As silicon technologies move into the nanometer regime, there is growing concern for the reliability of transistor devices. Device scaling may aggravate a number of long standing silicon failure mechanisms, and it may introduce a number of new non-trivial failure modes. Unless these reliability concerns are addressed, component yield and lifetime may soon be compromised.
Researchers at University of Michigan have developed a low-cost reliable system design approach which provides fine-grained detection, diagnosis, recovery, and repair of silicon defects that occur while the system is in operation in the field. While traditional approaches require at least 100% overhead due to duplication of critical resources, this online testing-based approach provides the same level of protection with an overhead of 5.8%. In addition, when a defect occurs, the subsequent degraded mode of operation was found to have only moderate performance impacts, (from 4% to 18% slowdown). Moreover, the coverage of this approach is good, with 89% of the total area of the prototype design effectively protected against silicon defects.
Applications and Advantages
- Online testing for microprocessors defect
- Protection for microprocessor pipelines and-nl-on-chip cache memories
- Ultra low-cost
- Very efficient with high coverage-nl-and low performance impacts