- DRAM read disturbance (RowHammer and RowPress) causes unintended bitflips in unaccessed cells, posing critical safety, security, and reliability concerns.
- Prior experimental work has characterized bitflips and proposed mitigations, but device-level models do not fully explain all empirical observations.
- The study aims to bridge the gap between experimental characterization and device-level modeling by identifying inconsistencies in three fundamental metrics: bitflip directions, bitflip counts, and the minimum activations to trigger the first bitflip (ACmin).
- Comprehensive TCAD simulations are presented that match real-chip experimental phenomena for RowHammer and RowPress.
- Results provide updated device-level error mechanisms and identify key modeling parameters that influence simulation accuracy.
- Findings have implications for improving experimental characterization methodologies and designing effective DRAM read disturbance mitigation techniques.