Apple Semiconductors: Brave New Macs

Apple Semiconductors: Brave New Macs

Summary: If Apple switched from the Intel architecture now used in its Macintosh computers to something of its own design, how exactly would they go about it?

SHARE:
apple-data-chip-processor-techbroiler

Recent reports originating on Bloomberg indicate that the Cupertino, CA-based computer manufacturer may switch from the Intel processor architecture used in its Macintosh personal computers to something of its own design.

But how would they go about it?

Apple already has laid some of the foundation for what would be a costly "Manhattan Project" of sorts for the company.

Back in 2008 Apple purchased PA Semi, a "fabless" chip design firm which the company has used to create custom "A-series" microprocessor designs that have been part of the company's products since the launch of the A4 processor in the iPhone 4 in 2010, and has been used in all of their mobile products since.

Like other Systems on a Chip (SoC) used throughout the mobile and electronics industry, the A-series processors that are used in the iPhone, the iPad and the iPod Touch are based on the ARM architecture, which is licensed intellectual capital that originates at ARM Holdings plc, a UK-based technology firm. 

ARM Holdings plc doesn't build processors per se. It creates reference designs, or "architectures" which are  essentialy the basic blueprints or the "DNA" of how semiconductors work, and then licenses those basic designs to other companies which in turn use them in their own chips.

The most popular of the ARM based architectures are the ARM Cortex-A7, A8 and A9, which are 32-bit microprocessor reference designs used in chips made by Apple, NVIDIA, Qualcomm, Texas Instruments and Samsung, among others.

Recently, Samsung introduced a processor known as the Exynos 5 which utilizes ARM's latest Cortex-A15 design, and is used in Google's Nexus 10 tablet that is due to ship shortly.

The instruction set that all of these processors use is referred to as ARMv7.

While the basic ARMv7-compatible core designs are licensed, each of the SoCs made by these companies have their own unique properties, such as number of cores and clock speeds, pipeline arrangement, memory and cache configuration and the type of Graphics Processor Units (GPUs) used. 

Currently, Apple uses Samsung to "fab" its A-series chips using its own ARM-based designs. It is said to be moving away from the Korean electronics manufacturing giant to TSMC, a Taiwanese firm, due to the legal rift that has been created between the two companies.

Today, Macintosh computers use Intel chips, the exact same that are used on PCs which run the Microsoft Windows desktop and server operating sytems. Moving the Macintosh to ARM-based processor designs would be a difficult and costly task, even for a company with the resources of Apple.

The first is the issue of actually having to port the Mac OS to a new chip architecture. Presumably, this effort has already been undertaken, even if the current generation of A-series chips aren't powerful enough to meet Apple's needs on the Mac.

In the early 2000's, Apple had a secret project called "Marklar" which to use Steve Jobs' own words at the company's World Wide Developer Conference in 2005 was a "Just in case" effort to port each successive release of Mac OS X (at the time which ran on PowerPC chips manufactured by Motorola and IBM Microelectronics) over a 5-yeard period to Intel chips.

Presumably, a similar "Marklar-style" effort exists or has existed for some time for ARM someplace at Apple. 

However, unlike the PowerPC to Intel transition, where the Mac had targeted chips to actually use over its 2-year transition period, and which were proven desktop processors used in Windows computers, a transition from Intel to ARM or a processor architecture of Apple's own deisgn would be a much more difficult effort.

To run on desktop or laptops, which would need to have much more demanding, CPU-intensive application requirements, such as HD video editing using Final Cut Pro and other "killer apps" such as Adobe Photoshop and Apple's Aperture photo-editing software, the Mac needs 64-bit processors. 

Additionally, Mac OS X would probably need to be adapted to handle significantly more cores than it does now (perhaps as many as 32 or even 64) in order to "Scale out" on ARM, using clusters of cores.

Right now, the OS kernel isn't heavily parallelized, nor is Mac OS X's application architecture.

Additionally, ARM processors as they exist today only run on a 32-bit architecture. Recently, ARM announced ARM-Cortex A50, which is a low-power 64-bit chip design that is now licensable to chip manufacturers.

Advanced Micro Devices (AMD) has already announced plans to produce server chips which use the new A50 architecture, using GlobalFoundries, a privately-owned subsidiary it spun off as an independent company in 2009.

However, it does not expect to do so until some time in 2014, and it could be later, or possibly even never, considering AMD's recent financial problems and questionable viability as a going concern. So we are probably least 2 years out from seeing a real A50-based chip from anyone using designs which have never been tested in the wild.

