Managing the Impact of On-chip Temperature on the Lifetime Reliability of Reliably Overclocked Systems

Abstract

Listen

Translate article

Physical and environmental variations require the addition of safety margins to the clock frequency of digital systems, making it overly conservative. Aggressive, but reliable, dynamic clock frequency tuning mechanisms that achieve higher system performance, by adapting the clock rates beyond worst-case limits, have been proposed earlier. Even though reliable overclocking guarantees functional correctness, it leads to higher power consumption and overheating. As a consequence, reliable overclocking without considering on-chip temperatures will bring down the lifetime reliability of the chip. In this paper, we analyze how reliable overclocking impacts the on-chip temperature of microprocessors, and evaluate the effects of overheating, due to reliable dynamic overclocking mechanisms, on the lifetime reliability of such systems. We, then, evaluate the effects of performing thermal throttling, a technique that clamps the on-chip temperature below a predefined value, on system performance and reliability. Our study shows that a reliably overclocked system, along with dynamic thermal throttling, achieves around 25% performance improvement, while operating within 355K.