Self-Attention to Operator Learning-based 3D-IC Thermal Simulation
Abstract
Thermal management in 3D ICs is increasingly challenging due to higher power densities. Traditional PDE-solving-based methods, while accurate, are too slow for iterative design. Machine learning approaches like FNO provide faster alternatives but suffer from high-frequency information loss and high-fidelity data dependency. We introduce Self-Attention U-Net Fourier Neural Operator (SAU-FNO), a novel framework combining self-attention and U-Net with FNO to capture long-range dependencies and model local high-frequency features effectively. Transfer learning is employed to fine-tune low-fidelity data, minimizing the need for extensive high-fidelity datasets and speeding up training. Experiments demonstrate that SAU-FNO achieves state-of-the-art thermal prediction accuracy and provides an 842x speedup over traditional FEM methods, making it an efficient tool for advanced 3D IC thermal simulations.
Keywords
Cite
@article{arxiv.2510.15968,
title = {Self-Attention to Operator Learning-based 3D-IC Thermal Simulation},
author = {Zhen Huang and Hong Wang and Wenkai Yang and Muxi Tang and Depeng Xie and Ting-Jung Lin and Yu Zhang and Wei W. Xing and Lei He},
journal= {arXiv preprint arXiv:2510.15968},
year = {2025}
}