Why You’ll Probably Never Own A Mac With An ARM Processor [Feature]


Image via Ars Technica

Ever since Apple launched the new MacBook Air, analysts and Mac fans alike have gone wild speculating that Cupertino might dump Intel and use custom-made, ARM-based chips in their laptop line instead. Yesterday, more fuel was thrown on the fire when it was revealed that an Apple intern worked on porting OS X to ARM devices back in 2010. Even Intel has said it would be “remiss” of them to dismiss the possibility that ARM might steal their Apple business. On the surface of things, it looks like ARM might make its way to our MacBooks soon.

Is ARM really a threat to Intel? Yes, absolutely, and especially as we transition into Apple’s Post-PC world. But there is next to no chance Apple will replace Intel chips for ARM-based ones any time in the next five years. In fact, there’s a good chance the exact opposite could be true, and Intel chips will be powering our iPhones and iPads by then. Here’s why.

Where ARM’s Power Efficiency Comes From

For most people, it’s not immediately clear what makes the chip inside your iPhone (an ARM-based chip made by Apple) different from the processor inside your MacBook Air (an Intel chip). So here’s a remedial layman’s primer.

All other things being equal — and as we’ll see later, all things are not equal — the main advantage ARM chips have over Intel ones is power management. But why are ARM chips so much more power-efficient than Intel’s that they can be used in the iPhone?

It’s because of a fundamental difference in the chips’ architectures. ARM’s RISC-based architecture has a distinct edge in power-efficiency over Intel’s x86, which was designed in the late 1970s. While computer architecture is a complicated thing, for the most part, RISC is more power efficient than x86 because it has to spend less energy figuring out where one instruction ends and the next begins.

With x86, an instruction to the chip can be any number of bytes. That means in any 64-byte chunk of memory, you can have any number of instructions… and as a result, a computer chip has to spend energy separating instructions before it can process them. With RISC, though, every instruction is 4 bytes: the chip knows that every 4 bytes, it can expect to see a new instruction. It doesn’t have to work as hard figuring out the grammar. Physically, this manifests itself in an ARM chip by allowing you to make your CPU cores smaller than their x86 counterparts, and for these CPU cores to draw less power.

Think of it like this. Which of the following sentences is easier for you to read?


The cat sat and ate his hat.

The second sentence is clearly easier to read than the first, because in the first sentence, there are many different types and lengths of words all crammed together. In the second sentence, all the words are spaced apart, and each word has exactly the same number of letters. A child can read the second sentence easily, while the first sentence would challenge many adults.

Again, we’re making this a lot less complex than it actually is in the interests of accessibility, but for the most part, the difference between the way RISC and x86 pass along instructions is like this example. An instruction passed along in RISC takes less energy to compute, and can be understood by a smaller, less advanced chip.

Those are all big advantages… but unfortunately, they have diminishing returns.

ARM vs. Intel

As we’ve seen, ARM is better than Intel chips at decoding instructions. But there are two other things every chip needs to do: execute those instructions, and put them into memory. And Intel has the advantage there, says David Kanter, principal analyst and Editor-In-Chief at Real World Technologies, a technology analysis firm specializing in x86 and RISC microprocessors.

“If you look at any modern, high-performance chip, what you’ll find inside is maybe 25% to 35% CPU cores, 35% to 45% cache memory, and the rest is other stuff like memory control, I/O and so on,” Kanter told Cult of Mac in an interview last year. “So while ARM can make their CPU cores about 20% smaller and more power-efficient than Intel can because of RISC, all things being equal, that’s only a really tiny advantage overall. Maybe 4%.”

And that 4% advantage disappears the second you put Intel’s massive manufacturing muscle into the equation.

Intel is the Apple of the microchip world: everyone is at least a year behind Intel when it comes to competing with their cutting-edge technology and design in the x86 space. Because of Intel’s manufacturing strengths, they can make the entirety of their chips smaller (and therefore more power-efficient) than anyone else around, not just their CPU cores.

The result? Intel’s CPU cores may need to be slightly bigger and less efficient than ARM’s, but the overall chip is smaller and less power hungry at the same speeds.

So why is Intel so behind ARM in the mobile space? Simple: Intel screwed up. Over the last decade, Intel spent years focusing on performance over power management, desktop over mobile. They were so focussed on winning the x86 Gigahertz war against the likes of AMD that they didn’t see the rise of ARM and Apple’s Post-PC world coming.

But Intel’s now seen the light. They have almost infinite resources to throw at the problem of catching up. And they’re going to do so quickly.

“By 2014, Intel will have gotten their power management ahead of everyone else and be using their manufacturing muscle as a major advantage in the mobile landscape,” predicts Kanter.

And when that happens, most of ARM’s advantages over Intel will go away, save one: Apple can design and tweak its own ARM chips to its heart’s content, adding all sorts of custom low-power graphics solutions and oddball sensors to the chip. That’s harder to make happen with Intel.

Chip mod-ability is a compelling reason why Apple will likely stick with making their own ARM-based, A-series chips for mobile devices. But when it comes to a laptop or desktop Mac form factor, getting these things baked into the chip itself isn’t as important, when you can add other hardware or software to manage the same tasks.

There Is No Threat To Intel From ARM In The Mac Space

“The fact is that there is no ARM processor today, nor any that will be coming in the next five years, that are suitable for Apple’s existing models of laptops and desktops,” says Kanter. “On a deep and profound level, there is no technical advantage right now for Apple to switch to ARM across its laptops and desktops.”

Why? ARM processors are still relatively slow, and unsuitable for the vast array of tasks we take for granted in a desktop or laptop. Compared to the Core i7 in your MacBook Air, the core of Apple’s A5 CPU is similar to that of a 1995-era Pentium Pro. A full-featured port of OS X simply can’t run on an ARM series chip right now, which is what gave us all iOS — a massively stripped down version of OS X — to begin with.

“Apple won’t use ARM in the MacBook Air or any other Mac laptops anytime soon, because by the time you’re done making compromises, you’d end up with an iPad,” says Kanter. And even if ARM chips could be quickly ramped up to match the power of Intel’s current chips, it’s not clear that they would be any more power efficient than Intel’s offerings.

But ARM doesn’t need to replace Intels in Macs to beat Intel. While Mac sales are booming, iPhone and iPad sales are exploding. So although Apple’s Mac business is very healthy, it’s becoming a smaller part of the overall pie. Last quarter alone, Apple sold 37 million iPhones and 15 million iPads compared to a mere 5.2 million Macs. If iPads and iPhones continue to overshadow PC sales, ARM doesn’t need to come to Macs to be a very real and very scary threat to Intel.

All the more reason for Intel to get serious about power management now and figure out how to more closely work with companies like Apple to offer an easier, more streamlined way to insert their own custom technology into Intel’s x86-based chips. But Intel knows they’re behind and are taking the lessons they’ve learned from ARM over the last few years seriously. More importantly, Intel has the resources and expertise not just to catch up, but to blow the competition out of the water in the next few years.

So Why Was Apple Working On An ARM Port Of OS X?

If ARM isn’t really suitable for running OS X, and if the power advantages of ARM mostly disappear when you make the chips as fast as Intel’s offerings, then why was an engineer working on porting OS X to ARM back in 2010?

We can only speculate, but there’s any number of reasons that an intern might be tasked by Apple to explore this route that don’t end in a commercial product. We already know that Apple puts new engineers on fake products until they can be trusted. It’s also possible that Apple would want a bare-bones version of OS X that they could show Intel to help at the negotiating table. Finally, Apple may have simply been doing preliminary groundwork in case of the eventuality that ARM does catch up to Intel in five-plus years or so.

(Update: Seth Weintraub from 9to5Mac wrote to inform us that the real reason Apple had an intern working on porting Darwin to ARM was to update Airport platform, which has devices running Marvell ARM-based chips. In other words, as we’ve said, moving the Mac to ARM had nothing to do with it.)

Apple’s not serious about ditching Intel for the Mac, but even so, the next few years will be very interesting as ARM and Intel trade body blows. If Intel plays its cards right, come 2015, we could all be talking very seriously indeed about whether or not Apple will be putting Intel’s new mobile chips in the iPhone 8 and iPad 6.

[image via Ars Technica]

[teaser-top]But could someday have an iPhone with Intel inside.[/teaser-top]
[teaser-featured]But could someday have an iPhone with Intel inside.[/teaser-featured]

Deals of the Day

105 responses to “Why You’ll Probably Never Own A Mac With An ARM Processor [Feature]”

  1. ddevito says:

    Excellent article.

