Toward the Graphics Turing Scale on a Blue Gene Supercomputer

TL;DR

This paper evaluates the raytracing performance on Blue Gene supercomputers, demonstrating significant speedups and analyzing scalability for scientific visualization applications.

Contribution

It provides the first detailed performance analysis and scaling behavior of raytracing on Blue Gene supercomputers, highlighting potential for scientific visualization.

Findings

822x speedup over Pentium IV

Nontrivial scaling at large processor counts

Application potential for high-resolution scientific visualization

Abstract

We investigate raytracing performance that can be achieved on a class of Blue Gene supercomputers. We measure a 822 times speedup over a Pentium IV on a 6144 processor Blue Gene/L. We measure the computational performance as a function of number of processors and problem size to determine the scaling performance of the raytracing calculation on the Blue Gene. We find nontrivial scaling behavior at large number of processors. We discuss applications of this technology to scientific visualization with advanced lighting and high resolution. We utilize three racks of a Blue Gene/L in our calculations which is less than three percent of the the capacity of the worlds largest Blue Gene computer.