DRAM Performance Sensor.

Human-machine environments require computers with capacity to store large amounts of data, and it’s important that memories do not fail, otherwise it will jeopardize the Human-machine interface. Moreover, as battery operated devices are becoming widely used in everyday Human-Machine environments, it’s also imperative to reduce power consumption in these devices. The present work presents a new DRAM performance sensor, to be used in DRAM memories of Human-Machine environments, especially in battery operated devices. Effects such as process variations (P), power-supply voltage variations (V), temperature variations (T) and aging (A) variations (PVTA – Process, Voltage, Temperature and Aging) are key parameters that affect chips performance and reliability. The new performance sensor for DRAM memories has the purpose to signalize when these PVTA variations, or any other parameter, change performance of the memory above a certain threshold limit, jeopardizing memory operation, signal integrity, and the Human-Machine system where it is used. Sensor’s sensibility to PVTA variations can be changed in run-time, which allows the sensor to be tuned during circuit’s life time. Another important feature is that it can be applied locally, to monitor the online operation of the memory, or globally, by monitoring a dummy memory in pre-defined conditions. These features allow the development of intelligent hardware to be used in Human-Machine systems which allow anticipating system failures and also improve power optimization. Moreover, as far as authors know, this is the first online performance sensor for DRAM memories.