  2. aardman says:

    Okay, I woke up today determined to ditch all my Intel holdings then you throw a wrench in the works with this article.  Thank you very much.  :-)

  3. prof_peabody says:

    while I agree that currently ARM is no match for intel in the laptop space, this article is less than convincing.  it reads like a blurry re-telling of another article but you don’t link to that article. 

    the main arguments seem to be that RISC chips are more power efficient (which is funny considering Apple famously switched from RISC chips due to their *poor* power management), and that intel has “better manufacturing.”  what?  

  4. David Nielsen says:

    One reason to work on porting OS X is that supporting additional architectures is a good way of weeding out certain classes of bugs and remove chip specific assumptions that are or can become problems down the road.

  5. brownlee says:

    Everything in this article is original research, and none of it has been “blurrily retold” from other articles. You’ve also fundamentally misunderstood the article: RISC chips are more power efficient in the way they parse instructions, but less power efficient than Intel chips in all other areas. And while Apple may have switched from PowerPC (a RISC chip), the A5 and A4 chips in the current iPhones and iPads are RISC-based, so power inefficiency of RISC as a whole had nothing to do with it. The reason Apple switched to PowerPC was because they were several generations behind Intel’s chips in all areas, including power management. 

  6. TomasCarlson says:

    Hmm… ARM=RISC… RISC=PowerPC… Déjà vu all over again? Guess I’ll have to dump my Intel-powered MacBook Pro as it will be obsoleted shortly… not.

  7. David Nielsen says:

    By manufacturing is meant that Intel currently are on 32nm for Sandy Bridge (what is used in your 2011 Mac models), 2012’s Ivy Bridge will switch to 22nm (and 2013’s Haswell likewise being 22nm but adding important features and more powerful graphics). Broadwell in 2014 will be 14nm.
    Compared to this Apple’s A5 is 45 nm, A6 might be shrunk down to 28nm but no details are released yet nor any long term plans. Note however that to get to 22nm Intel basically had to reinvent the transistor (with their popularly named 3D transistors), a major undertaking, one which ARM will have to license their way to and adjust manufacturing to using. Not to mention redesign their chips to use, if they do not they are likely to be unable to shrink beyond a certain size.

    With every bit of shrinkage you get a smaller die, which roughly means you can either use that improvement to increase performance or lower power consumption (a very simplistic view). Intel have a strong plan and have shown working models all the way to their Haswell 22nm design today.

  8. David Nielsen says:

    By manufacturing is meant that Intel currently are on 32nm for Sandy Bridge (what is used in your 2011 Mac models), 2012’s Ivy Bridge will switch to 22nm (and 2013’s Haswell likewise being 22nm but adding important features and more powerful graphics). Broadwell in 2014 will be 14nm.
    Compared to this Apple’s A5 is 45 nm, A6 might be shrunk down to 28nm but no details are released yet nor any long term plans. Note however that to get to 22nm Intel basically had to reinvent the transistor (with their popularly named 3D transistors), a major undertaking, one which ARM will have to license their way to and adjust manufacturing to using. Not to mention redesign their chips to use, if they do not they are likely to be unable to shrink beyond a certain size.

    With every bit of shrinkage you get a smaller die, which roughly means you can either use that improvement to increase performance or lower power consumption (a very simplistic view). Intel have a strong plan and have shown working models all the way to their Haswell 22nm design today.

  9. Mike Rathjen says:

    Macs with ARM means the end of Boot Camp, Parallels, and VMWare. Running Windows is very important to a lot of Mac users.

    • Speedy says:

      Windows RT runs on ARM.

      Most Windows software doesn’t run on Windows RT yet, but Microsoft themselves are pushing third-party developers to resolve that.

  10. RhysLadhani says:

    This was on macrumors like 2 weeks ago. COM slacked on this story. but id like to see ARM cpus in  macbook airs. i need my mbp to stay intel. but the smaller more travel based laptops should switch. battery life would be increased by ALOT and isnt that the point of laptops anway? to be able to use it without being plugged in all the time?

  11. sn0wball says:

    awesome read

  12. joewaylo says:

    Not necessarily the end. BC, Parallels, and VMWare would release an ARM version a few months after receiving a developer ARM MacBook Air. And Windows 7 already has an ARM version. But the Intel Windows programs you use on it would need to ensure their programs can run on an ARM Windows too which would take a few months for them to fix up.

  13. Mike Rathjen says:

    While these are possibilities, they are not trivial and the end result is a huge performance loss.

    x86 VM’s are “easy” and run close to full performance because Macs are x86. Remember in the pre-Intel Mac days that virtualizing an x86 OS on PowerPC was a difficult and slow endeavor. I seriously doubt Parallels would be ready in a few months, as we are essentially talking about a whole new product.

    The last I checked, Windows ARM is not intended to be sold on the shelf. Microsoft will make specific one-off versions of Windows ARM for very specific hardware. So this is a possibility, but only if MS partners with Apple in making Windows ARM specific to Mac models.

    I agree with Brownlee. This isn’t going to happen for at least 5 years, and by then Intel will have power efficient chips. In fact, I seem to recall an article here last year that reported Apple was specifically working with Intel to create more power efficient chips.

  14. David Preece says:

    So, so wrong.

    Apple has spent the last five years or so working on ways to spread the load across multiple cores – blocks, GCD etc. And, of course, have reasonably high performance dual core ARM’s of their own design shipping in the ‘millions per week’ territory. Remember, as well that Apple are now on their third ISA and on each occasion have managed to make the move with, really, very little trouble … and even now have of the order of hundreds of thousands of developers writing ARM code for iOS.

    Eight core ARM laptops are on their way and will be faster, more power efficient and will make Apple about $50 per box more money.

    • Speedy says:

      The *Mac* is on its third major ISA. Apple is on their … 5th? 6502 (Apple I, Apple II), 68K (Apple III, Lisa, Mac, Laserwriter), PowerPC (Mac, Pippin), ARM (Newton, iPod, iOS), x86 (Mac).

  15. imajoebob says:

    Just like all that stuff was when Apple switched from PPC to Intel.  We got past it.

  16. imajoebob says:

    For further reading see: “Why You’ll Probably Never Own A Mac With An Intel Processor”
    – by any Mac writer circa 2004

  17. Stephen Reed says:

    Intel has high margins and their mobile parts will be too expensive even if slightly more power-efficient than ARM parts in the next couple of years. I believe that Apple will continue to use their own ARM licensed parts. 

    Intel has avoided direct competition with the low-cost embedded CPU (i.e. ARM-licensed) manufacturers for decades, going so far as to sell off their own ARM-licensed XScale division to Marvell. But in the Post-PC era, with non-Mac PC sales actually falling in Europe, Intel (like Microsoft) must enter the mobile marketplace – but faces peril in the cutthroat margins realized by the current players: Samsung, TI, Broadcom, Freescale, nVidia, Qualcomm etc.One cannot preclude Apple creating at least a prototype Mac/ARM device to force Intel to lower its prices for laptop chips.

  18. jaymartin says:

    You actually have the comparison slightly off: it isn’t ARM vs Intel, it’s ARM’s design vs x86 design. Remember that ARM doesn’t manufacture chips – it only designs them and then licenses those designs to others to manufacture. Intel, on the other hand, does both. 

    I think the more interesting story here would be how Intel will fare vs the ARM-licensed chip makers – Samsung, Qualcomm, (perhaps AMD), etc. Can Intel continue to dominate with it’s approach of the one size (design) fits all OR will they be forced to move into more of a manufacturer (less of a designer) role? And, if the latter, will they also more fully separate it’s design business to mirror ARM in that respect?

    You touch upon it but it is the real story here IMO – not ARM vs Intel.

  19. Steve says:

    And they will have to write new drivers for everything that hooks up to it.  The $50 cost savings per unit will more than be lost.  And 8 cores of a weak processor still doesn’t add up to much.   Remember, they got the original Pentium processor to run on .1 watts.   You talk of great things, but Intel has a serious arsenal.  

    • Speedy says:

      Internal to Apple, porting drivers is a one-time cost. $50 per unit is indefinite.

      Most drivers are written in C, so “write new drivers” is mostly just “rebuild with different configuration”. If some third-party wrote their hardware drivers in x86 assembly, do you think Apple cares about them so much that they’d hold their entire platform back from an additional $50/unit profit? I can’t think of any company that Apple would do that for.

