Google Tensorflow ASIC and its impact

Discussion in 'STH Main Site Posts' started by Patrick Kennedy, May 20, 2016.

  1. gigatexal

    gigatexal I'm here to learn

    Joined:
    Nov 25, 2012
    Messages:
    2,677
    Likes Received:
    496
    this is why i will buy google's version of the echo - they're just too good at queries - and they probably already know me better than i do - and i've always wanted a device to be a butler/assistant/jarvis
     
    #2
  2. nrtc

    nrtc New Member

    Joined:
    Dec 3, 2015
    Messages:
    15
    Likes Received:
    2
    I think the TPU isn't as revolutionary as it may seem, since a large chunk of the performance increase is due to computing with 8 bit integers rather than 32 or 64 bit floating points. That alone will give you at minimum a 4x speedup. Nvidia's Pascal also moves in this direction with support for FP16 (reportedly 20Tflop/s at that FP16 vs 10Tflop/s at SP). The rest of the alleged performance increase (around 2x) will be due to the optimized ASIC implementation and integer arithmetic instead of floating point.

    The interesting question for me is why other large players have not jumped on the opportunity of easy 2x or 4x speedups by getting rid of SP/DP, since it's commonly known in the Deep Learning community that there's often no need for more than 8 or 16 bits of precision. Only Nvidia is moving slowly in that direction, basically sacrificing DP performance in many of their GPU's and the upcoming providing FP16 support. Would it be that Google with all their applications is in the perfect position to establish the exact requirements for deep learning at the moment?
     
    #3
  3. Alfa147x

    Alfa147x Member

    Joined:
    Feb 7, 2014
    Messages:
    101
    Likes Received:
    10
    It's the form factor here that makes this incredible and very interesting.

    We know cloud providers and other similar compute companies have custom hardware developed and manufactured to make use of every inch of a physical data center. The push towards modular units that can quickly swapped out when they die (similar to HDDs) seems like the move for many of these companies. This just is an evolution in blade server technologies that's bridging the gap between architectures.

    I can't wait till we can play with these units via cloud providers. I wouldn't mind paying $/h to run some experiments!
     
    #4
    gigatexal and Patrick like this.
Similar Threads: Google Tensorflow
Forum Title Date
STH Main Site Posts New MLPerf v0.6 Results Dominated by NVIDIA and Google Jul 10, 2019
STH Main Site Posts NVIDIA Tesla T4 Inferencing Power Grows on Google GCP Jan 16, 2019
STH Main Site Posts Google Enables Low-Cost Fast TPU v2 Pod Training in GCP Dec 16, 2018
STH Main Site Posts Google Cloud Filestore is a Cloud NAS with NFSv3 Jul 5, 2018
STH Main Site Posts Google Chromebook Pixel Running Ubuntu 18.04 Bionic Beaver Mar 12, 2018

Share This Page