

The ARM experiment will be restricted to MacBook Airs for casual users..for now.

Your MacBook Pro will run on Intel chips for another 10 years (well, start worrying now...lol).
 
I seem to recall somebody from ARM saying a while ago that in theory their designs, particularly the newer ones, could be good for servers. Presumably with some work somebody (like Apple) could make them suitable for high end workstations.

There are some states where over 1% of the entire electrical grid is reportedly going to data centers, much of it from fossil fuels. Many cloud applications are not CPU bound, thus ARM chips only have to get close, not beat x86 in pure performance. For this reason, there are several major players (a multitude that can often move faster at customization than Intel) looking to develop or use multi-core ARM64 server chips to reduce this energy/carbon footprint.

And developers and IT support for those systems might prefer the same architecture in their laptops. Apple could potentially lead this parade.
 
Let's fragment Macs along the lines of "these Macs can do Boot Camp, and these ones can't" and "These Macs can run Parallels/VMWare and these ones can't"
What's the difference between Boot Camp and Parallels/VMWare and which is better? Thx.
 
The ARM experiment will be restricted to MacBook Airs for casual users..for now.
.

And developers. Xcode and Eclipse will likely port to an ARM64 OS X quite easily. So will many power tools, such as Logic Pro.

Lots of kids are learning to code on the Rasberry Pi, a *nix ARM system. x86 assembly language means nothing to them.
 
It wasn't like apps could not be ported to Windows RT. They could. The Windows RT problem was far more so a chicken-and-egg situation. Are there going to be enough users to justify the port.

Additionally, the issue going to run into is similar to what RT faced. The legacy Windows ecosystem/API was much larger than RT. iOS is much larger than OS X. There is some movement to OS X from iOS but it isn't anywhere near 10,000's of apps let alone millions.


It also isn't just the ease of the initial port. The long term development costs are in running two stacks.

the Problem with rt was the dropping of the Win32 API for 3rd party app.
If Apple does OS X on ARM They certainly keep all common api, other than that there wold be no point in useind it instead of iOS.

Long term there is no point in keeping two stacks, no question. You know what thet means. no x86-64 and Performence orienteted Workflows get portet to CPU/Multicore and need to use the multithread agnostic programming language. There most APP can be parallelized, the ones that can't are mostly not that Important for the OS X Community the new MAC Pro shows where it goes tomorrow. As far as I recognised rarely one with modern SW complains about the new mac pro. Probably the Chinebench producers complain.
 
Everybody seems to be assuming this won't be for the high power systems like the Mac(Book)?Pro. I seem to recall somebody from ARM saying a while ago that in theory their designs, particularly the newer ones, could be good for servers.

ARM is only targeting a specific subset of servers. They are not targeting servers in general. The target they are aiming at are in the 'front end' cloud server ( e.g., generic web serve ) front ends ) and highly variable load servers. Think hosted virtual machines where the host company makes money by how many limited throughput 32 bit VMs they can pack onto some hardware. Or having "peaker load" servers than can rapidly spin up to handle load bubbles and then spin back down to extremely low power usage when the load evaporates.

Intel is now targeting the same market with their ATOM server offerings. All this has very little to do with where need full time computational horsepower. In short, in parts of server centers power efficiency was becoming a bigger selling point and ARM saw (and still sees ) an opportunity.


It used to be that folks used Core i3/i5/i7 or perhaps Xeon E3 or E5 class boxes at these kinds of workloads. Those doesn't work as well and data centers had very real power/cooling issues. Similarly some also had density limitations (since generic motherboards, chipsets , and designs didn't minimize space). That is only a subset of loads that servers in general run.

Back end databases and web services tend to have different issues where computation load and efficient large aggregation is much more an priority issue.



Presumably with some work somebody (like Apple) could make them suitable for high end workstations.

This kind of missing of the point of how Apple got to market first with a 64 ARM implementation ahead of others. Apple got there precisely because they were not looking to hit the sever market. The parts of the new ARM architecture that were highly effective for their "race to sleep" and more effiecient code ( dropping legacy ARM instructor set baggage) was what they were after. Expanded address space .... was entirely not the point. Better instruction set was. That it is tightly bundled to the address increase is only a minor side effect for Apple.

