Robust Design & HPC
New Licensing Solution for Parametric Studies
Performing simulation studies on many parameters results in dozens or even hundreds of required designs to evaluate. To speed the process, ANSYS offers a robust set of tools for parametric simulation. A key part of this solution is a new licensing product: The ANSYS HPC Parametric Pack amplifies the available licenses for individual applications (pre-processing, meshing, solve, HPC, post-processing), enabling simultaneous execution of multiple design points while consuming just one set of application licenses. Following the structure of the ANSYS HPC Pack license used to deliver scalable parallel processing for single jobs, the ANSYS HPC Parametric Pack license scales the user’s ability to run multiple simultaneous parametric analyses.
Improved Job Scheduling Environment
ANSYS 14.5 improves the process of HPC and robust design exploration by providing more capable and robust job submission/scheduling. The remote solve manager (RSM) can submit multiple design-point jobs, with each job executing on multiple parallel processing cores with third-party job schedulers. The speed of RSM file handling has been improved, so multiple jobs can robustly transfer data across the network to remote computational resources. A new installation and configuration wizard simplifies deployment of RSM in the HPC environment.
ANSYS has engaged with NVIDIA, the leader in GPU computing, to leverage GPU technology to accelerate solver technology. ANSYS 14.5 extends this feature to multiple GPUs per machine, resulting in an increase in scalable performance versus use of a single GPU. In addition, large models that exceeded the memory capacity of a single GPU (and disallowed its use) fit with multi-GPU memory and benefit from acceleration. ANSYS Mechanical customers can interchangeably use CPU cores and multiple GPUs — on a workstation or on a GPU-enabled server — to accelerate most simulation workloads. With ANSYS Fluent, view factors and ray-tracing (radiation) on GP-GPUs are fully supported features, leading to dramatic reduction in computational time for these steps in the simulation. Users can write UDFs to take advantage of GPUs.