GPU-accelerated micromagnetic simulations using cloud computing

TL;DR

This paper introduces a cloud-based, GPU-accelerated approach for micromagnetic simulations, leveraging open-source tools to enhance speed and cost-efficiency compared to traditional CPU methods.

Contribution

It presents a novel Python-based framework for remote GPU-accelerated micromagnetic simulations using cloud services, analyzing their scalability and cost benefits.

Findings

Significant speed improvements over CPU-based simulations

Cost-effective access to GPU resources via cloud computing

Effective scaling of simulations with cloud infrastructure

Abstract

Highly-parallel graphics processing units (GPUs) can improve the speed of micromagnetic simulations significantly as compared to conventional computing using central processing units (CPUs). We present a strategy for performing GPU-accelerated micromagnetic simulations by utilizing cost-effective GPU access offered by cloud computing services with an open-source Python-based program for running the MuMax3 micromagnetics code remotely. We analyze the scaling and cost benefits of using cloud computing for micromagnetics.