Thermal Modeling on Circuit Layout Thermal Modeling Data
1,001,000Speedup vs FEMMTA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MTAMethodology=Numerical / Compact Hybrid, Grid / Model Size=Multi-layer 3D2025.12 | 1,001,000 | — | — | |
| 3D-ICEMethodology=Compact 3D Solver, Grid / Model Size=Full 3D Stack2025.12 | 975 | 50 | — | |
| ThermGANMethodology=GAN (Generative Model), Grid / Model Size=128×1282025.12 | 300 | 7.5 | — | |
| LSTM+DCTMethodology=ML Regression, Grid / Model Size=128×1282025.12 | 200 | 19 | 2.26 | |
| ThermAlMethodology=NN + Physics Regularizer, Grid / Model Size=256×2562025.12 | 200 | 10 | — | |
| HotSpotMethodology=Thermal RC Network, Grid / Model Size=∼160×1602025.12 | 150 | 25.8 | — | |
| Power BlurringMethodology=Green’s Function Convolution, Grid / Model Size=192×1922025.12 | 40 | 0.1 | 2 | |
| PACTMethodology=Compact Thermal Solver, Grid / Model Size=Up to 512×5122025.12 | 20 | 200 | 3.3 | |
| COMSOLMethodology=FEM Solver, Grid / Model Size=Any mesh2025.12 | 1 | — | — | |
| DeepOHeatMethodology=Neural Operator Learning, Grid / Model Size=3D domain2025.12 | — | 5 | — |