A Novel Frequency-Spatial Domain Aware Network for Fast Thermal Prediction in 2.5D ICs

2025-04-19

AI Summary

In 2.5D chiplet integrated circuits, rising power density induces localized thermal hotspots, yet existing CNN/GCN-based methods struggle to capture global high-frequency thermal patterns, limiting prediction accuracy and generalization. Method: This paper proposes FSA-Heat, a frequency–spatial dual-domain aware thermal prediction network. It introduces a novel joint architecture integrating Fourier-domain spectral analysis with graph-structured modeling, comprising a Frequency-to-Spatial Transformer Encoder (FSTE) and a Frequency-domain Cross-scale Interaction Transformer (FCIFormer). A frequency–spatial hybrid loss (FSL) is further designed to suppress thermal gradient noise and spatial misalignment. Results: Experiments demonstrate that FSA-Heat reduces RMSE by over 99% compared to the state-of-the-art GCN+PNA baseline, achieves 4.23× inference speedup, and maintains robust generalization across multiple process corners and layout variations.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: Other Foundations of Computer Vision
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

In the post-Moore era, 2.5D chiplet-based ICs present significant challenges in thermal management due to increased power density and thermal hotspots. Neural network-based thermal prediction models can perform real-time predictions for many unseen new designs. However, existing CNN-based and GCN-based methods cannot effectively capture the global thermal features, especially for high-frequency components, hindering prediction accuracy enhancement. In this paper, we propose a novel frequency-spatial dual domain aware prediction network (FSA-Heat) for fast and high-accuracy thermal prediction in 2.5D ICs. It integrates high-to-low frequency and spatial domain encoder (FSTE) module with frequency domain cross-scale interaction module (FCIFormer) to achieve high-to-low frequency and global-to-local thermal dissipation feature extraction. Additionally, a frequency-spatial hybrid loss (FSL) is designed to effectively attenuate high-frequency thermal gradient noise and spatial misalignments. The experimental results show that the performance enhancements offered by our proposed method are substantial, outperforming the newly-proposed 2.5D method, GCN+PNA, by considerable margins (over 99% RMSE reduction, 4.23X inference time speedup). Moreover, extensive experiments demonstrate that FSA-Heat also exhibits robust generalization capabilities.

Problem

• Address thermal management challenges in 2.5D ICs
• Enhance global thermal feature capture accuracy
• Reduce high-frequency thermal gradient noise

Innovation

• Frequency-spatial dual domain aware network
• High-to-low frequency and spatial encoder
• Frequency-spatial hybrid loss for noise reduction