  20. semiwiz2002 says:

    Intel Investment Thesis
    Manufacturing Capacity
    In the 2010 fourth quarter earnings conference call, Intel management discussed a process shrink from 32nm to 22nm. They also mentioned that they have three primary fabrication centers. With a shrink to 22nm those three facilities could be reduced to two and have the same chip output. Surprisingly, the CEO announced that Intel would be going to a four fab model. This move effectively doubles the number of chips the company could produce relative to 32nm. What is the added capacity to be used for?
    From Paul Otellini:
    “As we approach our 22-nanometer transition, we are increasing our investments in manufacturing to capture what we believe is a significant opportunity for growth. Stacy will walk you through more details in just in a moment, but in short the market opportunities for our 22-nanometer products are outstanding. As a result, we are growing from the model of three high volume leading-edge manufacturing fabs to four
    Our 22-nanometer process will be the foundation for growing PC and server segments, as well as a broad family of Atom-based SoCs, serving smartphones, tablets, smart TVs, and other embedded devices.”
    With the move to 22nm, Intel has grown the lead over the best in class competition to as much as two generations of manufacturing technology. This 22nm technology will also feature TriGate transistors that will increase the performance of the transistors while reducing power consumed dramatically. Some recent information indicates a reduction of 95% in quiescent power consumption when compared to 32nm planar transistors.
    Intel is promoting a notebook format called the “Ultrabook”. This product is a thin packaging format similar to the MacBookAir. Intel is subsidizing tooling and supply chain establishment for PC manufacturers to the extent of $300 million. The Ultrabook, in most cases, will be too thin to use a hard disc drive, so flash based solid state drives should have huge growth as this plays out over the next 2-3 years. The boot time for an Ultrabook with a SSD will be seconds, the performance will be much higher relative to a HDD PC and it will be “always connected” even when in sleep mode. The elimination of a HDD and the low power level of the new 22nm TriGate Intel processors will extend battery life substantially.
    Intel makes Solid State Drives. The Intel/Micron partnership has produced the world’s first 128Gb flash chip in the Micron/Intel joint venture fab in Singapore:
     http://seekingalpha.com/news-a…   The article seems to imply an eight chip stack of these chips to produce a 128GB SSD in a thumbnail size format.
    Many analysts give Intel a bad rap due to their lack of involvement in smart phones and tablet computers. Some even feel that embedded ARM processors represent a serious threat to Intel long term.
    The real fact is that Intel has an architecture that is very high on computational power and also higher on electrical power than these mobile devices could tolerate. Looking at this from the recent past, Intel would not call the mobile business a served market. The devices neither needed the compute power offered by Intel nor could they tolerate the higher power level.
    That all changes at the 22nm TriGate node. The mobile devices need for more compute power than ARM processors can provide is growing and the latest Intel technology will meet or beat the electrical power requirements of these mobile devices.
    The bottom line is that Intel could not participate in this segment until now. They could, however, prepare for engagement in the mobile business. They have done this by building manufacturing capacity and designing low power functional blocks while waiting for their 22nm manufacturing plants.
    We can expect some interesting announcements at the upcoming Consumer Electronics Show in January.
    Bought Infineon’s baseband business in order to have a complete, hassle free, market proven baseband solution that can be embedded in an application processor SoC.
    In an unprecedented move, Intel is doing foundry work for a startup FPGA company. Intel is giving Achronix Semiconductor access to its 22nm TriGate process.
    The speculation is that the payoff for Intel is complete access to the Achronix technology for embedding with Atom processors in order to give mobile products OEM customers a level of product design flexibility not available from any other application processor vendor.
    A while back Intel bought McAfee, probably not just because they like to write big checks. McAfee announced DeepSAFE at the Intel developers forum. DeepSAFE provides security near the silicon level, beneath the operating system. It is very possible that Ultrabooks shipped with the new Intel Ivy Bridge CPUs will have a final solution to the exasperating problems of malware by putting “hooks” in the chip that makes all other security software obsolete. This could be rocket fuel that launches the Ultrabook next year.
    Apple and Samsung are locked into 30 different lawsuits in nine countries. To me it is obvious that Apple can no longer depend on Samsung as a supplier of critical component such as their “A” series of application processors.
    The scale of this business will soon approach 300 million devices between the iPod, iPad, and iPhone. The current A5 chip is 122 sq mm in size. That means that 500 chips can be produced on a 12” wafer. 300 million chips will require a leading edge manufacturing capacity of 600,000 wafers per year. That level of capacity/technology only exists at two companies in the world, Samsung and Intel. The Intel technology is two generations advanced from the Samsung process, so moving to Intel would produce a smaller A5 chip at higher speed and even lower power.
    The Citi semiconductor analyst feels this is a distinct possibility and could be made public around the end of the year.  http://www.cnbc.com/id/4332707…|headline|quote|text|&par=yahoo
    In the latest earnings conference call, Otellini was asked if Intel is doing foundry work for anyone. His answer was that they are doing a small amount in the FPGA area (Achronix), and “a couple of strategic customers that I am unable to discuss at this time”. Would Apple be considered strategic?
    While I’m at it, why would Intel do foundry work for a startup FPGA company? My guess is that Intel intends to include a chunk of FPGA in their first mobile SoC product. That would give unparalleled design flexibility to end customers who select that part.
     Side note: Apple and Intel have collaborated on the development of the Thunderbolt technology used in the Apple computers. This is an exclusive arrangement for some period of time, after which it will be opened to other equipment manufacturers. Intel makes the part.
    The Intel investment thesis
    ·         Ultrabooks will stimulate a new growth cycle in PCs.
    ·         Solid State Drives represent a $70 billion flash memory opportunity with Intel positioned well.
    ·         Intel has a huge fabrication technology lead over the next best competitor.
    ·         Intel is actively building out capacity to serve these new opportunities.
    ·         Intel’s mobile engagement will begin in 2012.
    ·         A final security solution.
    ·         An Apple/Intel relationship is very likely to come out of nowhere to surprise the doubters.
    ·         An Atom based SoC with embedded baseband and FPGA?
    ·         Intel’s additional manufacturing capacity could support a relatively short term doubling of the company.

  21. Steve says:

    Very good article!  I do like the comments at the end, and I can tell there are engineers and people with engineering minds who are following this closely.   Qualcom changed the name of their stadium in San Diego to “Snapdragon” in January for CES for two weeks.  Did you notice? Ask your non-techy friends what a Snapdragon is.   Or better yet, give them an option of three different processors that are in their phone with Intel being one of them and see how often they select Intel.

    The name “Intel” caries consumer confidence along with it and Intel will most like put “Intel Inside” on their phones and tablets.  Say what you want about how little this matters, you have to admit to unknowing consumer it does carry some weight.  

    My gut feeling is that Intel will throw it’s marketing muscle behind it’s phone processors and help pay for advertising like it did in the past.   It has signed on Motorola for phones and tablets and you can expect more to follow.  

    What the article didn’t talk about is the relationship between Apple and Intel.  Is it so bad that Apple needs to dump Intel?   I think it looks fairly strong.  They did work on joint projects like Thunderbolt.  In the future Intel will have the manufacturing capability to handle Apple’s needs in a manner that the competition simply won’t be able to match.

    Intel has a ton of money to power it’s lead over the industry coming up over the next few years.  You can talk about which technology is superior.  In the end money has a larger role to play in the equation than you may suspect.  Intel is buying their way into this market and locking up large contracts along the way.   Bet against Intel if you like, it very rarely pays off.  

  22. crustyjusty says:

    I haven’t seen anyone address the massive difference in price between ARM processors and Intel processors.  Sure, ARM is still way behind Intel in terms of processing power, and Intel is going to continue making headway on power management, but doesn’t Apple license ARM processors for $.30 per chip?  Add on $5-$10 for manufacturing costs and you’re nowhere near the price of an intel chip. So, even if the chip is a bit slower, this would enable the Apple to price certain products way below firms that continue to use Intel products, assuming they can tweak the performance to where they want it.

  23. campcamp says:

    And when has apple cared about that? They threw away the floppy drive, threw away the DVD drive. They move forward. 

  24. campcamp says:

    the article is crazy. It basically states that ARM is only temporarily superior to Intel in the mobile space because Intel wasn’t paying attention.

    For god’s sake, intel has been working on an ARM solution years. It’s not happening.

