Why ARM Isn't Afraid of Intel

With Intel (Nasdaq: INTC  ) finally making a push into mobile with its Medfield Atom processors, should ARM Holdings (Nasdaq: ARMH  ) and its ARMy of almost 300 licensees be shivering with fear?

Considering the significance of power efficiency in mobile computing, ARM's lead in the mobile race appears secure for the time being, even though Intel has already tapped Motorola Mobility and Lenovo as OEM partners to utilize Medfield chips. At the Consumer Electronics Show (CES) this year, interviews with a few ARM execs add some context to the technological arms race.

"Roughly good enough"
ARM CEO Warren East acknowledged that his company perceives Intel "as a serious competitor," and that it was inevitable that Intel would land some design wins within smartphones. However, East doesn't think Intel can ever be the leader in power efficiency, although "they have a lot more to offer." That last bit probably refers to sheer performance, a department in which Intel has always been top dog.

East isn't particularly impressed with Intel's smartphone ambitions: "They have taken some designs that were never meant for mobile phones and they've literally wrenched those designs and put them into a power-performance space which is roughly good enough for mobile phones."

ARM's forte has always been power efficiency, and he compares the mobile revolution to your car:

People want to do more things with their phones, but battery size remains constant. It's like having a car with a fixed-size fuel tank and you want to drive 100 more miles. You've got to make the engine more efficient. That's what we do for a living.

One example is ARM's new Cortex A7 design, which has the same performance as recent chips but sip only a fifth of the juice. That focus on minimizing power consumption is partially why Apple uses ARM blueprints for its custom A4 and A5 processors, since Cupertino puts battery life pretty high on its priority list, not to mention its marketing campaigns.

WinARM at last
East is excited that Microsoft (Nasdaq: MSFT  ) is opening up Windows 8 to ARM architecture, prompting ARM's slew of partners to start whipping up designs in hopes of a shift from Wintel to WinARM. One can't help wondering whether Intel feels a tad bit betrayed by its longtime partner in crime. Even though Windows 8 is expected out late this year, East has waited so long for this that another six to 12 months doesn't concern him.

He'd rather continue to sit tight and wait for Microsoft to get the experience down pat than compromise on the overall experience. ARM licensees NVIDIA (Nasdaq: NVDA  ) and Qualcomm (Nasdaq: QCOM  ) , among others, are also psyched about Windows 8, since their processors will have a place to call home, one that has long been occupied by Intel in PCs.

A second opinion
ARM president Simon Segars also gave The Verge an exclusive interview at CES. Segars said Intel's Medfield chip is "kind of what we were expecting," so its design wins with Motorola and Lenovo weren't surprises.

He detailed ARM's design philosophy and considers it a big advantage that ARM chips have been built from the ground up with power efficiency in mind, and they've always been packed into a single system-on-a-chip (SoC). That approach has been "deeply ingrained" into ARM's ethos; low-power has always been the starting point. ARM works to build onto performance while maintaining better power efficiency. Medfield is Intel's first true SoC offering.

On the other hand, Intel's approach has always been to start with higher-performance chips and continue to strip down performance and features to trim down power consumption. Naturally, ARM thinks its approach is easier, simpler, and better suited for mobile devices.

Separate but not equal
Segars also detailed how ARM chips are also designed to work well with other companies' intellectual property, which enables them to add features and differentiate themselves from other ARM chipmakers.

For example, some of Qualcomm's ARM-based Snapdragon chips use integrated basebands, whose additional efficiencies set it apart from other chipmakers as the only major SoC company to do this, earning the company handfuls of major design wins in the process. NVIDIA's first Tegra chip to use an integrated baseband, continuing its superhero nomenclature with codename "Grey" (as in X-Men's Jean Grey), isn't due out until later this year.

Core strength
When asked about the incredible speed at which mobile CPUs have advanced from single to dual and now to quad cores, which happened in only about 18 months, Segars brought it right back to power efficiency. Adding more processing cores adds more power suckers, and finding a way to optimally balance those cores in a way that idles them efficiently and switches them off when not in use is crucial.

ARM builds this feature into its system-level architecture to support multi-core processors. For example, NVIDIA's quad-core Tegra 3 actually has five cores; it uses this fifth "Companion" for ultra-low-power tasks while tapping the four primary cores when it needs serious horsepower.

You can't be mobile without battery life
The overarching theme in both execs' comments is power efficiency. In the days of traditional PCs like power-corded desktops, raw performance trumped power efficiency. To wow today's consumers, mobile devices need to be able to go untethered for hours on end. Ever notice how one of the selling points Apple frequently touts with the iPad is its massive 10-hour battery life?

