A Case Study for Using Dynamic Partitioning Based Solution in Volume Diagnosis

摘要

Diagnosis driven yield analysis (DDYA) has been widely adopted for advanced technology node product yield ramp [1]. However gigantic design size and high pattern count demand intense computation resources to diagnose volume failure data, and the diagnosis throughput becomes the bottleneck for the DDYA flow. This paper presents a case study which uses the fully automated dynamic partitioning based diagnosis solution to dramatically improve the throughput. Experimental results based on real silicon manufactured by a 16nm FinFET technology show more than 3X reduction for memory footprint and more than 4X improvement for runtime, which eliminates the throughput bottleneck.