FURI | Fall 2025

Advanced Thermal Analysis of 3D Stacked Heterogeneous Chiplets Using Machine Learning

With the global increase of computing power comes increased energy consumption. One of the largest components of energy usage comes from thermal inefficiency. The goal is to analyze chiplet design accurately and rapidly using machine learning to be able to optimize semiconductor design decisions. Those predictions drive an optimization loop that adjusts block placement and power distribution to lower peak temperature and temperature gradients. This helps to provide over 100× faster analysis, enabling cooler, more energy-efficient chips and fewer design iterations.