Fast Prediction of Thermal Throttling Design in M.2 Solid State Drive

Abstract

This paper demonstrates a predictive approach based on R-C network which can fast predict and accurately describe transient thermal throttling behavior of M.2 Solid State Drive (SSD) with 6 major components. The thermal impedance of each component can be described as a function of time and constants of R and C, which are obtained by the non-linear fitting between result from Computational Fluid Dynamics (CFD) simulation and calculation from thermal impedance. Once the constants of R and C are known, the prediction can be easily made. Using such principle, transient thermal behavior of NAND memory component would be evaluated in M.2 SSD drive. Moreover, it is efficient to finish prediction of SSD with 6 heat sources by editing in Excel spreadsheet. The prediction results are proved to be within 1.5°C difference compared with full CFD simulation in a single heating process. Finally, the temperature response of thermal throttling design with 3 power levels in M.2 SSD has been predicted by combining both heating and cooling processes. The predictive temperature response profile of targeted NAND memory agrees well with full CFD model. However, the computation time with this fast prediction approach is much less than that with a full CFD model.