Power efficiency is also on the forefront throughout our society as we focus on weaning our reliance on fossil fuels and move toward greener ways, which has helped ARM chips start to test the waters in the important server market. Medfield's performance benchmarks topped comparable ARM chips, but it was pitted against dual-core processors as opposed to the newest quad-cores, and it missed target goals in power consumption.

Intel will keep trying, but ARM's power efficiency advantages should keep it ahead in the mobile game.

The mobile revolution is set to become The Next Trillion-Dollar Revolution. There are lots of companies that are set to cash in on it, but one in particular has excellent prospects. The company is one of few players that will help power the mobile devices of the future, and it also has exposure to the explosive growth in China. I'm so bullish on the stock that I've given it an "outperform" CAPScallGet access to this 100% free report to find out what company I'm talking about!

Fool contributor Evan Niu has sold bullish put spreads on NVIDIA and Qualcomm. He owns shares of ARM Holdings and Apple, but he holds no other position in any company mentioned. Check out his holdings and a short bio. The Motley Fool owns shares of Qualcomm, Microsoft, Intel, and Apple. Motley Fool newsletter services have recommended buying shares of Intel, Microsoft, NVIDIA, and Apple, writing puts in NVIDIA, and creating bull call spread positions in Microsoft, Apple, and Intel. Try any of our Foolish newsletter services free for 30 days. We Fools don't all hold the same opinions, but we all believe that considering a diverse range of insights makes us better investors. The Motley Fool has a disclosure policy.


Read/Post Comments (6) | Recommend This Article (6)

Comments from our Foolish Readers

Help us keep this a respectfully Foolish area! This is a place for our readers to discuss, debate, and learn more about the Foolish investing topic you read about above. Help us keep it clean and safe. If you believe a comment is abusive or otherwise violates our Fool's Rules, please report it via the Report this Comment Report this Comment icon found on every comment.

  • Report this Comment On January 18, 2012, at 1:36 AM, jdwelch62 wrote:

    Readers Beware! Evan states in this article that Intel's Medfield chip "missed target goals in power consumption". What he doesn't tell you is that the Intel chip consumed fewer mWatts of power than both the Samsung Galaxy II S and the iPhone 4S that it was tested against (a little less than the Galaxy but a lot less than the iPhone), all while "Medfield's performance benchmarks topped [those] comparable ARM chips". So, I don't know what those "target goals in power consumption" were or who set them, but the hidden fact here is that the Intel smartphone chip both outperformed and consumed less power than the ARM chips it was stacked up against. The words Evan chose to use (and those he chose not to use) in his reporting of "facts" about power consumption is a classic example of journalistic "spin". Here's another fact that didn't get reported: the reference design phone with the Intel Medfield chip also racked up a 12 hour battery life. Yes, Intel is late to the mobile party, and has a long road to hoe, but the party's just getting started. Medfield is here now, and Haswell (Intel's 22nm Tri-Gate successor to Medfield) is right behind it. I'm going to check back with Evan in a year, and we'll see then how far Intel will have come by then in terms of design wins & actual market share. Intel is a marathoner, and the mobile processor race is not a sprint...

  • Report this Comment On January 18, 2012, at 9:02 AM, winklerf wrote:

    The problem with the ARM proponents is that they aren't looking at where mobile computing is going. The processors are going to resemble what is now a high performance chip, Sandy Bridge or the upcoming Ivy Bridge from Intel. Instruction sets might differ, but the problems of pushing a lot of data through a multi-core processor quickly is not easy. In fact, AMD's inability to keep up with Intel since Athlon suggests very clearly that it gets much harder. Phenom and Bulldozer haven't been complete failures for no reason. AMD just doesn't have the engineering infrastructure to get it done. Why do we expect a lower income firm like ARM to do better?

    The power consumption of x86 chips has been marching downward very quickly. The TDP, an Intel term for the absolute maximum power draw a processor can accomplish, of the Sandy Bridge ULV chips was the same as the original Atom a few years ago. High performance computing is coming to the mobile space and ARM has a lot of work to do on their architecture.

    None of this takes into account the fact that TSMC and the IBM alliance are having a harder time keeping up with Intel in manufacturing every year. Their processes are lower performance and also lag Intel. Mostly, they have been shrinking as quickly, but including fewer features that improve power and performance. The TSMC CEO has come out and said he doesn't know how much longer they can keep up this pace.

    If ARM isn't actually concerned about Intel, then they are foolish. A lot of other companies that weren't are dead and buried.

