Computex 2013: Intel makes case for Atom in tablets and smartphones

Computex 2013: Intel makes case for Atom in tablets and smartphones

Summary: After many false starts, Intel may finally be gaining momentum in mobile with several wins in tablets and hybrids announced this week at Computex, and new Silvermont processors and a global 4G LTE modem on the way.

SHARE:

One day after the launch of the fourth-generation Core processors, Intel held a separate event focused on its Atom processors for tablets and smartphones. The company didn’t break any news, but Dr. Hermann Eul, head of Intel’s mobile group, explained the company’s strategy to compete with ARM and why it may finally be gaining momentum in mobile.

IntelMobile1
Intel's Hermann Eul explains the company's strategy to break into smartphones and tablets at Computex.

Eul said that Intel’s strategy is to develop “leadership products” (from entry-level to high-performance), provide complete platforms, and deliver the scale for mobile markets. “You all know that Intel can create markets,” he said. The foundation of all of this, he said, is Intel’s architecture and process technology.

After years of false starts this strategy seems to be paying off with a steady increase in the number of Intel-based tablets and smartphones in the market. And it is paying off with positive reviews. Eul said that the current Atom single- and dual-cores are very competitive with the best dual-core ARM solutions. (ARM disputed this at its own press conference earlier this week.)

Intel is now shipping Clover Trail+, a dual-core processor that delivers twice the performance (three times the graphics capabilities), lower power and longer battery life than the previous generation. Lenovo’s K900 smartphone was the first device to use Clover Trail+ (the 2.0GHz Atom Z2580). The first device in the market to use Clover Trail is the with 2.0GHz Atom Z2580. Asus CEO Jerry Shen came onstage to talk about two new products, the MemoPad FHD 10 tablet and Fonepad Note FHD 6 phablet, which both use the 1.6GHz dual-core Atom Z2560 processor. Eul mentioned several other products using Atom--several of which were announced this week--including the Acer Iconia W3, Asus Transformer Book Trio, Lenovo IdeaPad Mix 10, Samsung Galaxy Tab 3 and ZTE smartphones.

Intel announced its next architecture, Silvermont, on 22nm with Tri-Gate transistors, a few weeks ago. Silvermont will be used in variety of platforms including Bay Trail for tablets and Merrifield for smartphones. It will deliver three times more peak performance or five times less power. The first Bay Trail tablets running Windows 8.1 or Android will be available by the end of the year. Eul said these will have twice the performance of the current platform, better graphics and about 8 hours of battery life. Intel demonstrated the performance of a Bay Trail tablet on video transcoding (2.5 times faster than Clover Trail) and playing a “full PC game,” Runic Games’s Torchlight 2.

On smartphones, Merrifield will deliver 50 percent better performance, longer battery life and more advanced imaging features than Intel’s current platform, Eul said. The first Merrifield smartphones will ship next year—expect lots of announcements at Mobile World Congress in February.

What’s really holding back Intel on smartphone isn’t performance, or even power; it is the lack of a global 4G LTE baseband. Intel is currently shipping a 3G baseband, the XMM 6262 and will begin shipping an LTE multimode baseband, the XMM 7160, in “coming weeks.” Eul said the chip will be smaller and use less power than competing solutions. The XMM 7160 will be used not only in smartphones, but also in tablets, 2-in-1s and standard clamshell Ultrabooks. The Galaxy Tab 3 10.1-inch tablet will use the 1.60GHz Atom Z2560 (Clover Trail+) and either the XMM 6262 or XMM 7160. Eul demonstrated both a Bay Trail tablet and an Ultrabook running on Far EasTone’s 4G LTE test network. Intel is also working on a single-chip solutions with integrated 3G and 4G modems, but it won’t comment on the timing.

Intel has yet to land a global hero phone, but its mobile business is clearly headed in the right direction. Altogether Intel’s mobile group expects to ship more than 400 million chips to customers this year, Eul said. What’s interesting is that Intel can also leverage all this mobile technology to “reinvent” PCs as well. A thin and light convertible with a Bay Trail quad-core, integrated global 4G LTE and Windows 8.1 at prices starting around $400 should be pretty compelling.

