Intel's new Nervana platform...

Discussion in 'Graphics and Semiconductor Industry' started by bdmosky, Nov 19, 2016.

Tags:
  1. bdmosky

    Newcomer

    Joined:
    Jul 31, 2002
    Messages:
    167
    Likes Received:
    21
  2. manux

    Veteran Regular

    Joined:
    Sep 7, 2002
    Messages:
    1,546
    Likes Received:
    394
    Location:
    Earth
    I would look at the intel business model to decide how much they are going after the market. Are they seriously investing to both software and hardware ecosystem or just throwing some hardware to it and hoping it sticks somehow?

    Intel seems to try and half ass a lot of things. They did that with mobile(phone/tablet) and it looks like they are now done with wearables:

    https://techcrunch.com/2016/11/18/intel-layoffs/

    Who knows if intel is really serious about ai or not. It's a lucrative but not very easy market to enter.

    It's very difficult to discuss intel solution as they don't really give tangibles to be discussed. Near future Nvidia products are not well understood either. What is nvlink2 and volta exactly? What is capability of the oak ridge facility being deployed on 2017. Nvidia's xavier is known but not known so it's difficult to really dissect and discuss that chips capabilities and how it compares to competition.

    I believe amd is also going to try to play in this space but who knows how it goes.
     
    #2 manux, Nov 19, 2016
    Last edited: Nov 19, 2016
  3. Pixel

    Regular

    Joined:
    Sep 16, 2013
    Messages:
    981
    Likes Received:
    437
    Bold promises to keep stock prices high and appease investors? Will not supplants Nvidia.
     
  4. manux

    Veteran Regular

    Joined:
    Sep 7, 2002
    Messages:
    1,546
    Likes Received:
    394
    Location:
    Earth
  5. keldor

    Newcomer

    Joined:
    Dec 22, 2011
    Messages:
    74
    Likes Received:
    107
    They say they intend to use it for training. The question, then, is how flexible will it be? If it's too fixed function, it risks being left behind since the most effective ways of training a network is still very much an area of active research, as well as varients in the structure of the network themselves.

    Deep learning networks are currently very "rudimentry" - input -> output processes, but it seems likely that the next step will be some sort of stateful network - that is, networks with memory. We have some very basic types right now, but getting them to full effectiveness may require some tweaking to the training algorithms, and if Nervana isn't flexible enough to do this, it will be useless in that field.
     
    #5 keldor, Nov 20, 2016
    Last edited: Nov 20, 2016
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...