IR Drop Estimation and Optimization on DRAM Memory using Machine Learning Algorithms

The IR drop effect on the supply network increases very fast with technology scaling. This can lead to an integrated circuit (IC) speed reduction and timing issues. For IR drop effect reduction, different techniques are used, such as decoupling capacitor insertion, wire sizing, etc.In DRAM memories also, the IR drop effect plays a significant role, as aside from timing issues, in DRAM, IR drop can lead to more issues, up to stored-value loose.In this paper, an approach to IR drop estimation on DRAM memory is used, which is providing data about the possible issue on memory banks using a machine learning algorithm and provides solutions on IR drop reduction. Testing results show, that with using the proposed method, IR drop reduces by 13%, but the available routing track count is decreased by 14%.