  • Report this Comment On January 18, 2012, at 9:31 AM, TMFNewCow wrote:

    @jdwelch62,

    For additional detail on how I derived that statement, I was referencing the Caffeinemark 3 benchmark, where the Medfield scored 10,500, higher than the Tegra 2's 7,500 and Snapdragon's 8,000. It used 2.6W idle and 3.6W during HD video playback, while the target goal was 2W and 2.6W, respectively.

    http://www.gottabemobile.com/2011/12/28/intels-atom-based-me...

    That being said, you and fidgewinkle have very valid points, and I agree that the reliance on TSMC can be limiting and even a potential liability down the road. Either way, I still feel confident that ARM has gathered enough momentum that Intel will have a lot of difficulties breaking into mobile. It can certainly be done, but I think odds are stacked against it.

    I appreciate the substantive comments, since we Fools love a good healthy debate with different viewpoints. The next couple years will be very interesting to see unfold in the mobile CPU space.

    Fool on,

    -- Evan

  • Report this Comment On January 19, 2012, at 4:01 AM, hkwint wrote:

    @jdwelch:

    Currently, there are no real comparisons between the Intel SoC and competing SoC's as far as I know. You can only compare one SoC to another one if you 'remove' all other components, and as far as I know that's not done.

    12h battery life is totally meaningless if it doesn't specify what task load, what screen is used, which brightness settings, what 3G / 4G chip, if radio chips were turned on / of, and so on:

    Besides, one should notice that it seems this Intel SoC is not yet available in mass quantities.

    @fidgewinkle:

    Currently, Cortex CPU's support some advanced functions which Atom's didn't. So of course, ARM sees where this all is going. If you need to put lots of data through and you can do parallel tasks, one uses the GPU (OpenCL) and not the CPU.

    You should remember, the CPU is only a tiny part of the entire SoC.

    The part about TSMC is also a little bit off I'd say, they're doing pretty fine, they can run Cortex A9's they produce on 3gHz if they want (but don't for power constraints). And besides, they only make about 1/3 of all ARM Smartphone-SoC's, as in contrary to Intel SoC's, those are _not_ dependent on just one foundry at all.

    There's the Samsung/GloFo/IBM/ST-M alliance, and nowadays they're interchangeable. So you can move from one factory to another.

    Then there's UMC Taiwan, where TI OMAP's are made.

    And there's TSMC. Apart from probably some other smaller facilities. So in contrary to Intel, there's choice for the foundry, and competition amongst them.

    And to the question why we expect ARM to do better than AMD: That's a totally senseless comparison. ARM doesn't make chips:

    It's the whole coalition of TI, Broadcom, Samsung, Apple, Renesas, LG, nvidia, RockChip, AmLogic, AppliedMicro, Creative, ST-Ericsson and more companies. Those combined are a little bit larger than "just AMD", and have much, much more experience in SoC design than Intel does. Besides, 6 billion+ chips a year on ARM are shipped, a number Intel simply can't match, they're not up to the task.

    This ARM-system provides much, much more choice than Intel on their own could ever offer.

    ARM-chips also find their way in cars, TV's (where Intel was recently ditched), coffee machines, hard drives, network cards, set-top boxes, healthcare devices, PND's and so on. A market where Intel is no player at all at this moment. So even with some competition in the smartphone-market they still billion of chips based on their architecture will be sold.

    Disclaimer: Currently I hold some ARM-shares.

  • Report this Comment On January 19, 2012, at 10:04 PM, winklerf wrote:

    @hkwint

    Comparing AMD to ARM does matter if you want to talk about architecture. Especially since AMD doesn't do manufacturing anymore. AMD has repeatedly flubbed the datapath in their architectures. This is design, not process technology. It is a part of the architecture which ARM is utterly inadequate at. Further, it is the part of the architecture which GPGPUs from nVidia and AMD are proving ineffective. Whenever one is trying to shove multiple instructions through an architecture simultaneously, whether it be through SIMD, MIMD, or multiple issue, the hard part is managing the data as it goes through. This is because the data moves slower than the processor and keeping the processor busy isn't a simple task.

    As for TSMC's process technology, it might seem to a layman that they are competitive when they aren't. I will neglect UMC, because they aren't superior in any way to TSMC. One might think TSMC is only about 1/2 a shrink behind since their 28nm and Intel's 22nm are coming out at the same time. However, if you look at the product that is being produced on TSMC's 28nm versus what is being produced on Intel's 22nm, then it is evident that the development cycles aren't comparable. Ivy Bridge had functioning silicon a year ago on 22nm. It takes a year after first (functional) silicon for a cutting edge microprocessor to become a quality product. The redesign is usually in the metal layers, which minimizes the number of masks that need to remade. It is a function of the chips being too big and complex to simulate well enough to get a final product on the first run. Everyone has this problem to some degree, but GPUs are just a lot of repetitions of the same thing and ARM CPUs are just not that big or complex. So, Intel is pretty much 2 years (a full process node) ahead of TSMC, their closest competitor with regard to shrink at least.