The folks who focused on when phones are going to get to over 4GB of RAM entirely missed the point.

Apple is going to roll out another ARM design team just for Macs that has to compete head-on-head with what Intel can do? That is kind of asking for hurt. That has major problems.

1. ARM isn't trying to compete with Intel's classic dedicated server offerings at all. So instead of leveraging ARMs basic R&D Apple is off by their lonesome.

2. The unit numbers of Mac is far smaller than iOS devices. The complexity of the SoC would like be higher to compete with Intel's offerings.... so that means a bigger design team for a smaller market. That's one reason why PowerPC had long term problems. Nobody, including and primarily Apple, wanted to pay for a world class CPU design team just for Macs. ( in contrast Microsoft , Sony , etc went and bought custom PowerPC designs for their systems. But those were not trying to go head-to-head with Intel's mainstream offerings in complexity. )


Could they have been thinking of this when they designed the new Mac Pro? I gather it can get pretty hot, using ARM might help mitigate that.

Not very likely at all. The current Mac Pro doesn't have major problems getting rid of its 400W budget of power. ARM is for when have an order (or two ) of magnitude less system power budget.

The Mac Pro is much more an example of Apple using "right tool for right job" and not being caught in some dogma of "all have is a hammer everything is a nail". (the latter is also x86 (or ARM) cores solves all problems fan boy dogma). In FLOPs/watts modern GPUs are general better than x86 cores so Apple moved the design that way.

Apple's A7 (and likely any A8 that comes along) are not targeted to problems that Mac Pro , MBP , iMacs , or even Mac Minis address.
 
Please no. I can't take another processor transition...

Why does everyone assume they are ditching intel? Could possibly be for a Low Cost Mac Running IOS. Could be a box similar to a chrome book with everything stored in the cloud. If you give me a minute I could probably think of 10 other reasons. AND... They wouldn't fracture the ecosystem any more than it is now.
 
What could happen, is that it could run both.

Both re-compiled OS X app as well as iOS apps (modified for larger device displays, which may happen very soon :). Mouse-based apps present targets that are too small or hard to see on touch devices. But touch apps present easy to hit, well organized targets for a mouse or a trackpad, and are thus even easier to use by older people who need reading glasses (the biggest growth populace these days).

I run many of my own iOS apps in the Simulator on my MBA. Works great. Too bad developers don't make more of these builds available.

That's an interesting idea.

I am thinking: no way does Apple want to have apps optimized for two different user interface models running on one device (touch vs. pointing device).

But I admit I could be wrong...
  • Like you, I've run iOS apps in the simulator (as I've developed them) and it didn't feel "wrong" to use the mouse pointer with them. Using the simulator doesn't work well for two-finger gestures, but I've been using a mouse. A trackpad, of course, will do that quite well.
  • I've done Windows "Metro" development, which supports pointer and touch and I think it works well either way. (Judging from the Internets I might be the only one?) The point being: it seems reasonable to me for a single UX design to support both a pointer-based and touch-based UX that works quite well.

I'm half convinced you're right. (Maybe slightly less than half convinced -- if I had to bet right now, I'd bet no.)

I think Apple will really want to avoid complexity, confusion and ambiguity in their UX.

I think that was the fundamental sin of Windows 8: it runs in two distinct UX modes. The user needs to figure out how/when/why to switch between them and has to deal with the fact that everything works differently between modes. It's ghastly. E.g., when you go to open a web page you stop and think, wait do I want to open this in Metro-mode or regular mode? And when it turns out you get it wrong you curse yourself.

If Apple can create a UX that feels unified between iOS-based app and OS X apps then I think it's possible. The iOS UX itself will have to evolve. Right now on iOS apps are all full-screen one-window apps. That will have to change.. but perhaps that's already in the works with side-by-side multitasking?

Interesting idea.

(Just to point out, though: this is all independent of a CPU architecture change. That is, while it makes some sense to do it together, Apple could start supporting iOS apps on OS X with or without a CPU change.)
 