    Regardless of whether or not x86 can compete theoretically—where is the intel phones? where’s the adoption? Where’s the gigantic app ecosystems? 2012 is a watershed point where more than 50% of US will be using smartphones. And intel isn’t even throwing dice.

  25. harold de vries says:

    x86 instructions can only be up to 15 bytes ever, everyone knows that. The 64-byte block is also somewhat misleading – there are no 64-byte blocks involved anywhere*. Intel uses 16-byte blocks of code and newer AMD’s use 32-byte blocks.
    * ok the loop buffer of Core2 can effectively store up to 64 bytes of code, but it’s really “four 16-aligned 16-byte blocks that are not necessarily adjacent”

  26. Martin Plamondon says:

     We shouldn’t forget about AMD and VIA.  AMD’s new APUs are on the right path for power efficiency while VIA always made cheap and power efficient x86 chips.

  27. Playos says:

    His point is that battery life wouldn’t nececerly be increased a whole lot when actual usage and similar software is installed.

  28. Playos says:

    Y’all got past it because you have Parallels and Roseta… Neither are possible with a switch to ARM.

  29. Playos says:

    Yep, and for 4 of the 6 smartphone OSes (iOS, Android, WP7, BBos, WebOS, Meego) they could care less about processor subsystem. Java, .Net, and HTML+js don’t care if they’re on ARM, x86, ext. So an all out switch wouldn’t be difficult. iOS is Apple, and they can switch out the platform and their devs will follow suit.

  30. Hung Ho Nguyen says:

    64bit memory addressing :). even if you have a 16bit instruction, the compliler/processor will convert it into 64bits (fill the rest with zeros). The result: risc like cpu and more memory addressing space.
    Disadvantage: apps take more memory and space but since memory and hard drives are cheap. So it’s not really important.

    Most arm cpus are using 32 bit instructions (you rarely see 4bytes = 128bits)

  31. paulbeard says:

    I haven’t seen it but there are claims that OS X (using VNC to handle video) can run on the ARM-based AppleTV 2, the one that isn’t a variant of the Mac Mini. I doubt it’s terribly useful. 

    And it’s worth bearing in mind that, as noted below, OS X was running on Intel hardware for years before we saw it running there (I remember seeing CPU tests in init scripts in 2002 [http://goo.gl/jz623] that wouldn’t have been necessary if it was only built for PowerPC). Competition is good, right? I wonder how many different architectures are humming away in the labs on Infinite Loop. 

  32. paulbeard says:

    And the earlier switch from 680×0 to PowerPC. 

  33. paulbeard says:

    My recollection was that Apple switched from PowerPC because IBM wasn’t able to deliver a fast enough laptop CPU that managed energy and heat to their liking. For whatever reason, maybe they weren’t making enough from Apple’s tiny market share at the time, IBM didn’t spend the resources on it (to be fair, chips are only part of what they do, unlike Intel). Look back at how quickly the CPU speeds shot up once the transition was underway. 

    A similar parallel can be seen in the iPhone and 4G: when power requirements (which drive battery size and form factor) are in line with their other specifications, we may see a 4G iPhone. 

  34. MattS says:

    Please  share with us your background in microelectronics and integrated circuit and microprocessor design.  There is no research that I see.  Just the retelling of some vague tech marketing stories.  The truth is that Intel’s x86 architecture actually uses in some places, concepts that came from RISC.  The current Intel x86 are considered by many to be a hybrid of CISC & RISC.  Maybe you should stick to writing for Playboy.

  35. Scott Cushman says:

    Even when ARM has designs out that addressed more than 4 gb of RAM, I don’t see that switch happening. It’s not a question of RISC vs. CISC (x86) architecture anymore; our computers will use both. Did I say, “will use?” I meant, “are using.” We already see the advantages of RISC architecture in Macs, PCs, etc. It’s called the GPU.

    As CUDA, OpenCL, and other general processing on the GPU schemes mature and standards emerge, we’ll see more and more of the brute processing move to the many-core GPUs, freeing up the CPU to handle the more complicated instructions that the x86 architecture can address with a comparative advantage. I think we will see a day when the GPU matters more to general performance than the CPU does. In fact, I can easily imagine iOS-like apps and Java VMs designed to run solely on the GPU. The hardware is there, the standards are emerging, we just need to see more OSes and software tuned to use our powerful GPUs efficiently for something other than pixel pushing. Let’s face it, only 3d gaming and a few fairly niche applications actually employ the full processing capabilities our systems already possess.

  36. David Weintraub says:

    There’s a really good reason why an Intern would be assigned to port Mac OS X/Darwin to ARM: Learning the ins and outs of the operating system. Nothing gets you more familiar with an OS than attempting to port it to another chip. What better way to train an engineer who’s working on CoreOS?

  37. WaltFrench says:


    It’s not actually clear that Intel “screwed up.” They continue to dominate the X86 market, with fat profit margins. Meanwhile, they have given up on the high-growth, but low-profit-margin marketplace. All well and good for cellphone manufacturers to have Qualcomm, TI and oh-so-many others competing for their businesses; that’s undoubtedly helped the cellphone market explode. But the profits have been miserable. It’s not clear that Intel could compete here in a way that they’d want to.

    Maybe once they figure how to leverage some of their fab technology, such as those 3D transistors we hear about, Intel will be able to get some share AND the profits it likes. Or maybe they’ll keep biding their time until the right time to get in, risking being marginalized by ARM. That’s a risk, but I don’t see it as a clear-cut screwup that our fine article writer has proclaimed.

  38. johndavidstutts says:

    Ouch, makes me wonder if some people is betting on the wrong horses

  39. WaltFrench says:

    “Intel Inside” makes good sense for a consumer wondering about a $1000 laptop. Especially if two machines are almost identical, and both running Windows Home Premium, it’s a nice differentiator. 

    Much less so for a phone that seems to cost $99. Nobody knows who makes the chips in a given phone. Hell, most don’t even know the name of the OS.

  40. twilightmoon says:

    Intel has a serious arsenal, but Apple has even more. Apple has over 100 billion in the bank waiting for a good reason to invest it. How much does Intel have?

    Apple makes nearly all of it’s money from mobile products: iPhone, iPad, and MacBooks. Very little money comes (comparably) from other sources such as desktops, software and services, and iTunes.

    How much of Intel’s income is from mobile products by comparison? How much in the way of margin is Intel able and willing to give up to capture the mobile space, where they have nearly no presence below laptops? This is not a slam dunk either way, and as pointed out by David, Apple could spread some of the load off the ARM processors to the GPU. Sure they would have to do more of the design themselves, and sure they’d have to write new drivers and such, but when has Apple been afraid of all that?

    I’m not convinced Apple is switching soon, but I hardly think any certainly as far out as 5 years is justified in this rapidly moving space.

  41. Dennis Munsie says:

    Porting Darwin to ARM is not the same as porting MacOS X to ARM. Darwin is the kernel that powers both MacOS X *and* iOS. The revelation that Apple had someone working on porting Darwin to an ARM SoC a few years ago should not be taken as they were porting MacOS X to ARM.

    As for speculation that Apple has MacOS X running on ARM in some lab somewhere, as Steve Jobs said once, they like to have options. MacOS X was running on Intel for at least 6 years before Apple transitioned publicly to Intel processors. Keeping a few engineers maintaining an ARM port would be relatively cheap insurance for them.  But it isn’t a sign that Apple is going to release an ARM based Mac any time soon.

  42. twilightmoon says:

    Hard to say. Certainly the competitive advantage of running stock Windows on Intel would be lost. But the exact level of power efficiency and battery life able to be achieved by Apple with control over both the CPU hardware and all the software and drivers is not really known.

    There’s more to this than simply the efficiency of the CPUs themselves. Put another way, while a MacBook Pro might not be suitable for this switch, how often is a MacBook Air running at full CPU speed? How many of them are pushed to the max, and how many would do just fine with a slightly slower chip that had other massive advantages on battery life?

    What if Apple could sell an ARM based 13″ MBA for $800 USD. What if those laptops ran for another 1 to 2 hours? Lets assume they are slower, but what if the difference in speed still allows for decent performance for word processing and web pages and email?

  43. twilightmoon says:

    I agree this is an issue for some, but how many? iPads have no such ability and they are replacing Windows systems at nearly 3x the rate as Macs, and that pace is accelerating. I am not saying Apple would necessarily replace all Macs whole sale with ARM in short order but in 3-4 years from now, when this might make sense, what will the computing landscape look like?

