Comparing Clusters and Supercomputers for Lattice QCD

TL;DR

This paper compares the cost effectiveness of clusters built from commodity PCs and traditional supercomputers for running Lattice QCD simulations, highlighting key design features and performance considerations.

Contribution

It provides an analysis of the design features and cost efficiency of clusters versus supercomputers in the context of Lattice QCD computations.

Findings

Clusters can be cost-effective alternatives to supercomputers for Lattice QCD.

Key design features influence performance and efficiency.

Empirical comparison based on MILC QCD code runs.

Abstract

Since the development of the Beowulf project to build a parallel computer from commodity PC components, there have been many such clusters built. The MILC QCD code has been run on a variety of clusters and supercomputers. Key design features are identified, and the cost effectiveness of clusters and supercomputers are compared.