Astrophysical Particle Simulations with Custom GPU Clusters
National Astronomical Observatories of China, Chinese Academy of Sciences, 20A Datun Rd., Chaoyang District, Beijing 100012, China
IEEE 10th International Conference on Computer and Information Technology (CIT), 2010
@inproceedings{spurzem2010astrophysical,
title={Astrophysical Particle Simulations with Custom GPU Clusters},
author={Spurzem, R. and Berczik, P. and Nitadori, K. and Marcus, G. and Kugel, A. and Manner, R. and Berentzen, I. and Klessen, R. and Banerjee, R.},
booktitle={Computer and Information Technology (CIT), 2010 IEEE 10th International Conference on},
pages={1189–1195},
year={2010},
organization={IEEE}
}
We present our new parallel GPU clusters in Beijing and Heidelberg and demonstrate the nearly optimal speedup and performance for parallel direct astrophysical N-body simulations with up to six million bodies. We reach about 1/3 of the peak performance for a real application code. The clusters are used to simulate dense star clusters with many binaries and galactic nuclei with supermassive black holes, in which correlations between distant particles cannot be neglected.
May 19, 2011 by hgpu