Topics: Processors, Laptops, Smartphones, Tablets

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

Talkback

20 comments
Log in or register to join the discussion
  • I Don't Know Why They Bother

    Why this irrational attachment to x86? Why not leverage their legendary design and fab prowess to implement a more efficient architecture less burdened by decades-long accumulations of legacy baggage that only matters to desktop Windows users?

    In other words, why not make ARM or MIPS chips?
    ldo17
    • Why they bother...

      Correct me if I'm wrong, & that's not a rare occurrence but what are the implications of that? As I understand anything written for the x86 platform is relatively easy to port to multiple systems (Linux, Mac or MS).

      While ARM seems to be a great CPU for power saving (phones & possibly lightweight tablets), what are the benefits for systems used for anything other than email & the odd document?

      I'm not saying these Intel CPU's are the be-all and end-all, far from it (most Intel based mobile devices are pretty average from my experience), but what do we have as a real alternative - maybe AMD can come up with something as their APU movement seems like the right way to go for the mobile market (tablets & the like).
      thrakath_z
      • Re: anything written for the x86 platform is relatively easy to port to mul

        If that were true, then the original Windows NT multiplatform concept would not have failed.
        ldo17
        • No. That was due to sloppy software.

          Nothing fixes that better than actually porting software.

          MS just didn't like being shown up - nearly everything would have to be rewritten.
          jessepollard
          • Re: That was due to sloppy software

            Windows NT was sloppy software?
            ldo17
        • There wasnt anything inherently wrong with Windows NT multiplatform

          For those that don't know, Windows NT was originally intended to be available for more than Just x86. Microsoft had different hardware abstraction layers (HAL) for different platforms. I remember PowerPC and DEC Alpha (might have been others but this was a long time ago).

          I actually helped a client set up a pair of graphics workstations that were running Windows NT on 500 Mhz DEC Alphas. They were real screamers for their day (I'm going from memory but I think the top Intel chip at the time was a Pentium 133 Mhz). They worked great. But not many bought Windows for anything other than Intel. Other HAL's were still there for Windows 2000 but gone in Windows XP/Server 2003.

          But I admit, the concept did fail. From a technical standpoint I didn't see anything wrong with it. It did work fine in my limited experience with it and performance was orders of magnitude better than the Intel CPU's of the day. My theory is that People buying Alphas generally wanted Unix and Windows just was not on their radar weather it worked or not. Also. x86 caught up in performance.

          Despite the theoretical advantages of RISC processors, Intel's ever improving manufacturing processes were always able to just pack more transistors into smaller and smaller packages that could be sold at low enough prices that they have been able to stay ahead.

          But...their whole approach of just packing in more and more transistors and making them smaller and smaller seems like kind of a brute force approach but one that works in a desktop PC...but also the reason its been hard for them to go mobile.

          Still, they have the engineering know-how, the manufacturing expertise, and most importantly enough money where they can afford to take a loss for a few years. They will eventually be successful in the mobile market...not sure they are there yet however.
          cornpie
          • Re: But I admit, the concept did fail

            Someone in another thread tried to put the blame on the other architectures targeted by NT (PowerPC, Alpha etc) being too low-volume. But that doesn't explain why the MIPS port didn't succeed either: MIPS is the other architecture (besides ARM) currently outshipping x86.
            ldo17
    • Lets list the reasons!

      1. More performance with x86.
      2. 64bit here and now. ECC memory
      3. USB 3.0
      4. Out of order execution

      And all that legacy baggage is such a small part of it.
      DevGuy_z
      • Re: Lets list the reasons!

        1. That greater performance comes at a greater power cost. Have you noticed this particular article is about Intel trying very hard to lower that power cost, to be more competitive with ARM?
        2. 64-bit would be good. I believe ARM A15 provides that.
        3. USB 3 would also be good. Interestingly, Intel's own chipsets have been slow to provide that.
        4. ARM A15 has OOO.
        ldo17
        • Seems to me...

