A Novel Simulation Methodology Reflecting System Power Scenario using a Markov-Chain-based Stochastic Random Power Model

In this paper, we present a novel power integrity simulation method based on Markov-Chain-based Stochastic Random Power Model (MCSRPM). In this method, we analyze numerous memory traffic cases of an actual system, create a Markov-Chain table which can reflect in system memory scheduling, and create MCSRPM by applying it to DDR5 DRAM state diagram. In practice, we have successfully verified that MCSRPM can predict power noise measured on a DDR5 server system. Also, it is possible to verify many power scenarios predictable in DDR5 systems at once, and confirmed that the analysis time has reduced by more than 60% compared to CPM-based simulation which can only verify few power scenarios.