Abstract:Thermal issues are becoming increasingly critical due to rising power densities in high-performance chip design. The need for fast and precise full-chip thermal analysis is evident. Although machine learning (ML)-based methods have been widely used in thermal simulation, their training time remains a challenge. In this paper, we proposed a novel physics-informed separation of variables solver (PISOV) to significantly reduce training time for fast full-chip thermal analysis. Inspired by the recently proposed ThermPINN, we employ a least-square regression method to calculate the unknown coefficients of the cosine series. The proposed PISOV method combines physics-informed neural network (PINN) and separation of variables (SOV) methods. Due to the matrix-solving method of PISOV, its speed is much faster than that of ThermPINN. On top of PISOV, we parameterize effective convection coefficients and power values for surrogate model-based uncertainty quantification (UQ) analysis by using neural networks, a task that cannot be accomplished by the SOV method. In the parameterized PISOV, we only need to calculate once to obtain all parameterized results of the hyperdimensional PDEs. Additionally, we study the impact of sampling methods (such as grid, uniform, Sobol, Latin Hypercube Sampling(LHS), Halton, and Hammersly) and hybrid sampling methods on the accuracy of PISOV and parameterized PISOV. Numerical results show that PISOV can achieve a speedup of 245×, and 104× over ThermPINN, and PINN, respectively. Among different sampling methods, the Hammersley sampling method yields the best accuracy.

NIO: A fast and accurate verification method for PVT variations

Silicon-Validated Power Delivery Modeling and Analysis on a 32-Nm DDR I/O Interface.

Multi-Task Evolutionary to PVT Knowledge Transfer for Analog Integrated Circuit Optimization

Knowledge Transfer Framework for PVT Robustness in Analog Integrated Circuits

On the efficacy of input Vector Control to mitigate NBTI effects and leakage power

NBTI and Leakage Reduction Using an Integer Linear Programming Approach

Assessment of Circuit Optimization Techniques Under NBTI

NBTI and Power Reduction Using an Input Vector Control and Supply Voltage Assignment Method

Studies of NBTI in Pmosfets with Thermal and Plasma Nitrided SiON Gate Oxides by OFIT and FPM Methods

Investigation on NBTI-induced dynamic variability in nanoscale CMOS devices: Modeling, experimental evidence, and impact on circuits.

Performance Robustness Analysis of VLSI Circuits with Process Variations Based on Kharitonov's Theorem

Reliability- and Process Variation-Aware Placement for FPGAs

PISOV: Physics-Informed Separation of Variables Solvers for Full-Chip Thermal Analysis

Worst case analysis for evaluating VLSI circuit performance bounds using an optimization method.

Leveraging the Core-Level Complementary Effects of PVT Variations to Reduce Timing Emergencies in Multi-Core Processors

Improving Performance under Process and Voltage Variations in Near-Threshold Computing Using 3D ICs

An Approach to Enhance the Efficiency of RISC-V Verification Using Intelligent Algorithms

Low-cost High-Accuracy Variation Characterization for Nanoscale IC Technologies Via Novel Learning-Based Techniques.

Oscillator Phase Noise Verification Accounting For Process Variations

Temperature-aware NBTI Modeling and the Impact of Input Vector Control on Performance Degradation

A PVT Variation Testing and Compensating Method for Sigle-Chip Digital- Analog Mixed TCXO