GPU-acceleration of ANSYS Fluent CFD simulations for improved workflow in product development

Bhushan Desam

NVIDIA Corporation


Predicting product performance reliably using CAE simulations in early phases of design is of vital importance for effectively managing new product development process. However, engineers frequently limit the size or detail in simulation models to limit the turnaround times due to computing capacity constraints, thus compromise the accuracy or limit the potential of simulations in product innovation. These challenges can now be overcome with the growing use of general-purpose graphical processing units (GPGPUs) in scientific computing as these offer greater floating-point operations and memory bandwidth rates compared to CPUs.
ANSYS 15.0 now leverages GPUs in several of its flagship products,
including the most widely used CFD product Fluent. Used in conjunction with CPUs, GPUs accelerate CFD simulations by handling parallel tasks on thousands of cores, reducing simulation times by at least half while limiting power consumption. In Fluent 15.0, the algebraic multigrid (AMG) solver is implemented on GPUs to speed up the flow portion of the simulation, which significantly benefit the coupled CFD calculations. As a result, more simulations can be performed in the same amount of time or increase the complexity of the models without increasing the turnaround times, which improve the workflow in simulation-driven product development. This presentation highlights the GPU-accelerated capabilities in ANSYS Fluent 15.0 along with typical CFD applications that can benefit from GPUs.