    To further their manufacturing lead, Intel also has the best performing transistors. They have the largest gain, which translates to faster with less power consumption. Further, they bring in features like metal gate and fin-fets, not just earlier than the competition, but at larger feature sizes. For example, TSMC is introducing Hi-K metal gate at 28nm just now while Intel brought it in at 45nm four years ago. When do you think they are going to get fin-fets? All of this translates to more efficient performance and better shrink in the case of fin-fets.

    Why may you ask is Intel only now bringing a chip into mobile that might be competitive? Most importantly, Atom was designed as a low cost chip, not as an efficient one. It has been designed on lagging technologies for its entire lifespan. Medfield being on 32nm is the closest it has been to a cutting edge process. Too bad it is a full process node behind Intel's best even though it is a simpler design than Ivy Bridge. The reason this is the case is Intel has changed their attitude from "We can just throw a low cost chip at this" to "We are going to put this on our best process technology". Part of the delay in getting up to their latest process is that they are working at the SoC and power gating challenges, which aren't their historical expertise. However, Medfield suggests that they are improving in these areas.

    While Intel has been testing their weaknesses and showing some signs of improving, ARM has shown nothing in silicon, through a partner or whatever, that they can address their datapath shortcomings. Calxeda is doing something with their instruction set that might pan out, but it won't be ARM architecture. It still could help ARM, but it is quite a long shot considering we've been down this road before with the likes of Transmeta. The takeaway message is that Medfield isn't Intel's best shot. It will only get harder from here as they put a priority on the mobile space and move the designs on to their latest technologies and improve their power gating and SoC expertise. There is no doubt ARM will improve their architecture, but AMD's repeated failure in architecture suggests that the prospects aren't good, especially when you have to play this much catch-up.

    As for the parts of the SoC that matter, there are two, the CPU and GPU. Everything else is trivial by comparison. Further, the point where it becomes CPU and vector math co-processor is fast approaching since there are limits to the need for rendering while the needs for flops will keep on growing. This means that the GPU blocks will end up more as specialized CPU blocks. In this arena ARM will become very reliant upon nVidia.

    As for the momentum that ARM has as far as market share, the barrier to entry in smartphones and tablets is microscopic. App markets are springing up all over the place. Apple, Google, Amazon and Microsoft come to mind. Further, most of the code isn't compiled so it can run on any architecture. Any processor that is superior in some way can have a piece of this market. All you need is a decent relationship with one or more of the markets and you're in. Big deal.

    The real head scratcher here is that ARM has a price of 15x Gross Revenue, not profit, revenue. It is already baked into their price that they are going to dominate the mobile processing market and take most of the computing market away from x86. They might be able to make a really big revenue move if they take a lot of market share, but they are going to need to beat x86 silly to get it. There are really a lot of people that don't understand the electronics industry and don't understand all of the challenges that think ARM will win because it is the big mover right now. Sun-Sparc, IBM/Moto-Power, DEC/HP and many others were all the big movers at one time too. They had multi-company alliances to try and defeat Intel-x86. In the end they all lost to the process technology advantage Intel had over them.

    Finally, I've worked with IBM alliance(includes Samsung, Global Foundries, and UMC as well as others) and TSMC process technology. Mostly, I learned how things should not be done. From what I've heard about how Intel does things, they do it much better than their competitors. This goes beyond the fact that they outspend them by a multiple greater than two on R&D the last I heard. I suspect it has gone up from there.

    Disclaimer: I hold INTC shares

  • Report this Comment On January 22, 2012, at 2:45 PM, dividendgrowth wrote:

    The correct bet for investors is to buy both ARMH and INTC.

Add your comment.

Sponsored Links

Leaked: Apple's Next Smart Device
(Warning, it may shock you)
The secret is out... experts are predicting 458 million of these types of devices will be sold per year. 1 hyper-growth company stands to rake in maximum profit - and it's NOT Apple. Show me Apple's new smart gizmo!

DocumentId: 1759385, ~/Articles/ArticleHandler.aspx, 10/25/2014 2:34:57 PM

Report This Comment

Use this area to report a comment that you believe is in violation of the community guidelines. Our team will review the entry and take any appropriate action.

Sending report...


Advertisement