• DocumentCode
    646763
  • Title

    Power delivery modeling for full switching voltage regulator on high performance computing system

  • Author

    Wei Xu ; Jiangqi He ; Dong Zhong

  • Author_Institution
    Intel Corp., Chandler, AZ, USA
  • fYear
    2013
  • fDate
    5-9 Aug. 2013
  • Firstpage
    599
  • Lastpage
    603
  • Abstract
    High performance computing (HPC) platform power delivery simulation is typically based on state average mode for simplicity. However, it is not able to capture the dynamic ripple noise and non-linear effect of output voltages, as well as the phase information. The correlation between state average model and lab measurement often lacks accuracy in low to middle frequency range. As a result, platform level power delivery performance assessment has to rely on hands-on measurements to fine-tune the board in order to meet platform specifications, which poses a stringent challenge to product readiness and development cost. To alleviate this problem, single shot design with accurate predictive switching VR modeling to achieve performance and cost balance has been identified as the goal. In this paper, we demonstrated a switching VR modeling methodology with accurate correlation and prediction results using On Semi voltage regulator (VR) model and Intel Romley platform customer reference board (CRB). This methodology applies to other platform simulation with different VRs and motherboards, and can be proliferated to external customers to improve their product design.
  • Keywords
    electronic engineering computing; switching circuits; voltage regulators; Intel Romley platform customer reference board; full switching voltage regulator; high performance computing system; power delivery modeling; semivoltage regulator model; Computational modeling; Correlation; Current measurement; Integrated circuit modeling; Predictive models; SPICE; Switches;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Electromagnetic Compatibility (EMC), 2013 IEEE International Symposium on
  • Conference_Location
    Denver, CO
  • ISSN
    2158-110X
  • Print_ISBN
    978-1-4799-0408-2
  • Type

    conf

  • DOI
    10.1109/ISEMC.2013.6670482
  • Filename
    6670482