10x More Cores for CFD (with Linear Scalability)
Engineering companies have to increase their productivity continuously, which requires more throughput from their simulation tools, making high performance computing (HPC) a core strategic technology. HPC gives you enhanced insight into product performance and improves productivity by allowing more design variants to be considered. ANSYS 17.0 continues the history of improving HPC capabilities of ANSYS CFX and ANSYS Fluent.
Due to further optimization of partitioning, communications and AMG solver, the Fluent solver has now demonstrated scaling to 129,000 cores at 90 percent efficiency, which is unprecedented in its scale and efficiency for parallel processing, as well as in its model complexity. In addition, optimized algorithms for sliding interface intersections and load balancing have resulted in improved scalability of moving deforming mesh (MDM) and sliding mesh cases, such as brake pads, engine crankcase lubrication models and internal combustion engine (ICE) cases. For example, a 1.6 million cell ICE case shows an improvement of up to 55 percent on 384 cores. Last but not least, the case/data IO time has been significantly reduced with the HDF5 format.
Thanks to multiple improvements made in its discretization and linear solver, as well as to other key solution steps like the calculation of general, nonconformal grid intersections, ANSYS CFX simulation time has been significantly reduced for a large variety of cases. For example, a full transient hydroturbine simulation and an internal turbulent flow simulation with CFX show time savings of up to 30 percent on 512 and 4,096 cores, respectively. Furthermore, HPC simulations involving large and complex CFX cases with a very large number of regions or face sets also show a significant reduction in wall clock time for file read/write operations.