          ...we are approaching convergence as time goes by. And Intel has deep enough pockets to stay the course even if it looses money for a while.
          cornpie
          • Re: Intel has deep enough pockets

            After it has spent its way into this market, how is it going to refill those pockets?

            ARM customers are not accustomed to paying x86 prices.
            ldo17
    • One major plus for x86 ...

      ... is its code density. X86 instructions are encoded in different bit lengths depending on their complexity. This results in simpler (i.e., the most common) instructions requiring fewer bits and, thus, enabling code to be cached very effectively, greatly boosting performance.

      This is one reason ARM developed their Thumb and Thumb2 ISA's allowing simpler instructions to be packed into fewer bits, albeit not as efficiently as X86.

      X86 processors also have lots of internal sophistication and features (e.g. deep pipelining, out of order execution, etc.) that results in compiler generated code executing very, VERY quickly.

      ARM processors, on the other hand, tend to be simpler and thus execute comparatively slowly, and require more memory accesses, which further slows execution performance.

      ARM is now increasing sophistication of its chips to improve perf, but that means more transistors and that means more power consumption.

      Intel is improving its ability to power-down as much of its chips' internals as possible until the dormant resources are required.

      Ultimately, both are racing towards one another whereupon it'll be a much more closely matched fight in terms of power vs. Perf.
      bitcrazed
      • Re: which further slows execution performance.

        The interesting thing is that ARM has struck a better balance between performance and power consumption than Intel has been able to manage. (Hence this article.)

        Intel has to match ARM on 3 factors: performance, power consumption and cost. So far it can just about manage two. All that x86 baggage makes it impossible to score all three.
        ldo17
      • But isn't requiring "more memory accesses" a significant bottleneck?

        ...on mobile devices that are lower on RAM and tend to rely on slow flash memory?
        cornpie
    • By the way

      It's not just the hardware capabilities of x86 that make it superior. Having access to more than 4 million Windows desktop apps, and being able to turn your tablet into a PC whenever you want is a big deciding factor. I think everyone (especially enterprises) takes a Windows 8 running tablet over an Android one any day, if the Windows tablet can compete in the battery department (which apparently Intel can pull off).
      Ehsan Irani
      • Re: Having access to more than 4 million Windows desktop apps,

        None of which matter a jot in the mobile space, as Microsoft has discovered.
        ldo17
        • Its a mixed bag

          I recently got a Surface pro at work while my personal tablet is a Rooted, Rom'ed and overclocked Transformer Prime. For personal use which is mostly web browsing, entertainment, and email, I definitely prefer my Android device.

          On the other hand, for work, I actually do prefer the surface pro for a variety of reasons. What the surface pro could really use though is a good docking station (and no I don't mean just the keyboard). I've experimented with plugging a USB hub into it and then connected a USB Keyboard, USB mouse, and USB ethernet adapter and also an external monitor via the display port. The device really is powerful enough that a user with relatively low demands (like office applications etc) might be able to use one as their only computer. Hooking up all the wires to it was too much of a pain other than as an experiment to see that it worked pretty well if I went to the effort. With a real docking station though?? Yes it could work for a lot of people. Not me mind you, but a lot of people.

          As a tablet though, its really still more of a laptop that can do double duty as a tablet when needed - which fills a certain niche but is definitely a different sort of device than my Android device or an iPad.
          cornpie
    • I don't think . . .

      . . . Intel just wants to be an ARM fab. They design chips. I think I'd rather they were competing with ARM than just being assimilated by ARM. I own and ARM-based phone and an ARM-based tablet, so I've got nothing against ARM, but the world is more interesting with competing chip designs.
      FDanconia
      • Re: I don't think . . . . . . Intel just wants to be an ARM fab

        They can be a MIPS fab, then. That also is outselling x86.
        ldo17
  • The x86 is already more power efficient than ARM in inferior process node

    THE ABI research showed that the intel clover trail+, the five years old core built on 32nm process, is already faster and more power efficient than ARM A15, A9 and krait built on 28nm processs.

    http://www.pcworld.com/article/2041056/analyst-intels-smartphone-chip-offers-significant-power-savings-vs-arm.html

    It really need a genius to drive intel alway from x86 now :D
    wsw1982