Has this been done in the labs at Apple HQ?
Probably yes.

Will this ever make the Foxconn production lines to be sold to the public anytime soon?
Not a chance. This prototype is forward thinking if it indeed exists. To cover all bases and because Apple can basically.
 
That may change entirely in a couple weeks. Wait for WWDC.

Apple shifting from some fixed multiple ( 2x or 3x ) .... not holding my breath on that one. There will probably be some new ones. However, variable/abitrary sizes? The impact of that is far more than just an API change... the impact is on the app design which not particularly just the function/procedure calls.

Even if Apple rolls out a more universal layout manager for iOS apps.... That drops the need for OS X as much as increases it.. iOS getting the capability means a hardware with variable screens ( like an iOS iMac or various laptop form factors ). If iOS systems already do it ..... need for OS X system goes up? Not particularly.
 
They were on PowerPc and then switched to Intel because PowerPc was moving ahead too slowly.
And then they went "mainstream" with Intel, with a supported CPU which received updates regularly, along with everybody else.

In "mobile" they can do what they want since they have such a high market share and sell in volumes. But Apple didn't start its mobile adventure with a custom made CPU/GPU. They first went "mainstream" there as well.

So, if Apple uses ARM for their Mac line. What happens? They will:
a) have to move the entire Mac line to ARM (from the Air to the Pro) or
b) have OSX run on two different architectures simultaneously. Which is a complete mess.

If Apple moves to their own CPU (sort of like the A series) then they will have to constantly upgrade and develop an entire CPU/GPU system just for their Mac line, which doesn't sell enough to support such an investment.
So all in all.
No. Just no.

It doesn't make any sense now to switch from Intel to ARM (in my opinion) because ARM is pushing slowly into "computer" territory and doesn't offer a wide enough array of CPUs (to my knowledge) to cover the Mac line from "Mini" to "Pro".

PowerPC was not developing too slowly. IBM just ****ed up with apple and made a deal with Sony focusing on CELL and server class cpus. That's it.
 
I seem to recall Tim Cook mentioning something about new products in new categories recently...
 
Yes.
First, you are missing an entire class of software. Drivers. The systems that general OS X users operate tend to involve more than solely Apple hardware and/or Apple drivers. Push a button and recompile is likely an over simplification.

Good arguments and clarifications, thanks for those.

Are drivers nowadays still hardware-related, and made with low level coding and causing problems in this case. If, that is bad news for my current external device collection? Firewire and usb? I once put Arch Linux for my NAS, but did not check how its hardware support was. The box went to a nonresponsive state pretty fast when I started to experiment with that…. :).

Second, the majority of software development is not spent pushing the "compile it" button. Much of the process is spent in fixing stuff that is broken. Some failure modes can like in parts that aren't so similar. Generalized enough, all instruction sets are the similar. Developers don't have to write in low level instruction specific assembler, but higher level compilers are much better at isolating developers when everything goes right rather than when things don't.

That hints to me, that this hardware transition could comes with a heftier software price tag than we originally hoped for, and also sometimes can result with crashing software, when testing is left to the users.

For customers with limited Internet bandwidth they had to download twice as much stuff if have both x86 and ARM Macs. Developers now has twice as many uploads and applications to test. Apple has twice as many applications to test and approve.

Again, costy. The ARM chips may be cheaper which partly compensates for that, but I start to see a my wallet getting thinner and thinner.

Have you used an iPad with a bluetooth keyboard? Not completely elegant but quite pragmatically useful if largely engaged in writing lots of text. It isn't like the trackpad has to be highly leveraged in all workloads. An additional trackpad can be largely in the same "boat" as an additional keyboard.

Yes, I have. For writing it works indeed, and solves one big problem with iOS. the virtual keyboard taking a lions share of display when editing. There are some inconveniences, like having to keep the iPad in not so optimum angle, and iOS display is simply too small for my eyes if it is kept too far away, but those could be solved with laptop like form factors. As I understand, ARM can be put in any form factors, and it works. I actually wonder how powerful Mini-sized ARM machine could be, CPU/GPU-wise. Top end Macbook Pro or just Apple TV capable?

