부산대학교 도서관

This paper presents Virtual FA, an innovative approach to enhance the DRAM production quality in Samsung and field quality in customers. By leveraging telemetry data from customer deployment fleet, our method classifies failure types and perform risk assessment without physical analysis, enabling swift and effective issue identification and resolution. The study, involving several experts across DRAM design, process, assembly, and quality team, targeted 15nm DDR4 products. With vFA, we accurately categorized failure types and determined risks across all products and cases, leading to improved production quality in Samsung and field quality in customers with timely feedback.