Intel "Ice Lake" (10 nm)

Discussion in 'PC Industry' started by iMacmatician, Jul 13, 2015.

Tags:
  1. Gubbi

    Veteran

    Joined:
    Feb 8, 2002
    Messages:
    3,522
    Likes Received:
    853
    They have been heavily limited by bandwidth. The push for higher speed DDR4 and LPDDR4 is what enables these new integrated GPUs.

    Cheers
     
  2. ToTTenTranz

    Legend Veteran Subscriber

    Joined:
    Jul 7, 2008
    Messages:
    9,856
    Likes Received:
    4,466
    Nah, Raven Ridge with the same 128bit DDR4 2400MT/s and similar TDP gets twice the performance.


    [​IMG]



    Not to mention they developed Crystalwell to circumvent possible bandwidth limitations.


    Intel's strategy in keeping the same tiny GT2 GPU over and over throughout the years was simply to keep the chips small to make them cheaper to produce.
    Because they could, obviously.
     
  3. Gubbi

    Veteran

    Joined:
    Feb 8, 2002
    Messages:
    3,522
    Likes Received:
    853
    Obviously.

    Intel were binning SKUs across different segments. You have laptops, which often ships with LPDDR 1833, where a big GPU would be bandwidth limited. You have the business segment where a GT2 "is enough" and you have the gamer crowd where the GPU is just dark silicon anyway. A bigger integrated GPU for the mobile market would add cost across the other segments without any benefit.

    I think you are right that AMD has forced Intel to up the ante in the GPU department. But I also think Intel's shift towards 10nm has helped. The production problems and the process characteristics means Intel is targeting mobile first with 10nm. Since they no longer bin similar dies across different segments, adding GPU resources doesn't incur a penalty to those segments.

    Cheers
     
  4. ToTTenTranz

    Legend Veteran Subscriber

    Joined:
    Jul 7, 2008
    Messages:
    9,856
    Likes Received:
    4,466
    LPDDR4 at 3200MT/s has existed in the market since 2015 with the Snapdragon 810.
    Intel has just shown they didn't need a drastic architecture or node change to support LPDDR4, since Comet Lake U supports LPDDR4X.

    Higher GPU performance would have been welcome for mobile (most of the consumer market AFAIK) and NUC-sized solutions.
     
  5. Dayman1225

    Newcomer

    Joined:
    Sep 9, 2017
    Messages:
    56
    Likes Received:
    77
    Thats because they back ported their 10nm LPDDR4x IP to 14nm.
     
  6. ToTTenTranz

    Legend Veteran Subscriber

    Joined:
    Jul 7, 2008
    Messages:
    9,856
    Likes Received:
    4,466
    Whatever the IP block is in there, Intel could have developed a LPDDR4 controller to launch in a late 2015 product. They're using one right now together on a 14nm SoC with SkyLake cores, which is what they already had back then.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...