    Look at Europe, we’ve seen Mac sales exploding, and Windows collapsing. Is this trend going to continue? For how long and how deep into the Windows marketshare?

    If, by using scale, and the lower cost per unit of custom ARM chips Apple is able to bring their costs down enough to ship laptops for $100 to $200 less, how much impact would that have?

    I can hardly imagine that most commercial software won’t be available for Macs in the next 3 or 4 years, given the trends that are already occurring. So costs will really be the only major benefit of Windows, and is it safe to assume that Apple won’t find ways to bring their costs down and dip into the lower price points?

    If you were running a company that competes with Apple would you feel safe making that bet?

  44. Fraser Scott-Morrison says:

    So why then is MS porting Windows to ARM?  

    Its not exactly a hugely cut down version of Windows if its running desktop Office and Mac OS could be considered to have about the same demands as Windows.

  45. twilightmoon says:

    It’s unlikely that Apple would switch all their systems over to ARM at same time. More than likely it would start with the MBA and maybe the Mini.

  46. tomtubbs says:

    Why should we believe that Intel can pull off this trick? I’m just wondering, as I remember the previous cries of how Intel could power MIDs and iPhones – and that didn’t really work. See how far Atom had to go and how it stacked up against its rivals as well. Currently it’s being a bit cheeky by selectively having cores in effect heat up till the TDP is hit (that and their graphics issues, most recently manifested with a racing game video demo…) – its useful – but also laptops with them in get hot v quickly and the heat is focused. We don’t need as high a performance in a chip for a phone. Yes, as much as you can pack in – but it’s a balance still tipping in ARM’s favour i’d say.

  47. FlopTech says:

    Another ARM advantage for Apple is cost.  They pay a licensing fee to ARM Holdings for the privilege, but they don’t pay off-the-shelf boutique prices for Intel CPUs.

    And let’s not forget that the vast majority of MacBook Air users are not heavy into Adobe bloatware or helicopter aerodynamics simulations.  Email, surfing, tweeting, iLife, and iWork are all 90% of MacBook Airs will ever need to do.  A quad-core Ax chip could handle that.  With no cooling fan.

    And Apple has proven that powerful apps can run with very little memory or file system footprint. Especially when the heavy lifting is done in iCloud. They did it with iOS, and they could do it with OS X as well.

  48. WaveRunnr says:

    What about AMD? They could be the happy medium since their processors now include graphics.

  49. Playos says:

    My point is that if it has the speed for equal web browsing and word processing, the x86 chip from Intel will be more power efficient for the same level of processing. The x86 chip will also have better hardware options for graphics (a true way to save massive power without sacrificing performance).

  50. Mike Rathjen says:

    I made one point only, and I thought it was pretty clear. ARM Mac means no Windows. Other people have challenged my point with some reasonable responses.

    Strangely, you turned my very simple point into something else in your head, “no way will Apple switch to ARM because of Windows”, then argued against that point instead. That is not what I posted here at all, therefore I will decline to defend it.

    I actually DON’T think Apple will switch to ARM for Macs, but my reasons have almost nothing to do with Windows compatibility.

  51. twilightmoon says:

    Fair enough, there is room to interpret your comment in more than 1 way, and I get that maybe I did not interpret it the way you intended. I do believe that the ability to run Windows is a selling point for Macs, but I also do not believe it is at the top of Apple’s list.

    You put no time frame on your assertion that Apple will run Macs on ARM. I do not think it will happen in the next 2 years, but I do believe it is nearly inevitable. I feel Apple can match Intel’s strengths given time, they certainly have plenty of money. Apple has stated numerous times they like to control the technology that goes into their gear.

  52. DocNo42 says:

    Neither may be as necessary with the advancements Apple has made in complier technology…

  53. delmiller says:

    I don’t know about this John, the whole CISC vs RISC debate of the 1990s has become more or less moot as RISC designs have become increasingly more CISC-like and vice-versa. ARM’s middle name may be RISC, but that is more historical than accurate.

    The “Reduced Instruction” part of RISC wasn’t primarily about instruction size but had more to do with other architectural implications that have been diluted as CPU designs have implemented the complicated computing processes required for multiple cores, out of order execution, multilevel caching, etc.

    It’s really hard to implement today’s architectures without deviating from the spirit and letter of RISC philosophy. You don’t see the more pure RISC designs, like the DEC Alpha or HP’s PA-RISC or the MIPS, prevalent today for a reason. CISC and RISC architectures have both evolved into something that is really both and neither.

    I imagine that for the ARM design to become as powerful as Intel chips or for Intel chips to become as power efficient as ARM chips, the architectures would have to borrow even more heavily from the other than they already do-which is significant.

    Now you may be right that ARM devices will not soon be implemented in Macs, but the RISC/CISC issue has almost nothing to do with it.

  54. Sigivald says:

    Because MS is insistent that Windows Is Forever, On Everything, For Every Purpose.

    Windows has been on Other CPUs for ages – it was on the Alpha (I remember running NT4 on a DEC Multia, myself).

    It was on PowerPC for a while before PREP/CHRP imploded.

    It was on the SH4 for the Dreamcast.

    None of those meant those platforms went anywhere. (Okay, PPC had a good desktop run, but still.)

    Microsoft has wasted an awful lot of effort on porting Windows to dead-ends before.

    (And that excludes Windows CE/Embedded, which started running on ARM in 1997.

    It’s not like Windows On ARM is some sort of amazing groundbreaking thing that tells us about Microsoft’s Entire Plan For The Future.

    It’s a footnote.)

  55. Sigivald says:

     Also, apart from the kernel and hardware drivers, pretty much nothing in OSX should need to be ported to run on ARM.

    Unless it has inlined assembly [which it shouldn’t, because that’s insane], a simple recompile should make any decent software simply run on the new target platform.

    A Darwin port to ARM combined with driver ports for the subset of hardware they might want supported? That’s 99% of the work.

    (Agreed that it is not a sign that ARM Macs are likely – I’ve heard tell that it might be to help support AirPort devices somehow, which makes a little more sense.)

  56. Sigivald says:

     Can Intel continue to dominate with it’s approach of the one size (design) fits all

    Are we thinking of the same Intel?

    The Intel I know makes everything from 10-core server chips (Xeon E8860, $4616 at release, 130W TDP) to single-core embedded CPUs (Atom E620, $19 at release, 2.7W TDP).

    That’s not a lot of “one size fits all”; I suspect it’s actually more variation than exists in the ARM space.

  57. pat555 says:

    Does any consumer look for “Intel Inside” on a tablet?  The iPad still has over 90% market share.  If there’s anything consumers are looking for on a tablet, it’s an Apple logo.

    The processor architecture is actually one of the easiest things to change about a tablet.  First they have to get the software right.  There’s no evidence that anyone outside of Apple has figured out how to make tablet software (and the accompanying content ecosystem) work like consumers want, regardless of processor.

  58. bdkennedy says:

    I think Apple is going to either buy ARM or they are going to manufacture their own processors.  The processor and RAM is the only technology remaining that Apple doesn’t own inside of their computers.

  59. Nick Valvo says:

    Three things: 

    1) Presumably some machines in the Apple lineup would still run Intel. The PowerMac G5 hung on for awhile after the Intel switch, for example. 

    2) Isn’t Windows being ported to ARM now?

    3) For how long is Windows even relevant? The trend lines are pretty brutal for Windows versus OS X/iOS. PCs + Tablets, Apple’s offerings are already about 1/3 of the market. That’s way up from five years ago, needless to say. I get the considerable mindshare that Windows has in the enterprise, but still: where are we in five years?

  60. jaymartin says:

    The x86 design is what I’m talking about. When Intel can come up with a cost-effective low-power design that can beat ARM then we’ll see. The problem is that it took a long time for Atom to get where it is today – and ARM designs are still ahead of the game.

    So, yes, we’re thinking of the same Intel. The question remains – can Intel complete with ARM using x86 derivative designs. They haven’t gotten there yet.

  61. Playos says:

    I’d say MS has a better chance on that front than Apple.

  62. Mike Rathjen says:

    I believe it is far more likely that Intel can make their powerful CPUs
    very energy efficient than it is for ARM to make their energy efficient
    CPUs very powerful.

    Ever since Apple threatened to leave Intel over power efficiency back in 2010, Intel has been driven to reduce power consumption. In just a little over a year, we now have x86 processors that use only 15 watts and are still powerful enough for Windows 7, Lion, and modern applications.

    In contrast, ARM is nowhere near being able to run a modern desktop OS with demanding applications.

