XPS: FULL: FP: Design and Synthesis of New Energy-efficient Self-healing Computing Electronics with Real-time Configurability

Northwestern University, Evanston IL

Investigators

Abstract

The energy consumption and material cost for modern data-centric computing systems have been rapidly expanding over the past decade. Unfortunately, a significant amount of the energy and material expense is being wasted to create enough tolerance to manufacturing defects, process variation, reliability degradation, etc. As the technology scaling is becoming prohibitively expensive, it is highly desirable to develop a novel computer system with real-time reconfigurability to cope with the variability stemming from both manufacturing and user demand. Unfortunately, conventional CMOS technology has not provided us such a capability because once an integrated circuit is fabricated, it can no longer be modified. However, the very recent development of non-volatile memory devices, specially the emerging memristor device, have opened the door for a new design paradigm with real-time reconfigurability. Leveraging the memristor?s real-time tunability, large on-off resistive ratio and CMOS process compatibility, this project explores a new design and synthesis methodology of large scale integrated circuits with real-time reconfigurability. By intelligently integrating novel memristor device based self-healing circuits with detection and tuning functionalities, a new computer system can perform real-time adjustment without allocating a large amount of design margin and thus achieve significant energy and cost saving. With this in mind, this project will perform (1) device level modeling and design kit development for memristor integrated VLSI and mixed-signal design, (2) circuit level design to create robust self-healing digital and mixed-signal system, and (3) design automation development to enable large scale integration of the proposed reconfigurable design methodology. For broader impact, the project provides a unique training opportunity to the students to obtain a broad knowledge base and out-of-box thinking capability through the cross-layer research activity in this project. The developed design kit and methodology will be available to the community to inspire innovative VLSI and mixed-signal circuit design with memristor devices.