Test cost reduction for the AMD™ Athlon processor using test partitioning

Abstract

The application of SOC-style test partitioning to a monolithic microprocessor design results in considerable benefits, including simpler and faster ATPG, reduced ECO impact, faster debug, and, most surprisingly, reduced test application time. These results challenge the orthodoxy that flat, top-level ATPG is the best method to produce an optimal pattern set. The granularity of the partitioning was the key factor in achieving the results: a 33-element partition of the AMDtrade Athlon CPU chip resulted in better than a ~80% reduction in test time compared to aflat model of the entire chip. This paper describes the ATPG experiments and quantifies the design overhead required for implementing wrapper cells at partition boundaries.