Application of Imagery Modeling and Deep Learning in Chip Thermal Analysis

Currently, with the reduction of transistor size and the increase in the number of transistors on the chip, thermal simulation in the chip design process has become increasingly important. Thermal simulation can predict the working conditions and temperature of chips in the early stages, providing valuable feedback for chip design to ensure the optimal working temperature. Currently, thermal simulation software typically faces challenges of higher accuracy and longer computation time. The use of deep learning methods helps to quickly obtain high-precision thermal simulation results, which helps to shorten chip design time and accelerate the chip design process. The purpose of the Deep Learning based chip thermal analysis method proposed in this article is not only to verify the feasibility of using deep learning in chip thermal analysis, but also to reduce the time used for thermal simulation in the chip design process and ultimately accelerate the entire chip design process. The experimental dataset is the temperature of multiple layouts of a chip, and using mature image processing deep learning models, high-precision predictions can be achieved.