ARM CEO foresees netbooks taking over

Could ARM-based netbooks one day dominate the industry? In an interview with PC Pro, ARM CEO Warren East predicted that netbooks could eventually make up 90% of the PC market, and he expressed a desire to see his company’s ARM application processors power a significant portion of those systems.

East sees the netbook takeover happening "over the next several years." As he points out, today’s netbooks may have Intel microprocessors, but they already include ARM logic, too: "chances are it’s an ARM in the Wi-Fi or Bluetooth. More often than not there’s an ARM in the hard disk drive and sometimes there’s an ARM in the integrated camera as well." The next step, in his view, is for ARM-based CPUs to replace the Intel chips.

We’ve certainly seen a few ARM-based netbooks (or smartbooks) recently, but because ARM processor cores aren’t x86-compatible, Windows 7 doesn’t run on them. That means smartbook makers have to opt for the less-capable Windows CE, a Linux distribution like Ubuntu, or Google’s Android or Chrome OS operating systems.

Speaking on that topic, East stated, "But actually the trajectory of progress in the Linux world is very, very impressive. I think it’s only a matter of time for ARM to gain market share with or without Microsoft." He added that there’s little point in ARM pleading Microsoft to make Windows run on its processors—"Microsoft knows us very well, it’s worked with us for the past 12 years, all its mobile products are based on ARM." East acknowledges that getting the extra support from Microsoft wouldn’t hurt, though.

Comments closed
    • Manabu
    • 13 years ago

    WinCE is walking alongside WinMobile for some time. The latter is for touchscreen, the former isn’t. I doubt Microsoft will completely stop WinCE development.

    • Manabu
    • 13 years ago

    The RISC “instruction set” inside Intel’s and AMD’s processors is probably very different from ARM instruction set. That means that you would need an decoder anyway, or would need to redesign from scratch anyway.

    And right now, who wants an high performance, power hungry ARM? People generally buy intel/amd because x86. If not, then they can chose PPC, SPARC, MIPS, and other risc architetures more optmized for speed.

    ARM is aware that optmizing a bit for performance will be needed if it wants to conquer netbook/laptop markets, and for the first time anounced an performance optmized hardmacro of an Dual core Cortex A9, consuming an whooping 2W in GP process, only for the CPU! It may even need an heatsink!

    But that is already more than enought for those aplications. We are entering in the good enought computing era, where any 2Ghz dual core is already very fast for most common tasks. Power efficiency is more important than more performance, and that is why ARM is becoming relevant again. They need to continue optimizing for power efficiency.

    • ludi
    • 13 years ago

    Yup. In fact I read past the headline.

    • ludi
    • 13 years ago

    I see them quite a bit, Aspire Ones and a handful of Dells mostly. I expect the average college campus has /[

    • blastdoor
    • 13 years ago

    Yeah, and I’ve seen literally NONE either in my house, at my office, or on the way to work. So I bet they don’t even exist!

    • jcw122
    • 13 years ago

    Netbooks will never take over. People say they are becoming popular, but I’ve seen only 2 over the past several months on a college campus.

    • blastdoor
    • 13 years ago

    I tend to agree…

    One could argue that much of this was true back during the 1990s RISC vs CISC wars, and that Intel clearly won back then. The difference, of course, is that power consumption was not as big a deal back then, and Intel had a substantial lead in fabs.

    Today, power consumption is a huge issue and there are a lot of very well capitalized foundries out here, so Intel’s investment lead isn’t as big. And ARM is a stronger competitor than PPC was, because it’s so widely licensed — you aren’t totally dependent on IBM for designs or for fabrication.

    I think Intel is going to lose this time around, and as we see more and more software written for ARM, we may even eventually see ARM move into laptops and eventually desktops (although that’s several years in the future).

    • djgandy
    • 13 years ago

    You are cherry picking and ignoring markets where atom is perfectly capable and successful. Computing doesn’t just accelerate in terms of raw computing power you know.

    • Bombadil
    • 13 years ago

    The Tualatin was severely limited by its slow bus. Just shrinking it would be completely pointless. All the good parts lived on in Banias and Dothan and I’m sure some is still there in Westmere.

    • JustAnEngineer
    • 13 years ago

    Yep. Atom has about the same performance as a 10-year-old desktop processor. The problem is that you’ve also got to give up 10 years of progress in software to remain compatible with Atom. That probably also leads to a regression on the software side as developers work to dumb down things to be compatible with Atom rather than bringing us new and innovative software that fully exploits the power of modern 64-bit processors.

