May 15, 2009 -- Ansys, Inc., a provider of simulation software and technologies designed to optimize product development processes, has announced "significant high-performance computing (HPC) milestones" achieved with its Ansys 12.0 platform. The company says new release delivers impressive performance gains that enable product development teams to increase the value of simulation by considering large, high-fidelity models in shorter turnaround times.
Key HPC achievements in the recent release include optimized parallel computing performance on multi-core processors, expanded support for large simulations, scaling breakthroughs, and support for parallel file systems. These product enhancements deliver best-in-class HPC capability for multiphysics simulations, engineered to scale from multi-core desktop workstations to departmental clusters and large enterprise supercomputers.
"Engineers today are including more geometric details in their analyses and looking for a more-realistic treatment of physical phenomena. Both of these factors drive the need for outstanding performance on the latest computing platforms," comments Jim Cashman, president and CEO of Ansys, Inc. "Ansys 12.0 is our most HPC-capable release to date, and it delivers tremendous value to customers who need faster turnaround and the ability to consider high-fidelity multiphysics simulations. HPC is an important part of simulation driven product development, since it delivers efficiency and productivity improvements to engineering organizations seeking to gain a competitive advantage in today's business climate."
The Ansys 12.0 technology incorporates optimization for the latest multi-core processors and benefits greatly from recent improvements in processor architecture, resulting in highly efficient use of parallel processing to reduce the turnaround time for simulation. Improved algorithms for model partitioning, combined with optimized communications and load balancing between processors, have yielded additional parallel scaling breakthroughs. For structural mechanics analyses, dramatically improved scaling is observed on desktop parallel systems, and a major performance milestone of over 1 teraflop has been achieved via cluster computing on 512 cores.
According to the company, with the new release, Teraflop performance is well over 100 times faster than the fastest single-core performance currently observed, reducing run times from days to minutes on the most challenging simulations. For fluid dynamics simulations, nearly ideal linear speedup has been demonstrated out to 1,024 cores, roughly doubling the core count for ideal scaling compared to previous releases. At 2,048 cores, scaling remained at approximately 80 percent of ideal linear performance.
The companion Ansys Fluent 12.0 software introduces parallel input and output (I/O) of files, dramatically reducing turnaround time for large simulations that involve extensive I/O and removing file handling as a bottleneck for scaling on large clusters. In addition, Ansys 12.0 technology includes important enhancements that enable larger simulations than ever before, setting the stage for customers to consider highly detailed physical phenomena and full-assembly models in their fluids or structural simulations.
Recognizing that HPC deployment and management are key concerns, Ansys 12.0 is pretested for interoperability with multiple hardware solutions, including registration under the Intel Cluster Ready standard and support for Microsoft Windows HPC Server 2008, as well as Linux and UNIX-based clusters.
"In today's environment, product development teams are looking to bring their products to market faster with digital prototyping of large, high-fidelity models," offers Kirk Skaugen, vice president and general manager of Intel's server platforms group. "The performance of the Intel Xeon processor 5500 series with the tuned and optimized Ansys products has delivered speedup of more than two times for Ansys Mechanical and Ansys Fluent applications. This kind of speed is key for today's mission-critical engineering simulations."
HPC performance data for the new release and supporting materials are available at http://www.ansys.com/corporate/partners/partners-hpc.asp.
On the Web:
www.ansys.com
