9002: GPU-acceleration of CAE simulations for improved workflow in automotive product development
GTC Japan 2014
2014年7月16日
Bhushan Desam, Ph. D.
Sr. Alliances and Marketing Manager-CAE
NVIDIA Corporation
CAE simulations that are commonly used in automotive product design and development processes, such as computational fluid dynamics (CFD), computational structural mechanics (CSM), and computational electromagnetics (CEM), can now be accelerated with NVIDIA GPUs for shorter turnaround times. These applications are traditionally parallelized and run on CPUs, however faced challenges with scalability and power consumption that had limited the number of CPU cores used in a simulation to achieve the necessary turnaround times. GPUs eliminate this problem by utilizing thousands of energy-efficient cores for handling parallel tasks while coordinating with CPUs for serial tasks to complete a simulation. As a result, CAE users can experience speed ups of 2-4X relative to CPU-only solution, which significantly benefit the work flow by allowing them to do the same number of simulations in a less amount of time or increase the complexity of their models without increasing the turnaround times.
This talk will highlight and discuss various GPU-accelerated CAE applications, such as ANSYS Fluent, ANSYS Mechanical, OpenFOAM, Abaqus, MSC Nastran, including several benchmarks to demonstrate the typical speed ups in automotive and manufacturing applications.