    • djgandy
    • 13 years ago

    That’s exactly what I said.
    And I also said Sandy Bridge is a low power centric design.

    Would a PIII get outrun by an atom in terms of performance? Also remember atom is in the mW area (Moorestown) and it is based on the old P6 arch so technically it is an improved die shrunk older processor 🙂

    • djgandy
    • 13 years ago

    Yep that is how WinCE works and it’s a complete PITA.

    • vipw
    • 13 years ago

    This is exactly how Windows CE is sold. Not that I think Microsoft is smart — you can look at the operating system design and know that isn’t true.

    • Shining Arcanine
    • 13 years ago

    It will not happen period. Intel never die shrinks old designs after more than one process generation. Just imagine how much power they could save if they die shrank the Pentium III.

    At Intel, their mantra is basically performance first and efficiency second. If they cared about power consumption, they would have die shrunk their older chips and given us low power processors a long time ago.

    The 1.4GHz Pentium III Tualatin had a 32.2 TDP and was on a 130 nm process. If they die shrank it to a 32 nm process, the die size would be 6% of the original chip’s die size and correspondingly, I would expect the power consumption to scale with that. (32/130)^2 * 32.2 watts = 1.95 watts.

    • MadManOriginal
    • 13 years ago

    I’m not a Windows fan or hater just a realist. Apparently the world *does* care if their ‘computer’ runs Window, as I said just look at the quick turnaround in Windows dominance on netbooks. How do you explain that hmm? Now maybe in the future that will change but there’s a lot of ‘future’ out there so it’s easy to just say that as a guess.

    To repeat myself on the other point, for more CE-like devices with more closed ecosystems, whether by firmware or software, ARM is great. If someone wants to try to argue ‘but if it runs Linux it will be open!’ doing so puts things right back in the ‘complicated computer’ arena. Let’s not be stupid and argue for ‘open Linux’ on the one hand and ‘easy to use CE devices’ on the other.

    • stmok
    • 13 years ago

    I’ve seen Apple fans be completely deluded by their beliefs. Now I see a Windows fan completely clouded by his own narrow perspective of the world.

    The world doesn’t really care if its Windows or not. ARM is in no rush to get something into market to establish a marketshare ASAP. (Time and time again, its proven that long term strategy of consistency is superior to short term “try-fail-retry” approach.)

    The world only cares if a solution does what they want with relative simplicity.

    Effectively, if you offer a solution that is tuned to the consumer; it wouldn’t matter if it wasn’t Windows.

    That’s why to ARM and its CEO; *[

    • Sahrin
    • 13 years ago

    Did you?

    Headline: “ARM: our netbooks will fly with or without Windows”

    Sounds like someone responding to a customer’s demands. No, that can’t be it. It sounds like someone pontificating on their own superiority. There, that’s right.

    • djgandy
    • 13 years ago

    But for a netbook < 5W is a fine, which a low power, lower clocked die shrunk nehalem could do. That wouldn’t happen though, they’ll use a low power design like Sandy Bridge instead.

    • Shining Arcanine
    • 13 years ago

    All x86 processors have the penalty of an extremely power hungry decoder. Even if die shrinks allow for a lower power version of one, a processor designed with similar specifications and a RISC ISA will have lower power consumption and similar performance. There is no getting around that.

    Putting it another way, people want computers that are light, high performance and have long battery life. x86 has a performance advantage because of Intel’s massive investment into it but its CISC design makes it inherently weak at weight and battery life because the more power a chip draws, the lower a system’s battery life will be and the more weight will need to be added to support the same battery life with a larger battery. x86 processors are at an inherent disadvantage because of their decoders’ power requirements.

    • puppetworx
    • 13 years ago

    I think 90% is towards the high end of possibilities. The reality is 90% will not come about unless the core netbook hardware/experience is improved. If ARM can deliver a better platform for netbooks and gain market share this will start a great technology race with Intel.

    I can’t help but think about the MaxiPad. It’s a device which will really capture the mainstream netbook audience and steal customers from Microsoft. With Google and Apple innovating massively in the rapidly expanding mobile computing department I think Microsoft has to step in. On these tiny mobile screens the WIMP environment really isn’t ideal. I hope we see some of these rumours about a Courier device or OS come to fruition real soon.

