Newall, Matthew, Holmes, Violeta and Lunn, Paul (2014) GPU Cluster for Accelerating Processing and Visualisation of Scientific and Engineering Data. In: Proceedings of the Science and Information Conference. SAI 2014 . IEEE, London, UK, pp. 140-145. ISBN 978-0-9893-1933-1

The ability to process, visualise, and work with large volumes of data in a way that is fast, meaningful, and accurate is an essential part of many fields of scientific research today. The success of video game industry has resulted in on-going developments in the complexity of Graphical Processing Units (GPU), as well as rapidly falling cost per core. Their characteristics make them excellently suited to any task exhibiting a high level of data parallelism. Recent development of GPU architectures is aimed at HPC systems and applications. In this paper we are presenting our experience in designing and deploying a small dedicated GPU based cluster for processing and visualising data generated by engineering and scientific application. This GPU cluster is helping our researchers to analyse complex data using visualisation, and to accelerate large data processing. We have shown that our GPU cluster solution can achieve five to ten times speed up compared to the CPU system. As a result of our work we can demonstrate that even a small GPU cluster can benefit Higher Education institutions.
Keywords—GPU, CUDA, GPU Cluster, Visualisation

Restricted to Registered users only

Download (953kB)
Add to AnyAdd to TwitterAdd to FacebookAdd to LinkedinAdd to PinterestAdd to Email