Win8 Metro mouse is more different than inelegant. It is also not particularly necessarily to implement that specific way either.

I personally find it inelegant with mouse, because the controls including buttons are too large and scattered to every corner. I have to move eyes and the cursor too much. Inelegant and impractical mean the same here. For WP8 or Surface, it is nice and practical. For large vertical displays with mouse, not so much. Definitely not what Apple should find acceptable.

There is no reason developers should compelled to make OS X software the exact price as iOS software. That is a consistency for consistency sake argument. There is little to do with economics to back that up.

I agree about pricing. iOS and OSX are not comparable. Fuller software has a right to be more expensive, because it is priceworthy. If the software is the same… paying twice is not tempting. Now, if iOS and OSX platforms will be mixed, can pro software of Macs still be succesful and temptating or will it be cribbled and complicated with cryptic business models, subscriptions and app in purchases.

ARMs in Macs similar for consistency (i.e., completely maximize the components that both Macs and iOS devices use) is similarly lacking in economic reasoning and far more homogenous for sake of homogenous line of reasoning. Apple already has a OS/app ecosystem for ARM devices. The simplest approach is to just deploy that to any ARM based laptop/desktop box. In that case the user experience works just like it does now. Done.

Putting the OS/apps into another physical form factor doesn't particular do much to the apps. Can add keyboards now where they don't come by default. A trackpad as an alternative point/touch device isn't a huge leap from a touch screen. Nor does it necessarily preclude a touch screen being present.

For deploying, I am not sure. Now when I think about it, are there UIs that work equally well for both mouse and touch currently and don't lose in comparison to more optimized interface on Mac side? That could be a quideline to follow. If x86 to ARM and ARM to x86 are relatively straight forward processes and the resulting UI were really usable and close to optimal, then why don't we have that vast iOS software library compiled and sold succesfully to our Intel Macs with mouses already.

Webpages are quite usable on both platforms, but they don't have complicated selection dialogs in general. Those, who have, tend to be buggy at least on my iPad. Touches are neglected and so on, which hints that touch areas do not correlate well with cursor.
 
Last edited:
WTF do they need a full computer for if that's all they do? They can use an existing iPad. It even has M$ Office now. That doesn't mean it's a good idea to split the prime OSX market into two groups again (like the overlap of PPC and Intel) and cripple software like VMWare and Apple's own Boot Camp into worthless oblivion when that's what sold a lot of Macs that would have otherwise been PC purchases. All that is GONE with Arm. Just plain gone. In its stead, you get 100% iOS on your desktop platform (a nightmare to many, especially if it replaces the desktop OS entirely in the end and Apple finally locks up software to only the App Store like iOS). I don't call that being a drama queen. I call that the logical reality of all Apple's moves up to now if they suddenly shift to ARM. OSX has been getting more iOS'ified as time goes on already.

This is the point isn't it. An ARM only OSX would never work. It's not going to have the power to run Say FCPX or dozens of other Pro apps. So everything going to ARM... .well it's just not going to happen is it.

But what people are not considering here is what if this is NOT OSX or some form or ARM RT thing at all ... but iOS in a laptop form. Perhaps touch screen / with a trackpad control too. could even be a hybrid device with a removable keyboard etc.

This is MacRumors after all and the actual source is not exactly spot on very often or infact....ever!
 
An ARM system is never going to be as powerful or flexible as an X86 one, the GPU of an ARM system is never going to match a Nvidia Titan for example, and who is releasing an ARM based server and what OS does it run? Because I only know of one ARM laptop and that's the Chrome Book.

The 'quite a lot' of people you mention would get the exact same experience from an iPad, they buy a computer because they expect it to do more and be more powerful, especially with a Mac due to it's high pricing. They don't expect a glorified iPad with a keyboard and mouse.

And that's the general public, not tech enthusiasts I am talking about. And I'm guessing you have Apple shares also? Because that can be the only possible reason I can see you would actually WANT Apple to make MORE profit?

Wat ? ARM is more flexible than x86. You are mixing 2 different things. A cpu and specialized gpu chip. But again ever heard of CELL ?
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.