MART (MBIST Algorithm Recommendation Tool), developed by iSTART-TEK, is an AI-driven test algorithm recommendation platform designed to optimize SRAM testing strategy selection for advanced semiconductor products. Traditionally, selecting MBIST algorithms heavily relies on engineers’ experience, manual rule matching, and repeated evaluation of trade-offs among defect coverage, test time, power consumption, and silicon area. MART transforms this workflow into an intelligent AI-assisted decision-making process.
The system integrates multiple design and test parameters, including memory type, memory size, port architecture, target DPPM, failure information, fail bitmap patterns, repair results, and test time budget. Based on these inputs, MART automatically recommends the most suitable combination of March algorithms, UDA (User-Defined Algorithm) elements, and diagnosis flows to build a comprehensive memory test strategy.
A key innovation of MART lies in its multi-objective optimization engine. Instead of using conventional rule-based filtering, MART applies AI-driven weighted analysis to balance product quality, testing efficiency, and implementation cost simultaneously. This enables engineering teams to identify optimal MBIST strategies more quickly and accurately, significantly reducing decision overhead and shortening development cycles.
As AI processors, automotive chips, and high-performance computing devices continue to increase in memory capacity and architectural complexity, achieving ultra-low DPPM and high reliability has become critical. MART addresses this challenge by shifting MBIST planning from experience-driven selection to data- and AI-driven optimization. This innovation marks a new generation of intelligent memory testing, enabling semiconductor designers to improve yield, reliability, and time-to-market with greater efficiency.