    • Shining Arcanine
    • 13 years ago

    It helps somewhat, but they still need to be able to deal with legacy single threaded code and computations whose algorithms cannot be multithreaded. For those two cases, they will need higher level Instruction Level Parallelism.

    Since Intel’s processors are technically a RISC core behind a CISC decoder, it should be theoretically possible to adapt Intel’s RISC core to the ARM instruction set and drop the decoder, which would yield lower power consumption and x86 level performance with the ARM instruction set, but there are only two ways of doing that. One way is for Intel itself to do it, which will happen when hell freezes over. The other is for some third party to develop a similar performance RISC core that implements the ARM instruction set, which would require a massive investment in terms of resources.

    Designing a high performance processor using a non-x86 instruction set is something that many companies would be unwilling to do because of the history involved with doing that. Intel’s resources have repeatedly allowed it to design new chips to move ahead of its competitors, leaving no market for other processors unless they also happen to implement x86, in which case, they would be delegated to the status of a budget alternative, a position no company wants to have. In addition, Intel’s chips are so widely deployed that they can add extensions to their ISA to compensate for deficiencies in their own processor designs. However, companies whose processors implement ISAs that are widely licensed cannot do that because they are one small slice of that ISA’s market and unless their processors become a big portion of it, no one will want to support their extensions because very few people would likely benefit from such support in the first place.

    Although I would like to see ARM processor performance reach parity with x86 processor performance, the reasons I have stated make me think that is unlikely in the near future. The best chance any company will have for achieving this would be when the miniaturization barrier is hit and die shrinks are no longer possible, at which point the CISC decoder in x86 processors will become a liability as processors will need to make the most of the transistors that they have, because adding more as a means to greater performance will become increasing expensive as the die size increases. This would also neutralize the process advantage Intel has enjoyed over its competitors, allowing potential competitors to more confidently invest in processor development with the knowledge that a superior design will not be surpassed by an inferior design through a die shrink.

    • poulpy
    • 13 years ago

    He gave a quick and dirty 90/10 split of the PC market -aka Personal Computer- do you really think engineers working on Solidworks or equivalent are that common?

    Your average user will check emails, use facebook, watch some dumb ass video on the internet, listen to MP3s, chat with skype and maybe play WoW or the Sims.

    You have a market for high-spec workstations or game oriented PCs, just they aren’t the average computer the masses need.

    • djgandy
    • 13 years ago

    In 2 shrinks time (3-4 years) we could have equivalent of 2 core Nehalem’s for netbooks providing there are cutdowns in cache and < 2 GHz clock speeds.

    The netbook is all about balancing the price for the market. They are slightly underpowered at the moment, but they are capable enough for their audience and target application. In 2/3 years time they will be more than powerful enough for their purpose though and other purposes.

    We always need faster computers, but lately we are at the stage where for casual use you really don’t and even for gaming. A 2006 mid spec (E6600 or similar) dual core is still more than enough for most peoples daily use.

    • IntelMole
    • 13 years ago

    In some respects, the general movement to simpler multi-core in the next decade will help them a lot in this regard.

    The Atom being completely in-order is but the first sign of this.

    ARM, being unencumbered by x86 legacy requirements, can in theory put smaller simpler cores on a chip. To avoid putting too much decoding logic on a chip, Intel opts for fewer SMT-capable cores. However, this introduces it’s own complexity.

    So with the move to massively multicore on the horizon, with the same fab tech, for the same performance, an ARM design will be smaller, cheaper and cooler.

    In theory anyways.

    There’s the obvious requirement for software to catch up, but as a software developer, I think it’s not impossible to do. For instance, developers coding sorting algorithms may begin to favour merge sort over quicksort, because one is implicitly a divide-and-conquer algorithm ideal for parallel execution. Another thing that will help is the “everything is a thread” mindset. I work with a functional/scripting language that has a single-threaded main control loop, but allows you to run a function over an input vector in parallel by simply using the parallel version of the command. It gets you to thinking that whatever non-serial dependencies in a function you have can be split into functions themselves and run in parallel this way.

    This is all subject to Amdahl’s law of course, but like I say, this is not insurmountable.

    • anotherengineer
    • 13 years ago

    All engineers are concerned about what software it can run, cant do work if it cant run it, but I guess according to you we are not average user.

    I guess all PC gamers are not avg users either, I want your to borrow your crystal ball please.

