FINE™/Turbo: Unrivaled HPC Performances Demonstrated on the Jaguar Cluster
|
|
Recent improvements in the load balancing and virtual domain decomposition have resulted in significant increases in the scalability of the FINE™/Turbo solver. These improvements include a more efficient virtual domain decomposition that is completely transparent to the user, and which improves performance at higher process counts. Tested on a 3.5 stage compressor model with 80 million cells, generated using Autogrid5™, linear scaling was documented up to roughly 500 processors. The model included 6 rotor-stator interfaces, with non-matching connections, making it representative of the type of high fidelity, quantitative simulations required in today's turbomachinery industry.
The simulations were performed on the Jaguar cluster, which is part of the National Center for Computational Sciences (NCCS) at Oak Ridge National Laboratory, in Oak Ridge Tennessee. Established in 1992, the center is designed to provide computational scientists with access to powerful computational resources for unclassified research in “grand-challenge” science and engineering applications.
<- Back to previous page