Comments Locked

21 Comments

Back to Article

  • therealnickdanger - Monday, February 14, 2011 - link

    I know that Android can run on Moorestown (or any other x86 platform with proper porting), but if Medfield can fit in phones as the article states... are we looking at mainstream availability of phones with the power and ability to dual-boot Android and Windows in the near future? Like Q4 '11 or Q1 '12?
  • sprockkets - Monday, February 14, 2011 - link

    The Moto Atrix already can do this on ARM. Whether or not x86 can do better with the same power constraints remains to be seen.
  • Ryan Smith - Monday, February 14, 2011 - link

    Medfield will not be able to run Windows. Windows requires a number of "legacy" I/O systems including PCI, which smartphones don't include for power saving reasons (polling and idling don't mix).

    For Medfield with Windows support, you'd want the successor to Oaktrail.

    http://www.anandtech.com/show/3752/moorestown-wont...
  • DanNeely - Monday, February 14, 2011 - link

    I wouldn't hold my breath. Windows wouldn't run on the smartphone variant of Moorestown because the PCH didn't have a PCI controller (PCI's bus polling requirement would trash idle power levels) and windows wouldn't boot without it. The noPCI constraint will remain for power reasons, so unless Redmond makes a version of windows that will work without a PCI bus it won't be possible.
  • Shadowmaster625 - Monday, February 14, 2011 - link

    It only requires a few thousand logic gates to emulate a PCI bus to make windows happy. We're talking about adding 47 cents to the cost of a SoC.
  • Cullinaire - Monday, February 14, 2011 - link

    You're joking right? You're saying +47c/unit like that's an insignificant number? We're not talking about XE $1000 CPU's here.
  • Roland00Address - Monday, February 14, 2011 - link

    47cents a unit is a small fraction of the cost considering you are also going to pay $30 to $80 dollars for the license of the Windows OS.
  • PubFiction - Monday, February 14, 2011 - link

    Why couldn't Intel try to give M$ incentive to make an alternate version of windows that could side step this issue? Well other then the fact that Intel keeps messing with linux and making M$ mad.
  • DanNeely - Monday, February 14, 2011 - link

    I wouldn't be surprised if MS does do this for win8 (unless a coding horror makes doing so very difficult). There hasn't been a plausible candidate windows platform for years that didn't support PCI at the chipset level, so MS hasn't had any reason to handle the case where it's not present.
  • nafhan - Monday, February 14, 2011 - link

    Not Windows 7, but it seems like Windows 8 would be a possibility. Presumably, some of the device requirements that keep Windows 7 off of SoC's will be going away on the ARM version of Windows 8, at least.
  • DanNeely - Monday, February 14, 2011 - link

    Unlike other SoC's Moorestown didn't support attaching the ram chip on top of the CPU, and the interface with various other peripherals (eg camera, touch screen) was in a separate chip (Briertown) as well, leaving Intel with 4 chips on the PCB (CPU, PCH, Ram, mixed signals) where almost all of their competition had only one. Medfield will be integrating the PCH, but what about the other 2?
  • IntelUser2000 - Monday, February 14, 2011 - link

    The mixed signal chip, Briertown is a power management unit too, so its not that behind the other SoCs.
  • DanNeely - Monday, February 14, 2011 - link

    How common is it as a separate chip? Annand's Moorestown coverage when it was first released counted it being a separate chip as a liability; and you're the first person I've seen claiming otherwise.
  • IntelUser2000 - Monday, February 14, 2011 - link

    I've heard it was nearly always seperate. Usually the PMIC is one of the smaller chips and is more complicated to integrate so they don't.

    Since I can't link search for these chips:

    iPhone 4: Dialog D1815A Power Management IC
    HTC Surround: Qualcomm PM7540 power management IC
    Samsung Galaxy Tab: Maxim 8998
    Even the iPod Nano has one: Apple 338S0783-B1 10298HLS

    Go to iFixit and read the teardown articles.

    If every article mentioned every chip there would be something that numbers close to a dozen even on a smartphone, but they tend to be insignificant so its ignored.
  • bplewis24 - Monday, February 14, 2011 - link

    Ryan, in Anand's Qualcomm/Krait announcement, he mentions that Qualcomm claims they will be launching a 28nm SoC in a "couple of months." To be fair, he does seem skeptical of that claim.

    Here, you suggest all of the SoC competition will be behind Intel's 32nm process until next year. How are we to reconcile the two conflicting statements?

    Brandon
  • DanNeely - Monday, February 14, 2011 - link

    Anand said it would be sampling in Q211 and available about a year later. sampling means low volume test production only, Intel is already sampling Medfield and intends to be shipping it later this year. Assuming Anand and Ryan's guess about when the chips will become available intel will have the lead in Q4 2011 and Q1 2012.
  • bplewis24 - Monday, February 14, 2011 - link

    Thanks. While I did not have an indication of what "sampling" was, I still clearly missed the part about it being in devices a year later. My fault.

    Brandon
  • mczak - Monday, February 14, 2011 - link

    Both memory controller and gpu moved to cpu for moorestown, mp20 only handles i/o stuff - this was different with z5xx where the chipset was home of the gpu and MC.
    So it's only one shrink for the gpu for medfield.
  • Ryan Smith - Monday, February 14, 2011 - link

    You're right, thanks for catching that.
  • (ppshopping) - Wednesday, March 2, 2011 - link

    welcome
  • lili53 - Friday, March 11, 2011 - link

    welcome

Log in

Don't have an account? Sign up now