Recent

% | $
Quotes you view appear here for quick access.

NVIDIA Corporation Message Board

  • ibexx ibexx Feb 27, 2010 12:10 AM Flag

    EETimes:Six reasons why no one wants an Atom-based SoC

    SortNewest  |  Oldest  |  Most Replied Expand all replies
    • Correction:

      then your processor is an order of magnitude higher 'in power consumption' than your competitors

    • "Did you really say that? "

      I think so...

      "the guy who has hundreds of physics phds on the payroll to assure they remain on the cutting edge of semiconductor manufacturing has so little investment in 'low power capability' "

      I remind you that they just spent $3B on Larrabee... with thousands of engineers on the payroll to assure they remain on the cutting edge.

      "that they need to go to the merchant market for this basic functionality?"

      It's about numbers, if they build a low-power fab, then they have to be able to fill it... But in this case, it more that they didn't predict the market would shift to low-power this soon. In your own words, they're not good at making quick course adjustments.

      They've been very proud they've got ATOM to 5W... and giving themselves bonuses for it too. This will take an x86 design from the ground up, with power-efficiency designed into it, rather than take an existing x86 design and try to make it low-power.

      YOu can have the best team of phds & engineers working on something, but if they're being payed to go in the wrong direction... then your processor is an order of magnitude higher than your competitors.

    • <<INtel doesn't have a low-power fab process, it's why they were looking into a deal with TSMC.>>

      Did you really say that? The largest semiconductor mfgr in the world, the guy who breaks new ground on process capabilities, for example made a big deal about copper layers and all their benefits a few years ago, the guy who has hundreds of physics phds on the payroll to assure they remain on the cutting edge of semiconductor manufacturing has so little investment in 'low power capability' that they need to go to the merchant market for this basic functionality?

      sorry, nfw

      A better assessment is they've probably got a whole prototype fab line dedicated to low power research. And they STILL can't get x86 to where it needs to be. I agree it will take years to get the power down, but they will still be behind ARM if they're trying to shoe horn in x86 into a handheld space.

      And that's the whole point isn't it? They're going to cede this market if they don't pull their head out. (Not that I want or expect them to. :)

    • From two different Anandtech articles:

      "Atom's pipeline is a fairly deep 16 stages, with a 13 stage mispredict penalty.
      Longer pipelines are generally associated with greater power consumption "

      about ARM

      "The shallower pipeline is very important for keeping power consumption low. Mispredicted branches have a much lower performance and power impact on shallow pipelines than they do on deep ones. "

      Anyway, if we're aware of it, then they're definitely aware of it. Among with tons of reality that they have ALOT of ways to rethink their architecture. As evident by ARM vs ATOM that there is an order of magnitude difference in power consumption... there is ALOT of room for INtel to improve.

    • INtel doesn't have a low-power fab process, it's why they were looking into a deal with TSMC.

      Bottom line: INtel is playing catch-up on low-power devices and it will be 2011, or 2012, 2013 before they get their act together on a low-power SoC.

    • <<I would not be surprised if Intel has a real fear about losing control of its intellectual property>>

      Oh please. No one is biting because there isn't anything to be gained here.

      <<if Intel really wants to be in this business>>

      Now this is really an interesting question. They obviously want to be in, but ODMs and consumers have voted that their products aren't compelling enough. So, they take it to step2, they license the greatest core in the world! and everyone should come flocking. Bzzzst. No takers. Huh?

      The truth is x86 can no longer compete in emerging markets. If they want to be handheld, they need a product that is a fraction of a watt in power consumption, and x86 architecturally isn't going to get where it needs to be (in comparison to competitors) with the next two or three process generations.

      Between this and Laughabee/neverbee, Paul O and the boys better regroup in a hurry. If they just keep screwing by trying to shoe horn in x86, this fully integrated SoC market will completely pass them by.

      NVDA on the other hand is nicely positioned as rich graphics a key differentiator and exceptional software competence.

    • "new SoC model might have some inside Intel a little scared"

      Paranoid !! is more accurate

 
NVDA
22.27+0.06(+0.27%)Sep 3 4:00 PMEDT