• DocumentCode
    230398
  • Title

    Novel Critical Path Aware transistor optimization for mobile SoC device-circuit co-design

  • Author

    Mojumder, Niladri N. ; Song, Seung Chul ; Wang, Jiacheng ; Ken Lin ; Rim, Ken ; Xu, Jie ; Yeap, Geoffrey

  • Author_Institution
    Adv. Silicon Technol., Qualcomm Technol. Inc., San Diego, CA, USA
  • fYear
    2014
  • fDate
    9-12 June 2014
  • Firstpage
    1
  • Lastpage
    2
  • Abstract
    We present, for the first time, a holistic system-circuit-transistor co-optimization method, named “Critical Path Aware (CPA) transistor optimization”, through which we demonstrate power reduction of more than 20% in a state-of-the-art SoC design. In this method, we simplify and optimize all paths (critical and non-critical) to guide device design point for maximum power-performance benefit. We introduce novel `Binning and Mapping of statistical Path delay (BMP)´ method as a key enabler of this optimization platform, which deduces complex block level circuit data paths to a set of manageable ring oscillators, which are then used to link product level power-performance metric to transistor level optimization, taking intrinsic transistor performance, multiple Vt and multiple Lg into account simultaneously. This holistic optimization method has the potential as an important tool to extend Moore´s law beyond 10nm node by maximizing performance and minimizing process complexity.
  • Keywords
    MOSFET; circuit optimisation; oscillators; system-on-chip; FinFET; critical path aware transistor optimization; holistic optimization; mobile SoC device-circuit co-design; power-performance benefit; ring oscillators; size 10 nm; statistical path delay binning; statistical path delay mapping; system-on-chip; Delays; Digital signal processing; FinFETs; Optimization methods; System-on-chip;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    VLSI Technology (VLSI-Technology): Digest of Technical Papers, 2014 Symposium on
  • Conference_Location
    Honolulu, HI
  • ISSN
    0743-1562
  • Print_ISBN
    978-1-4799-3331-0
  • Type

    conf

  • DOI
    10.1109/VLSIT.2014.6894379
  • Filename
    6894379