    You ask for a timeline. That’s a fair question but tough to answer. As we enter the post-PC era I think computing will change drastically. Most (but not all) computer users will fall into one of two categories: content-consumer and content-creator.

    In a few years (5 maybe?) content consumers will realize that everything they do can be done on their tablets. Tablets then will be very powerful and have better interfaces than they do now. Email, web, video, music, games, social networking, news, etc., will be just as good on a tablet as it would on a beige box, plus it’s portable, so why buy the beige box? It’s the end of desktops if you have a powerful tablet that can be both mobile AND at home connect to keyboard, mouse, big display, and big storage connected to your router (or storage provided by the cloud).

    iPads are improving at such a fast rate that technically we might be there in a year or two, but I say 5 years because it will take people a while to adjust to the fact that they don’t need a big beige box on their desk.

    Content creators, content hosters, scientists, PC gamers, etc., already need powerful workstations and servers, and will continue to do so. But they are less concerned about power and not concerned at all about things like battery life. Workstations will continue to be needed with full desktop OS.

    So I don’t know where the combination of ARM and OS X fits into this vision. The first group (content consumers) will be using iOS with an additional 5 years of advancements, so while they will be using ARM, they won’t need OS X. The latter group wants brute force over power efficiency, so they don’t need ARM.

  63. axoniclabs says:

    Apple won’t use ARM for MacBooks and they envision iPad to replace MacBooks in the long future. Intel will continue making CPU’s for MacBooks and Apple will continue to increase performance/power for it’s own ARM CPU’s. One thing the author left out is the potential performance gain from tight integration of CPU and software/compiler. See MacroScaler: http://arstechnica.com/apple/n

  64. Mike Rathjen says:

    1) Okay.
    2) Yes, but it won’t be boxed retail software. Microsoft has stated that they are making Windows ARM specifically customized to each tablet and that it isn’t meant to run Windows desktop applications.
    3) That’s an excellent question as we enter the post-PC era that Steve Jobs talked about. My personal opinion about this is quite long. If you are truly interested, it is posted just a few responses down.

  65. twilightmoon says:

    I’m unsure I buy that the graphic options of the x86 are better, they may be. I won’t disagree with you since I can’t prove otherwise.

    It may be true that x86 will win out here, but it has to overcome a pretty substantial price difference. There’s an entire industry that’s moving in the direction of ARM, and there are a few big companies making x86. I would not take it as given that x86 and Intel are going to win this. It’s not simply a matter of reducing the power consumption from x86, it has to overcome price, customizability issues, too.

    My personal belief is that Apple sees a major competitive advantage to having a high powered custom made ARM chip that they can control and have some form of internal project underway to make a high powered ARM chip that they can use to replace x86. Remember Apple thinks long term, so this might be 4 or 5+ years out, but I really do think this is coming.

  66. twilightmoon says:

    “In contrast, ARM is nowhere near being able to run a modern desktop OS with demanding applications.”

    This is not exactly a comparison. When this “race” started ARM was nowhere near able to run modern OS’s (if you mean Desktop OS’s–iOS is a very modern OS and is based on Mac OSX).

    How much did ARM catch up during the time Intel brought down its power consumption? How much progress was made on each side? ARM added duel cores, faster speed, and some other goodies, Intel reduced power. Your basis of comparison here is not really fair or accurate, it does not very clearly state how much progress was made on each side.

    Also, how much progress did Intel really make, and how much more can they make going forward? Do we know for a fact that Intel didn’t already have some fairly easy optimizations that they didn’t do because they were focused on raw speed, and now no longer have many easy options for further efficiencies?

    Think of it this way, if you were watching Football, and a team was down 47 to 0 at the end of the 1st quarter, you’d turn off the TV. But if you turned the game back on and saw that halfway through the 3rd quarter, the other team had recovered and the score was 53 to 48, would you say “see the other team has not caught up” or would you think that the previous losing team had a shot?

    ARM was about this far behind not that long ago, and they may not catch up, but they might. Keep in mind they do not need to be as fast as x86, they just need to be “fast enough” while being cheap, customizable, and power efficient.

  67. twilightmoon says:

    “Tablets then will be very powerful and have better interfaces than they do now. Email, web, video, music, games, social networking, news, etc., will be just as good on a tablet as it would on a beige box, plus it’s portable, so why buy the beige box? It’s the end of desktops if you have a powerful tablet that can be both mobile AND at home connect to keyboard, mouse, big display, and big storage connected to your router (or storage provided by the cloud).”

    Also I am rather puzzled at this. Why would you say “better interfaces” and then mention mice, and keyboards. Keyboards are certainly better than touch screens for a lot of text input, but they come at the cost of extra bulk and weight if you’re talking about a portable device. Mice are nowhere near as good as direct interaction for small phones and tablet devices. Email, games, social networking, news and internet browsing are already comparable or better on a tablet vs a desktop.

    Where tablets currently fall short is in areas you didn’t mention: development, photo video and sound editing, and certain heavy duty computing tasks. This is the main area where tablets will have to catch up. The iPad already does many of these, it’s really just a matter of how well.

  68. RawBob says:

    A juvenile analysis, at best.

  69. twilightmoon says:

    Market Cap 12.35B. It’s not impossible, but it would make it the largest acquisition in company history. Apple is historically *very* stingy handing out money on acquisitions, so I am skeptical. Apple would be more likely to buy 5 or 6 smaller chip design firms, or spend the money on increasing it’s manufacturing capacity.

  70. KarlWa says:

    I doubt Apple would ever buy ARM; they don’t need to. ARM is an IP company and Apple already has a (clearly very permissive) license to their chip design. ARM is profitable because they’ve colonised the market with royalty-bearing designs. If Apple bought the company, it’d almost certainly stop licensing those designs. The return isn’t big enough for Apple; it’d take a lot of money to run and not provide much benefit over their current license.

    ARM’s business model is totally different to Intel’s. Unlike the x86 platform, different ARM systems could have radically different architectures because the product is the IP. It’s a flexible enough model that companies like NVidia are claiming that they’ll have 64-bit, quad-core ARM cores before ARM themselves. There’s no way of telling where ARM cores will be in five years: even ARM themselves can’t really say for sure (evidently). There’s no reason to think Apple isn’t pushing the envelope even further themselves.

    That’s the power of ARM’s business model. By selling licenses to the “source” of ARM’s platform, they’re getting the kind of rapid, distributed development you usually see in open-source projects. Intel can’t match that flexibility as a supplier to Apple.

  71. Paul Annesley says:

    ARM doesn’t manufacture processors, they license their architecture to other manufacturers.

    Currently Samsung are manufacturing the processors for Apple, but they could move it in-house if that made sense for them.

    There’s still plenty of third party technology in Apple computers; AMD/ATI graphics, Broadcom networking/bluetooth, etc.

  72. jamesdbailey says:

    This is a fairly superficial report. While I can’t fault any of the information for the current generation and maybe next of ARM and x86, the world of CPU design is probably about to change radically. The longer term direction of CPU architecture is about to hit a wall unlike anything the IC industry has ever seen.

    We are likely near the limit of what can be done to shrink transistor size on ICs using current silicon substrates. If new technologies don’t appear soon, die shrinks after the 14 nm or perhaps 11 nm process is implemented are done. I know people have heard this before but this time we are really hitting physical limits based on the size of electrons themselves. There is early research from IBM using grapheme as an alternative but it shows no signs of being ready in the next 5 years or so needed to continue Moore’s law unabated.

    The consequence of this upcoming limit is that all of Intel’s vaunted manufacturing ability only gives them a short window of opportunity to pull ahead. Once the 11-14 nm die shrink is commonplace, any chip fabricator will be able to duplicate it even if Intel got there first. When that happens, Apple and others will be able to leverage the work they’ve done on the ARM designs to level the playing field and still retain the advantages that you’ve mentioned like the custom boutique design Apple has employed in the A4 and A5.

    Apple is well aware of this likelihood and is almost certain to factor it into any long term decision they make to switch CPUs either to ARM for MacBook Airs or away from x86 to ARM for iOS devices.

  73. KarlWa says:

    “I suspect it’s actually more variation than exists in the ARM space.”

    Not even close (http://www.arm.com/community/p

    ARM doesn’t make chips. They provide the core designs that go in SoCs. SoC stands for “System on a Chip”. It’s an entire computer (CPU, GPU, various IO and memory controllers, the memory itself, etc) on a single package.