That's a big difference from Macs leaving PowerPC and landing on proven desktop Intel desktop chips in 2006.

Presumably, Apple could license the A50 blueprint from ARM and produce its own line of 64-bit semiconductors and optimize them for use on desktop and laptop computers, and continue to outsource the "fabbing" to TSMC and other firms.

But it's much more likely that Apple would seek to buy a firm like GlobalFoundries and keep semiconductor production completely internalized.

The other option is to use more of that cash and to buy real estate in Asia or Latin America (such as Brazil or possibly even Mexico) and build new "fabs" of their own from the ground up. This would be much more ambitious then buying someone else's fabs, and probably a lot more costly.

However, it would guarantee a constant and dedicated supplier of chips for the Mac and the iDevices, something the company has been struggling to do in order to keep up with the tremendous demand for their products.

In addition to building designs based on ARM, there is also the possibility that Apple could do something totally unexpected, which would be license a different chip architecture that none of their competitors are using (such as IBM's POWER or Oracle's SPARC, both of which are powerful 64-bit RISC-based chip families) or to build an entirely new chip architecture that is exclusively for their own use.

There's a bunch of reasons why this makes sense. Apple has always wanted independence from the rest of the industry, whether it was their own operating system or by using proprietary components.

While licensing ARM intellectual capital and spinning their own custom versions of the architecture in their own chips has given them a leg up on their competitors, there is always the possibility that the British company could be purchased by an UK or European concern or possibly even another American or Asian company which would be hostile to Cupertino's interests. 

Certainly, Apple would probably like to own ARM Holdings plc and to keep the architecture from its competitors. With over $100B in cash assets, it has more than enough money to buy the company, which at the time of this writing was valued at around $15B and has been gaining traction in the stock market over the last year.

However, this would only get Apple the designs, assuming that the British Government and European Union would even permit such a thing, which is unlikely due to what would be clearly be percieved as an anticompetitive move by the company.

Even by owning ARM, Apple would still need to own fabs, and ARM is completely fabless. So it would seem that buying/building fabs, and creating a new processor design from the ground up is probably what the company is going to do, long term.

To do this, Apple not only needs to spend a great deal of money towards this effort towards building infrastructure, but it will also need to considerably amp-up its in-house design efforts and hire a lot of new semiconductor designers, presumably poached from Intel and other companies, staffed up to levels comparable to what ARM, Intel, IBM Microelectronics and Oracle has for their own unique chip design efforts.

The result will represent the 4th such transition effort for the company's flagship computer systems since their introduction in 1984. When these systems will appear is unknown. I postulated a few years ago we could see them as early as 2015, but this could be later.

Whenever they show up, they will be Brave New Macs. And they will change the face of Apple forever.

When will the "Brave New Macs" appear? And will they use 64-bit ARM architecture or something of Apple's own creation? Talk Back and Let Me Know.

Topics: Apple, Hardware, iPhone, iPad, Smartphones, Tablets, PCs

About

Jason Perlow, Sr. Technology Editor at ZDNet, is a technologist with over two decades of experience integrating large heterogeneous multi-vendor computing environments in Fortune 500 companies. Jason is currently a Partner Technology Strategist with Microsoft Corp. His expressed views do not necessarily represent those of his employer.

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

61 comments
Log in or register to join the discussion
  • Apple is going to release iPad MAXI 46'' FULL HD device and ditch MAC OSX

    The MAC OS X is not capable compete with new iOS devices and it is going to be replaced with new iPad MAXI product line, it is going to be the best choice for computer professionals and corporate users, it should drive Apple's capitalization up to $2 TRLN.
    Nikolayev
    • The Maxi Pad?

      Sure. Ok.
      William Farrel
    • Riiiiiiiiiight....

      Actually Apple is code-naming the thing "The Alan Parson's Project" and it will eventually settle on it's 8th chip preperation aka "Preperation H"... Austin Powers fans will get this...

      That makes as much sense as your iMaxi Pad
      athynz
  • Hmm

    Wouldn't it make more sense to just buy AMD? Then they could merge the two companies strengths and push high end graphics to offset the below par ARM Chips.
    slickjim
    • Possibly, if the purchase were cheap enough.

      But AMD has no fabs.
      jperlow
      • I thought they had one is Dresden

        Isn't that a fabrication facility?
        William Farrel
        • Fab 25

          In Dresden is owned by GlobalFoundries now. AMD has divested itself of all of its manufacturing capability.
          jperlow
          • I see they divested their remaining shares this year

            I hadn't realized that.

