PyPOD-GP: Using PyTorch for Accelerated Chip-Level Thermal Simulation of the GPU

TL;DR

PyPOD-GP introduces a GPU-optimized PyTorch library for chip-level GPU thermal simulation, achieving significant speedups over traditional MPI-based methods while maintaining high accuracy.

Contribution

The paper presents PyPOD-GP, a novel GPU-accelerated library using PyTorch for efficient chip-level thermal simulation with high accuracy.

Findings

Over 23.4x faster training speed

Over 10x faster inference speed

Only 1.2% error compared to device layer

Abstract

The rising demand for high-performance computing (HPC) has made full-chip dynamic thermal simulation in many-core GPUs critical for optimizing performance and extending device lifespans. Proper orthogonal decomposition (POD) with Galerkin projection (GP) has shown to offer high accuracy and massive runtime improvements over direct numerical simulation (DNS). However, previous implementations of POD-GP use MPI-based libraries like PETSc and FEniCS and face significant runtime bottlenecks. We propose a $\textbf{Py}$Torch-based $\textbf{POD-GP}$ library (PyPOD-GP), a GPU-optimized library for chip-level thermal simulation. PyPOD-GP achieves over $23.4\times$ speedup in training and over $10\times$ speedup in inference on a GPU with over 13,000 cores, with just $1.2\%$ error over the device layer.