    ARM currently design CPU and CPU core IP. The rest of it is stuff that companies like Apple, Nvidia, Qualcomm, Texas Instruments, etc get from other companies or develop themselves.

    This isn’t like a traditional PC where the macroscopic system architecture is fixed with interchangeable components – the entire system architecture can change from vendor to vendor. They can each have radically different ways of handling IO and memory, depending on what they were designed for.

    If you’re making a multi-SoC OS like WoA, it provides a challenge to support all of these. That’s why Microsoft limited themselves to 3 SoC partners that WoA will work on. It’s not Windows-on-ARM as much as Windows-on-specific-implementations-of-ARM-based-designs.

  74. Bill Coleman says:

    While an ARM chip may be about the size and complexity of a 1995 processor, it is running 10 times faster than the chips of that day….

  75. Playos says:

    The expanded instruction of x86 exists in large part to aid graphics subsystems, and virtualization systems. Intel and AMD both continue to add instructions to help graphics systems. Apple might see a great benifite from a custom high power ARM chip, but even Nvidia is having issues actually designing one that is really ahead of the curve and doesn’t explode power usage.

    These are extermely complex items and they live in the heart of their business. Spliting the Apple faithful among x86 and a custom ARM solution would seem like a really bad choice, and not in Apple’s corporate personality… and pushing everything over to an ARM system would seem like suicide if x86 systems could come with more support, equal power consuption, and more 3rd party elements.

    If Apple goes ARM, they litterally have to design and manufacture every chip and subsystem in the computer, plus coding drivers for them to interact. Could it happen? Sure, look at Nvidia, they are putting out some of the most complete ARM packages right now with manufacturing ability only behind Chipzilla and AMD, if even that, and the Tegra 2 wasn’t even that great and video decoding.
    The path forward for them seems safest to leave the Mac family x86 and push for greater power in the iOS family to zero-market the exact crowd who will be willing to take the preformance hit. Instead of a Mac Book Air, think new iBook with capable of running 2 or 3 iOS apps at the same time… That
    leverages the long term investments made before and tests the waters for
    more capable machines.

  76. twilightmoon says:

    I agree that it would not make sense to move the MBA to ARM for one or two systems, it would only make sense if they planned to move all of them there eventually. I suspect if they plan to sell Macs 5 or 10 years from now that they will all be ARM eventually, but they will start with the most portable systems where it makes the most sense.

    To think that ARM won’t eventually catch up to where Intel is today is foolish, and to think that Intel will be as cheap as ARM is now, is also foolish. ARM does not need to be where Intel is in 5 years to be worth switching to. Personal computing does not need to be dramatically faster, only cheap, power efficient, and “fast enough”.

    I think this will happen, but I do not think it is coming before 2-3 years time.

  77. Playos says:

    It’s not really foolish to think that ARM chips may not be able to match today’s Core i5/7 in the mid range (5 years or so). The gains in Ghz haven’t really paid out the big imporvements we’ve seen in x86 parts. I think it’s more foolish to belive that today’s basic requirements will maintain 5 years from now and on a much more limited architechture. Scripted langues take MORE processing power than native ones, and with HTML 5 applications coming around the bend in force we will see that today’s horespower isn’t always wasted.

  78. N8nnc says:

    I appreciate your effort, but I respectfully disagree. Intel has had plenty of time and resources to respond with lower power implementations but they haven’t. I conclude that they can’t or more likely don’t want to do it. It’s against their DNA. Their mentality is to go faster and go bigger. Apple wants more control and I believe has determined that fast is “good enough”. They can’t steer Intel, just like they couldn’t steer IBM. I suspect we will see ARM Macs, but there’s zero chance we’ll see Intel iPads.

  79. perfectlyreasonabletoo says:

    The reason you should believe Intel can pull it off is that they have the most advanced technology, unlimited money, and they’ve never tried to target a smartphone/tablet level of performance before.  No matter what, their stuff has always gone into some flavor of PC, and PCs put far greater demands on the CPU.

    I don’t know if you use Windows, but if you do, open up Task Manager and look at the Processes tab.  On my computer here, Chrome by itself is running 10 threads and is using 700 megabytes of RAM with only have 3 tabs open.  This is because Chrome is designed with a PC in mind, where maximum performance is all that matters.  But on a phone or tablet it would be designed to be capable of running with a fraction of the resources.

    If Intel doesn’t need the processor to be able to simultaneously handle 10 threads from just one program among dozens of others, there’s no question they can scale down power consumption dramatically.  Their next-generation “Ultrabook” platform that they’re pushing so hard uses just 17 watts – and it’s still a quad core chip running at between 1.6 and 1.8 GHz depending on the model, not to mention having a fairly capable GPU built in and a massive amount of cache.  And that “1.6 GHz” is clock-for-clock much faster than any “1.6 GHz” ARM design by far.

    To get an idea of what I mean, try running any cross-platform benchmark on your smartphone and then run it on a PC.

    BrowserMark on my PC: 768630
    BrowserMark on my Galaxy Nexus: 89041

    Linpack on my PC: 931 MFLOPS, completed in 0.09 seconds
    Linpack on my Galaxy Nexus: 32 MFLOPS, completed in 5.24 seconds

    SunSpider on my PC: completed in 172.7 ms
    SunSpider on my Galaxy Nexus: completed in 1870.1 ms

    This is hardly a scientific test but you can see the scale of the difference.  None of these tests are really capable of stressing my CPU anyway, in fact I doubt it even had a chance to come out of its idle clock speed (1.6 GHz idle, 4 GHz maximum).

    It’s too soon to say one way or the other if Intel is capable of pulling this off.  I think they’re more than capable if they actually commit to doing it – which they have.  In fact, there are supposedly going to be Intel-based Android test platforms coming out later this year, so we won’t have to wait a lot longer to see.

  80. perfectlyreasonabletoo says:

    Windows has exactly two things going for it, and even people who use Windows exclusively (myself included) will say the same thing:

    1) People are comfortable and familiar with it.

    2) There’s tons of software for it.

    Both of these will be gone with ARM-based versions of Windows since x86 programs won’t be compatible and the interface is totally different.

    So, why is MS porting Windows to ARM?  Because they’re desperate to get a piece of the action before Apple and Google take the whole tablet market for themselves.  But frankly, it’s a terrible idea no matter how I look at it.  I’d be interested in an x86-based Windows tablet since I’d be able to run anything I want on it, but I can’t figure out what possible reason there would be to buy an ARM one.  I am really, truly at a loss as to why they’d even bother doing this when it’s frankly such a terrible and short-sighted idea.

    “Its not exactly a hugely cut down version of Windows if its running desktop Office”
    … Come on, seriously?  Office?  That’s your idea of a benchmark?  “Can it run Office”?

  81. perfectlyreasonabletoo says:

    Windows has exactly two things going for it, and even people who use Windows exclusively (myself included) will say the same thing:

    1) People are comfortable and familiar with it.

    2) There’s tons of software for it.

    Both of these will be gone with ARM-based versions of Windows since x86 programs won’t be compatible and the interface is totally different.

    So, why is MS porting Windows to ARM?  Because they’re desperate to get a piece of the action before Apple and Google take the whole tablet market for themselves.  But frankly, it’s a terrible idea no matter how I look at it.  I’d be interested in an x86-based Windows tablet since I’d be able to run anything I want on it, but I can’t figure out what possible reason there would be to buy an ARM one.  I am really, truly at a loss as to why they’d even bother doing this when it’s frankly such a terrible and short-sighted idea.

    “Its not exactly a hugely cut down version of Windows if its running desktop Office”
    … Come on, seriously?  Office?  That’s your idea of a benchmark?  “Can it run Office”?

  82. perfectlyreasonabletoo says:

    AMD’s in the shitter as far as anything mobile is concerned.  It’ll be another 2-4 years before they’ll be able to match Ivy Bridge in power efficiency, let alone start on anything better.  They’ve screwed up so fantastically the last couple of years that at this point it’s questionable if they’ll even be able to last that long.

  83. Charbax says:

    Mac is dead. 99% of the activities all those Macbook fanboys do on their super overpriced Macs is to be in the Web browser and do a few very basic word processing, basic photo/video editing (iPhoto and iMovie stuff), iTunes, such features which can perfectly well run on an ARM Macbook.

    You are totally completely wrong on power consumption. An ARM Macbook can run 20 hours on a slimmer/lighter battery today. With a Pixel Qi LCD, it can run upwards 50 hours on a battery. Something Intel can never equal because Intel is bloatware.