    • Shining Arcanine
    • 13 years ago

    I think ARM will need to improve the performance of their processors to have the same Instruction Level Parallelism people are accustomed to getting from Intel processors before they move beyond embedded systems, but even without such major advances, they will likely be able to take a good portion of the netbook market, as netbooks do not require large amounts of processing power. Despite that, I imagine that ARM would likely need to modify its ISA to be 64-bit because few people want the lower number crunching performance of a 32-bit processor when 64-bit processors are available.

    Personally, I would like to see ARM processor performance rise to the level of Intel Core i7 processors because that would mean that computer processors will no longer need to use heavy duty instruction decoders that waste large amounts of energy. However, if it happens, I expect that it would come from a company that licensed the ARM instruction set rather than ARM itself.

    • OneArmedScissor
    • 13 years ago

    Protip: The average person doesn’t give a rat’s ass about a single one of those things.

    • anotherengineer
    • 13 years ago

    “ARM CEO foresees netbooks taking over”

    taking over what? desktops?? So I can play games on a netbook?

    Workstations?? So I can use Ansys, acad, solidworks, etc on a netbook??

    Do they have a nice 24″ IPS panel for doing photo editing??

    Hmmmmm I wonder what he sees netbooks taking over??

    • IntelMole
    • 13 years ago

    If these netbooks are extra smart, does that make them smARMbooks?

    OK, I’m done.

    • Shining Arcanine
    • 13 years ago

    They might be able to get around that if they made VIA a subsidiary of Nvidia. Then legally speaking, the license would not change hands.

    • kuraegomon
    • 13 years ago

    Those are the conditions as stated … but remember that little AMD settlement recently? And the assorted investigations (FTC et. al.) into Intel’s business practices? I suspect that even Nvidia _could_ manage a VIA buyout/x86 license terms re-negotation, given the current attitudes regulatory bodies and the Justice Department have towards Intel.

    The AMD settlement is a telling statement on Intel’s internal assessment of the gravity of the charges against it. Even Chipzilla doesn’t pony up a billion+, just to make a rival on life-support go away – unless they NEED to avoid litigation that would endanger their defence against anti-competitiveness charges levied by the FTC/et al.

    • blastdoor
    • 13 years ago

    I think that talking about this in terms of “netbooks” is too narrow. I think the real point here is that in the future, an ARM-based CPU could be credibly used in 90% multifunction computer-like devices, of which netbooks are an early, and kind of awkward example. These will be devices running Android, iPhone OS, or some Linux variant. They might also run some version of Windows, if MS ever gets its act together.

    And I’m not saying that it will be used in 90%, just that there’s no reason it couldn’t be. As time goes by, fewer and fewer people are going to need a CPU that costs more than $20. That really is the world of ARM.

    One other point — I think ARM is very smart to license their technology to all interested parties rather than jealously guarding an instruction set the way Intel does with x86. Intel’s bad attitude is hurting x86, which in the long run hurts Intel.

    • dpaus
    • 13 years ago

    If Microsoft’s smart (OK, OK, stop laughing), they’ll make WinMo7 (or, more likely, WinMo8) a test case for their new modular approach: build a tiny, fast core (ala QNX) and make everything else an optional service. As smaller devices become more powerful, more and more of the services can be comfortably run until you have the “full” Windows experience.

    • derFunkenstein
    • 13 years ago

    don’t x86 licenses expire if the holder changes hands?

    • Kurotetsu
    • 13 years ago

    VIA’s x86 license won’t transfer if it is purchased by another company. A partnership might work, but Nvidia has already made it clear that they’re not interested in VIA as anything but a bargaining chip.

    • WillBach
    • 13 years ago

    NVIDIA will have to buy North Korea and threaten the U.S. with nukes before Intel gives them an x86 license. Just sayin’.

    • BoBzeBuilder
    • 13 years ago

    What you guys are saying makes sense, as long as the same lazy programmers don’t decide to cut features instead of improving their code.

    • poulpy
    • 13 years ago

    I agree with the others, it’s exactly what I want them to do -to a certain extent.
    That’s the way you end up with well optimized programs -making sure it runs fine on slower hardware, the other way around and you’ll end up with a Crysis for breakfast mate..

    • BlackStar
    • 13 years ago

    Yes and yes.

    Personally, I’d be happy with well-designed web apps (WebGL/Canvas games, HTML5 video) for almost everything – but it’s Linux so ChromeOS will be able to run pretty much any Linux/Unix-based app.

