Oct 2026· International Symposium for Testing and Failure Analysis· 0 citations
Abstract
NAND-type 3D Flash memory—built on vertically stacked memory cells to achieve higher storage density and enhanced performance—has become a foundational technology in modern data-storage systems and AI workloads. However, continued scaling to further increase density and improve device characteristics introduces substantial process-integration challenges. As dimensions shrink and architectural complexity rises, defect rates tend to increase, and more critically, many defects manifest transient or intermittent electrical behavior despite being physically observable. Although process improvements remain essential, understanding the failure mechanisms that drive these intermittent or transient behaviors is equally critical. Such insight is vital for defining effective screening strategies and stress methodologies that support time-to-market goals, while meeting stringent DPPM requirements. In this manuscript, we investigate three classes of transient failures originating from defects located in the row circuits, column circuitry, and peripheral logic—key functional blocks of a NAND-type 3D Flash memory chip outside the memory array. Beginning with the observed failure signatures, we analyze the underlying root causes and explain the physical mechanisms responsible for their transient behaviors.
As NAND-type 3D Flash memory continues to scale in performance, speed, and storage density, the underlying architecture has grown significantly more complex. Global reference signals and clock networks must traverse large areas of the chip while driving numerous loads. A defect affecting even a single load can distur...
Xiao-Chen Zhu, L. Ngô, Mi-Mi Dang et al.· International Symposium for...· 0 citations
For 3D Flash memory products, competitiveness depends not only on die size but also on how quickly data can be programmed into and read from the memory cells. Achieving high performance requires minimizing delay throughout the entire data path— from the I/O interface, through buffers and sense amplifiers, to the memo...
Long Phi, Xiao-Chen Zhu, Elliott Rill· International Symposium for...· 0 citations
3D NAND flash has become the foundational non-volatile storage platform of the AI era, underpinning workloads from model training to large-scale inference and cold data archival. With roadmaps now targeting kilolayer stacks and tens of trillions of devices per die, scaling is no longer governed primarily by process int...
Prasanna Venkatesan, T. Song, G. Thareja et al.· 0 citations
Dynamic random access memory (DRAM) memory is organized as an array of individual storage cells, each holding data as electrical charge in a capacitor. Due to inherent charge leakage over time, periodic refresh operations are required to preserve data integrity. As semiconductor manufacturing processes scale down and m...
Khalil Gammoh, Roberto Capoferri, Adrian Drexler et al.· IEEE Transactions on Reliabi...· 0 citations
In Flash semiconductor technology, NAND-type 3D Flash memory continues to stack up more Word-Line (WL) layers e.g. 2XX for 8th and 9th generation 3D NAND. The complexity of interconnect structures and the density of metal layers have increased dramatically. This evolution introduces new failure modes that are difficu...
Shen Chen, Alex Su, Rita Hou· International Symposium for...· 0 citations
Static Random Access Memory (SRAM) remains the dominant on-chip memory technology, occupying over 50% of chip area in modern microprocessors and serving as the backbone of cache hierarchies, register files, and AI accelerators (Evolution of SRAM Architectures, 2026). As CMOS technology scaling approaches fundamental ph...
Shivani Patel, Mrs. Deepali Sahu, Mr. Deepak Sahu· International Journal of Adv...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 8, 2026
Exploring how generative AI could make machine vision more accessible to businesses. The post GenEye in a Box: Making Machine Vision Something You Can Just Ask For appeared first on GPT-Lab.
MIT News · Artificial Intelligence· news.mit.eduOct 8, 2026
Training AI agents with reinforcement learning can be challenging because their tools, context, and decision-making are managed by complex frameworks. Agent Lightning connects existing agents to RL training, making it easier to improve them without rebuilding them. The post Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses appeared first on Microsoft Research.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.