• DocumentCode
    2416163
  • Title

    Process-Tolerant Low-Power Adaptive Pipeline under Scaled-Vdd

  • Author

    Ghosh, Swaroop ; Batra, Pooja ; Kim, Keejong ; Roy, Kaushik

  • Author_Institution
    Purdue Univ., West Lafayette
  • fYear
    2007
  • fDate
    16-19 Sept. 2007
  • Firstpage
    733
  • Lastpage
    736
  • Abstract
    Designing low power pipelines in modern high performance microprocessors is becoming a challenging task due to increasing process parameter fluctuation associated with the scaled devices. Conventional low power design techniques typically make the critical paths of pipeline stages sensitive to parametric variations, degrading the yield. We implement a low-power and robust pipeline design methodology which is suitable for aggressive voltage scaling while maintaining high frequency operations. This is achieved by isolating the critical paths; making them predictable (by design) and ensuring they are activated rarely. At scaled supply (with frequency unchanged), any possible delay errors (under 1-cycle operations) are predicted ahead in time and avoided by adoptively stretching the clock period to 2-cycles. The test-chip implementing the design methodology for a two-stage pipeline in 130 nm process shows 40% power savings with only 13% performance loss (due to adaptive clock stretching operations) and ~9.4% area overhead.
  • Keywords
    logic design; low-power electronics; microprocessor chips; pipeline arithmetic; adaptive clock stretching operations; critical paths; delay errors; high performance microprocessors; low power design techniques; low-power adaptive pipeline; parametric variations; process parameter fluctuation; scaled devices; size 130 nm; two-stage pipeline; voltage scaling; Clocks; Degradation; Delay effects; Design methodology; Fluctuations; Frequency; Microprocessors; Pipelines; Robustness; Voltage;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Custom Integrated Circuits Conference, 2007. CICC '07. IEEE
  • Conference_Location
    San Jose, CA
  • Print_ISBN
    978-1-4244-1623-3
  • Electronic_ISBN
    978-1-4244-1623-3
  • Type

    conf

  • DOI
    10.1109/CICC.2007.4405835
  • Filename
    4405835