System-Level Calibration for Memories under Process Variability Impact

In microelectronics technology scaling is performed in orderto improve the performance and cost of embedded systems. Scaling to the minimum feature sizes below 100 nm, however, causes a host of problems. One of the most important problems is process variability which is a variation of device parameters after fabrication. Since process variability increases by downscaling the silicon integration technology minimum geometry transistors are heavily affected. Therefore the focus of this book are small on-chip memories in the lowest layer of a memory hierarchy. Process variability severely impacts the performance and the energy consumption of memory organizations. RTL simulations suggest that process variability slows down the memories and makes them consume more energy in most cases. This degrades the parametric- orsystem-yield after fabrication. To improve on the current standard corner-point design techniques, which produce huge overheads in energy and delay, IMEC introduces configurable memories. This book introduces a controller which is able to put such memories in a state were all requirements are met by minimum energy consumption.