Lifetime Electrothermal Stress Management for Multi-core Systems

点击翻译按钮获取中文摘要

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).The objective of this research is to develop a measurement-driven real-time wear-out model for the time-cumulative effects of electrothermal stress on the remaining life of multi-core systems. The approach is to develop highly efficient built-in sensors for accurate temperature, thermal gradient and current measurements and to use real-time measured data to build time-dependent conditional probability density models for electrothermal stress induced circuit failures and time-history dependent system reliability. The intellectual merits include new methods for designing accurate on-chip temperature and current sensors with significant reductions in area and improvements in power efficiency and new methods for real-time thermal gradient surface temperature profiling. Merits also include introducing the concept of time-temperature stress history based conditional probability assessment for developing optimal power/thermal management schemes. Broader impact includes benefits consumers derive through lifetime and reliability improvements in a broad range of elevated temperature limited electronic systems. Significant participation in this project by female and/or minority students will be maintained throughout this initiative. Research results on sensor design, reliability enhancement and statistical lifetime modeling will be integrated into existing and new courses that will be made available to both on-campus and remote access students. Timely dissemination of research results and technology transfer will be through conference and journal publications, WEB page postings, student industrial internships, and close collaboration with researchers at leading semiconductor companies both in the US and abroad.