Reproduce “MORE-Stress: Model Order Reduction based Efficient Numerical Algorithm for Thermal Stress Simulation of TSV Arrays in 2.5D/3D IC” on OpenResearch

About this paper

Tianxiang Zhu, Qipan Wang, Yibo Lin, Runsheng Wang, Ru Huang

Researchers at Peking University developed MORE-Stress, a model order reduction-based algorithm for efficient and accurate thermal stress simulation of Through-Silicon Via (TSV) arrays in 2.5D/3D ICs. The method achieved computational speedups of up to 504x and memory reductions up to 115x compared to commercial FEM, while maintaining errors below 1%.

Key insights

• The periodicity of fine structures like TSV arrays can be exploited through model order reduction to significantly reduce the computational complexity of thermal stress simulations for large-scale systems.
• By pre-calculating local basis functions for unit blocks and treating them as abstract elements, a large-scale array problem can be efficiently solved with fewer degrees of freedom while maintaining high fidelity.
• Combining a rigorous numerical method (FEM) with model order reduction allows for a solution that is orders of magnitude faster and less memory-intensive than full FEM, yet more accurate than simplified linear superposition methods.

Summary by alphaXiv.