Research On Stress-aware Runtime Reliability Management And Performance Boost Technology For 3D IC

With the Moore's law coming to an end,3D integrated circuit(3D IC)becomes a new possible breakthrough to continue increase transistors' density.But the application of 3D IC is greatly limited because the on-chip thermal and reliability problems are more severe than traditional integrated circuit.Reliability management method can take active management strategy to guarantee 3D IC running in an appropriate performance level,the method will be based on 3D IC's runtime reliability information,such as temperature.So,Accurate and fast reliability management is important for 3D IC.Other than runtime thermal reliability,3D IC also needs to maintain its structure reliability.Different with traditional IC,3D IC introduces through silicon via(TSV)to connect different silicon layers and TSV will bring tremendous stress challenge to 3D IC.However,due to the lack of stress information and difficulties in implementing management method,existing 3D IC reliability management methods suffer from low management accuracy and high system performance degradation.In this work,we propose a new stress and thermal aware reliability management method for 3D IC called STREAM.Unlike traditional methods which do not perform explicit stress analysis due to the large computing cost,they suffer from inaccurate management strategy due to inaccurate reliability information.While STREAM employs an artificial neural network(ANN)based stress model to estimate stress accurately at runtime.In order to further improve the reliability management accuracy and improve the system performance,a lifetime estimator with lifetime banking technology and a specially-designed lifetime model predictive control(LMPC)are integrated into the reliability management framework.Lifetime banking technology regards 3D IC's reliability information as a resource bank which can be withdrew and deposited,this technology enables STREAM to break the performance ceiling which restricts traditional management methods and dynamically improve the performance of 3D IC.Meantime,lifetime model predictive control method can solve the low control efficiency problem of existing management methods,it is based on advanced model predictive control(MPC)method which uses 3D IC's thermal model information.Our numerical results show that: The ANN stress model can estimate accurate stress information with negligible computation cost;lifetime banking technology can significantly improve the performance of 3D IC;lifetime model predictive control outperforms traditional control method in reliability and efficiency.In all,STREAM performs the stress and thermal aware full-chip reliability management with both high accuracy and speed.It is able to boost the performance of 3D ICs and outperforms the state-of-the-art 3D IC reliability management method.