            Thanks!
            William Farrel
          • Didn't you say the A-series ARM chips were "fabless"?

            If so then AMD might be the right fit if the chip is created by a "fabless" process... I've seen first hand how a fab works at a Motorola plant back in the day - a very complex and interesting process - but I have no idea how a "fabless" process works.
            athynz
          • Ok...

            Fabless means the company has no ability to manufacture (Fabricate) anything on its own. AMD, like Apple, only has the ability to "design" processors. They have to partner with a component manufacturer such as Samsung, TSMC or in AMD's case, GlobalFoundries, in order to manufacture a chip.
            jperlow
      • AMD is very cheap

        Apple could buy AMD 10X over with the amount of cash it has in the bank.
        Shameer Mulji
        • Again

          AMD has no fabs, so other than its own x86 technology (AMD64, which is heavily oriented towards the server business now rather than desktop) and the possibility of the ARM A50 stuff and owning the ATI ICAP (again, doesn't make sense for embedded GPUs) it may not make sense.
          jperlow
    • There fixed it for you

      "Wouldn't it make more sense to just buy GlobalFoundries?"
      erick.mendes
  • Not sure that Apple will ever own a chip manufacturing factory

    The problem is that it is economically effective only if it's production scaled to like tens of billions of USA's dollars per year, and even Apple's huge consumption by far will not be able to make it.

    It is similar to assembly business; Apple is perfectly happy with taking 30% of Foxconn's over 1 million people force to itself, but never the whole company.

    Also, owning a fabric itself bears always risk of lagging behind competition. Intel versus others is primary example: while 22 nm equipment is done in Europe and, while being superpricy at $100 million per machine, is available not only to Intel, the latter is the only company which was able to use this process for universal, all-purpose VLSI/ULSI such as CPUs. The lines of Toshiba and Elpida, among others, can only use this 22nm process to make RAM, which is much easier.

    The consequence of this unique advantage Intel has is that its rivals such as Global Foundaries, TSMC, Samsung Semiconductor are severely lagging behind with only recently starting to mass-produce universal chips at 32-28 nm process. Intel is at least two years ahead of all of competition.

    So Apple could be only interested in fabric that would close the gap and give the company exclusive advantage over all of rivals that would still had to use lagging GF, TSMC, SS.

    However, there is no such fabric now and it is not know how Apple could get technology of the level that Intel has -- unless Apple would buy Intel (though it would be almost impossible due to regulations).

    Hence there is no sense in either building/buying fabrics, nor in coming up with big fast architecture chips -- Intel will not allow use of their advanced XX-nm process, and chips produced in rival fabrics will seriously lag because of much worse process.
    DDERSSS
    • Technical note: "fabrics" is rarely used synonym of "factory"

      And I am sorry for the misspellings and issues with the phrasing; unfortunately, option to edit comments is not available.
      DDERSSS
    • Intel over ARM

      While Intel does have a fab process node advantage, it hasn't really manifested itself since all of the mobile chips have lagged behind Intel's main node by a generation as well, making them comparable to TSMC offerings. For example, the Medfield SoC is still on 32nm, like Samsung, while Intel has been pumping out 22nm chips since April/May. It's not until 2014 that Atom chips actually gain an advantage over what TSMC has by using Intel's latest fab process.

      And that still doesn't solve the problem of no one using Intel chips. I consider the Motorola RAZR i to be a joke since it has no LTE and isn't a Nexus, so it won't get the latest updates or be strongly pushed by any vendor. Same with the Lava Xolo 900 and Orange San Diego.
      Jeff Kibuule
  • Some more ideas about Apple diversification

    Computer Associates is known for external growth by buying other companies. Their next target is Apple. As usual, CA has no clues of what to do with their new purchase which will sink into oblivion. Otherwise next potential bidder will be Oracle. I have heard that Larry Ellison wants to popularize relational DB. Each iPhone and iPad will have a free single user version of Oracle relational Database.
    RelaxWalk
    • Where is Oracle

      Going to come up with 300 billion in cash, even if it wanted to do such a thing?
      jperlow
  • IOS MAC?

    I don't see Mac moving away from Intel as the Virtulization functionality is very popular. I can see an ISO Mac, a laptop that will run IOS apps!
    mgiron@...
    • Virtualization

      Exists on the ARM platform as well. However, Apple has no interest in it, and doesn't really care if a small percentage of their customer base uses it.
      jperlow