    As long as it can play fullscreen 720p video, I’m sold.

    • quock
    • 13 years ago

    Why not? Targeting a machine with a slower processor and less memory might mean that the programmers will be spending a bit more effort in writing software that’s smaller and efficient.

    • hassan272
    • 13 years ago

    Nvidia should just buy up VIA for the x86 license, and make a tablet using an x86 CPU along with Tegra2.

    • jstern
    • 13 years ago

    I was recently thinking about something similar. How a program which required a 50Mhz processor back in the day would require something like 500Mhz now, even though the features are similar. Or how a Super Nintendo game ported to the GameBoy Advance would be 5 times bigger, even though it’s basically more or less a carbon copy of the game, with less sound channels. So in some cases we can actually end up with more efficient programs if programmers also targeted netbooks. It’s probably why flash is soooo horrible and disgusting, inconsiderate programmers are probably not concerned about their simple programs being efficient.

    I don’t know much about programming, etc, so it’s just an observation.

    • SPOOFE
    • 13 years ago

    /[

    • MadManOriginal
    • 13 years ago

    When there’s a real Windows OS for ARM maybe – like it or not see Window’s quick turnaround to market dominance in netbooks. Until then ARM will be great for CE devices.

    • wira020
    • 13 years ago

    I cant agree that netbook will take over the market anytime soon but i do wish that desktops will become smaller and powerful… and cheap of course… I dont care whats the form factor, if it;s powerful enough, i’d use it..

    • BoBzeBuilder
    • 13 years ago

    Netbooks suck. I don’t want programmers target the Atom processor and 1 gig of RAM.

    • ludi
    • 13 years ago

    Did you read the article?

    • Sahrin
    • 13 years ago

    This guy has 100% the wrong attitude. Your business is about your customers, not about your own success. You need to be telling your customers how your going to offer a superior platform, not how you’re going to crush your competitors. Joe Computer Buyer doesn’t care about that.

    • JdL
    • 13 years ago

    I agree that netbooks will take a significant share from traditional laptops. However, without x86 support, ARM will have no place in them.

    • Bombadil
    • 13 years ago

    Normal Windows laptops have always been stupid devices but are increasingly popular. Windows CE certainly needs updating, but it has strengths too. Will ChromeOS have instant on and third party applications?

    • adisor19
    • 13 years ago

    I see what you did there..

    Adi

    • NeelyCam
    • 13 years ago

    I guess he’s admitting that Intel will have some of the market, and agrees to call all the devices with the name “netbook”.

    Besides, depending on what “netbook” means, I think this guy’s prediction is off. The current ‘netbooks’ are too slow to be used for anything semi-serious, and for your random every-day surfing needs, smartphones will take over after performance increases. Meanwhile, compact laptops are squeezing ‘netbooks’ from the higher-end with significantly higher performance and reasonable cost/size/battery life proposition.

    • CheetoPet
    • 13 years ago

    Would those cost an arm AND a leg?

    I’m done.

    • Beomagi
    • 13 years ago

    (leaping too far…)

    Next thing you know, arm will have “desktop versions” of their chips pushing 130 Watts…

    • nagashi
    • 13 years ago

    I thought the industry had decided to use the term ‘smartbooks’ for arm-based netbooks? Did that term get ditched due to some random litigation or something?

    • StashTheVampede
    • 13 years ago

    Windows CE is already dead to MS. They are working toward WinMo7 and I hope they get their shit together.

    • BlackStar
    • 13 years ago

    Windows CE is a joke. ChromeOS FTW.

    • StashTheVampede
    • 13 years ago

    Android supports multi-touch, yay!

    Tegra2 will play H264 video and flash, adding Android’s multi-touch AND a webcam will really bring the tablets/netbooks to the masses.

    • poulpy
    • 13 years ago

    For the vast majority of people some well polished Android + Tegra2-like in a Netbook or Tablet form factor would be more than enough.
    Android is still work in progress and needs more professional 3rd party apps but it’s going places IMO and the more exposure it’ll have the more of those will come.

    Now if you want to do serious gaming, video editing, 3d rendering then sure by all means get something with more oomph.

    Edit: it’s probably more Chrome than Android that will take care of that market

    • Hattig
    • 13 years ago

    And they also announced their Eagle processor today, which will be positioned above the Cortex A9 when it comes out, but not this year.

Pin It on Pinterest

Share This

Share this post with your friends!