AI 中文总结
针对先进3D/3.5D IC封装热管理挑战，提出感知冷却的点Transformer框架COOL，显式编码几何与冷却结构，引入PI-BC损失，在自建基准上NMAE达2.4%，速度较商用FEM求解器提升超15.7倍且优于现有学习方法。
AI 中文摘要
先进3D和3.5D IC封装显著提升了集成密度，但由于层间热耦合和复杂冷却结构，热管理挑战也随之加剧。传统求解器精度高，但迭代设计流程速度太慢；现有基于学习的方法要么无法捕获芯片间热耦合，要么将冷却结构视为静态组件，限制了其在实际封装协同设计场景中的适用性。本研究提出COOL，一个感知冷却的点Transformer框架，它将异构组件（芯片、中介层、TIM、散热片）表示为带标注的3D点云，嵌入几何、材料和功耗属性。COOL显式编码几何边界和冷却结构，并引入物理感知边界条件（PI-BC）损失，以强制材料界面和冷却边界的热一致性。大量实验表明，COOL在我们构建的多封装热设计基准上达到了显著的2.4% NMAE， substantially outperforming existing learning-based approaches while providing over 15.7x speedup compared to commercial FEM solvers.
英文摘要
Advanced 3D and 3.5D IC packaging significantly improves integration density but elevates thermal management challenges due to cross-layer heat coupling and complex cooling structures. Traditional solvers deliver high fidelity but are too slow for iterative design flows, while existing learning-based methods either fail to capture inter-die thermal coupling or treat cooling structures as static components, limiting their applicability in real packaging co-design scenarios. In this work, we introduce COOL, a cooling-aware point transformer framework that represents heterogeneous assemblies (dies, interposers, TIMs, heat spreaders) as annotated 3D point clouds embedding geometric, material and power attributes. COOL explicitly encodes geometric boundaries and cooling structures, and introduces a physics-informed boundary condition (PI-BC) loss to enforce thermal consistency at material interfaces and cooling boundaries. Extensive experiments demonstrate that COOL achieves a remarkable 2.4\% NMAE on our constructed benchmark of multi-package thermal designs, substantially outperforming existing learning-based approaches while providing over 15.7x speedup compared to commercial FEM solvers.
Comments7 pages, accepted at DAC 2026
Journal ref63rd ACM/IEEE Design Automation Conference (DAC '26), July 2026
