Intermittent faults and effects on reliability of integrated circuits

A significant amount of research has been aimed at analyzing the effects of high energy particles on semiconductor devices. However, less attention has been given to the intermittent faults. Field collected data and failure analysis results presented in this paper clearly show intermittent faults are a major source of errors in modern integrated circuits. The root cause for these faults ranges from manufacturing residuals to oxide breakdown. Burstiness and high error rates are specific manifestations of the intermittent faults. They may be activated and deactivated by voltage, frequency, and operating temperature variations. The aggressive scaling of semiconductor devices and the higher circuit complexity are expected to increase the likelihood of occurrence of the intermittent faults, despite the extensive use of fault avoidance techniques. Herein we discuss the effectiveness of several fault tolerant approaches, taking into consideration the specifics of the errors generated by intermittent faults. Several solutions, previously proposed for handling particle induced soft errors, are exclusively based on software and too slow for handling large bursts of errors. As a result, hardware implemented fault tolerant techniques, such as error detecting and correcting codes, self checking, and hardware implemented instruction retry, are necessary for mitigating the impact of the intermittent faults, both in the case of microprocessors, and other complex integrated circuits.