    The ARM SoC to be used in the ARM Macbook (similar to iPad3) costs below $50 to build. The motherboard with processor in a current Intel Macbook Air costs over $500 to build.

    It’s a fact ARM Laptop can easily run on 10x less power than Intel Laptop when the Pixel Qi LCD is used.

    What Apple is going to do is promote an official Keyboard Dock for the iPad3, and optimize a few apps such as Safari, Office stuff, iPhoto, iMovie and few other to work smoothly in Laptop mode on the iPad3. The iPad3-Laptop thus overtakes most of the Intel Macbook Air sales at less than half the price yet more profit per sale for Apple. End of story.

  84. David Longfellow says:

    I’ll never own a Mac, period.

  85. johnbiggs says:

    John, you completely ignored the fact that Apple is going to die because it sucks and people who use Macs are mouth-breathers who smell like shoe polish and dead fish. Please address this in future posts on Cult of Mac.

    Also great post.

  86. giromide says:

    This article understandably doesn’t talk about Intel’s server processor business, which isn’t exactly hurting. The demand for servers, though not quite as hot as the demand for mobile devices, will keep Intel relevant regardless of what they do to correct course in the consumer space.
    Apple historically is devoted to RISC and to owning as much of their stack as they possibly can; however, they know how to pick their battles. They aren’t moving away from Intel for their denser offerings in the next two or three years, but beyond that is anyone’s guess.

  87. Fourthletter58 says:

    “while the first sentence would challenge many adults.”
    Blogs are really dragging down literacy standards.

  88. ZangDooo says:

    Dude is like totally rocking in like every single way. WOw.
    Total-Privacy dot US

  89. hard_candy says:

    The author doesnt address that alongside the more efficient architecture and flexibility ARM offers, price is a third determining factor. Intel’s margins on desktop class processor is much higher then those on ARM processors per mm^2.

    So even if Intel uses parts of their latest process space for mobile processors (they now use even older generation wavers compared to the competition because of that exact pricing / margin issue) they will have to ask a higher price if they dont want to take a hit on profits, and will be less competitive in their higher margin products as they can ship less of those processors on the latest generation of wavers.

    That’s like, leaving the most critical aspect out of the equation to make this argument work.

    Reality is Intel will be more likely to loose their competitive process edge to the ARM competition exactly because profits and revenue on mobile processors is rising much faster then the desktop domain Intel gets its revenue from; giving companies like Qualcomm and foundries more money to throw at R&D, and more advantages in economies of scale going forward towards 2015 and beyond.
    Thirdly, we are hitting the wall in scaling on conventional present day manufacturing technologies not too far beyond  2015, the area in which Intel excels. After 2020-2025, technological innovation can come from areas where Intel does not excel in. The process playing field will recalibrate in that timeframe.

    This article reads too much like a fanboy article, even when referencing an authority like Kanter.

  90. Andriba says:

    Missing the Point:

    This article is very informative, but all of the comments, like the article itself, miss the point that the subject of discussion is Intel’s aging technology “catching up” with the ARM processor. 

    Why does anyone think that ARM and Apple will somehow put a hold on the work they have been doing for the past several years just to allow Intel to “catch up”?

    As Steve Jobs quoted Wayne Gretzky, Apple “skates to where the puck is going to be, not where it has been”. In this, Apple is now the leader and Intel is the follower. Before Apple switched to Intel, it was dependent on IBM, after that it was dependent on Intel. That was before the iPhone and the iPad.

    Today, if Intel disappeared, Apple would survive. 

    That doesn’t mean that Apple would not be harmed by the disappearance of Intel, nor that Intel’s survival depends on Apple. It means that now, Apple has a choice. If Intel can provide what Apple needs, that will be good. But if they cannot, Apple can now turn to other suppliers.

  91. sdbryan says:

    Even on iOS almost all app developers are using Xcode. Part of the standard development process is compiling an intel binary that is run on an iOS device emulator. The ARM binary is compiled from the same source code but if Apple chose an intel processor for iOS it seems like there could be 100,000 or so apps that could be recompiled and delivered in short order.

    There were some sinister motives ascribed to Apple over their Xcode only edicts, but I think having hardware flexibility was an important factor.

  92. imajoebob says:

    And there was Virtual PC and RDC before that, so what? In 2003 anyone who talked about running OS X on a PC was laughed out of the room. Technology adapts; people adapt with it.

  93. Jack Schofield says:

    > It was on the SH4 for the Dreamcast.

    That was CE ;-)

    In passing, Microsoft has written two operating systems — NT and CE — and neither was written on an Intel chip….

  94. Mijail says:

    It’s very hard to get a meaningful “cross-platform” benchmark; you will typically end up measuring something you didn’t want to. 
    In this case, it looks like you are wanting to measure the actually achievable processing power of the CPU, but you are also mixing in the quality of the underlying runtime (does the browser do the same optimizations in each platform?), or the quality of the benchmark port (does it follow best practices in each platform? is it better optimized in one of them?), possibly even the memory subsystem.

    Now you might want to take that into consideration, since anyway we are not using the naked CPU but the whole stack, so any optimization anywhere helps. But it didn’t sound like you are thinking that way. 

    For an example of what happens when you run openssl speed on a jailbroken iPhone and some discussion, see http://hmijailblog.blogspot.co

  95. twilightmoon says:

    I think long term this is why Apple is developing ARM as a substitute for Intel for the Mac. Apple wants options and leverage. How much can Intel charge Apple for its chips if Apple can simply say “okay, never mind we’ll just use our custom ARM processors, thank you for all of your help so far, we’ll take it from here.”?

    Also, if Apple can produce chips that are suitable for it’s Macbook Airs and possibly it’s entire range, Intel will need to do *better* than what Apple comes up with to be able to compete, and better enough to justify what will certainly remain a premium price.

    So even if Apple never uses ARM in it’s products, there are several reasons why it definitely should be making the effort to make it feasible.

  96. twilightmoon says:

    Awesome, you may want to practice this line: “Do you want fries with that?”

  97. twilightmoon says:

    Very good points Playos! Again, I’m a believer in Apple coming up with something really amazing with ARM, and it may never replace x86 in Macs, but I just would not rule it out. Apple has some serious muscle now, a lot of cash, and a real motivation to completely control their computers. I could be very wrong, but we will have to wait 4-5 years to find out.

    That’s a long time in this business.

  98. David A Stephens says:

    You’re a silly ass. Apple stock topped $500 and it’s the most valuable company in the history of the world. You prove that generic Haldol doesn’t work.

  99. Gonzalo Fernandez says:

    Don’t forget the games that do require more power than regular programs. For Apple is very important this market (as has been said many times) and Intel would have the most powerful SoC, what would come great to have the best games and sweep the competition. I imagine a future where games are created for both Pcs, tablets and mobiles and then only just adjust certain parameters depending on the power of each platform because the architecture will be the same. That is precisely what Apple wants and also they can afford to spend more money on components of their products if it ensures the best possible product. It has been leaked that Apple and Intel are collaborating on improving the efficiency on x86 and the market of Apple is so big that Intel could create for them specific SoC with the graphics chip that Apple wants.

  100. shinali says:

    The only technical detail this article got right was the difference of instruction widths between ARM and x86 architectures. They have very little to do with power consumption.

    In fact, modern day desktop processors employ a microcode based approach, where the CISC instructions get translated into a set of RISC-like instructions. The translation process is completely insignificant in cost, both in terms of hardware and performance. Machines are not humans, translations and instruction decoding are one of the smallest parts in processor architecture (though not any less important than anything else).

    The reason why ARM consume less power, yet with some modest performance compared to x86 architectures because that’s how they’re explicitly designed. Modern day desktop processors have deep pipelines, complicated branch predictions schemes and so forth because the target requirements allows them a large power and hardware budget. ARM processors cannot afford so much hardware because they’re made for low-power devices, so they cannot have a deep a pipeline like Sandy Bridge or anything else. An ARM processor with a similarly complicated architecture like Intel’s latest offering would probably perform just as well with the same instruction set as its low-power brethren.

  101. SSLMatrix says:

    It’s a very interesting and modern note.

  102. Vivek Rajawat says:

    BC Parallels,
    and VMware might release an ARM version a few months after receiving
    a developer ARM Mac Book Air and Windows 7 already has an ARM version. 

Leave a Reply

Your email address will not be published. Required fields are marked *