NVIDIA announced that ACUSIM Software has integrated support for NVIDIA Tesla 20-series GPUs into the company’s AcuSolve 1.8 release. Performance tests of the general-purpose finite-element-based CFD flow solver have demonstrated up to a 2x boost in performance with the Tesla C2050 GPU processor, compared with the latest quad-core CPU running the same simulation.
AcuSolve is used in a broad range of mechanical design applications and deployed by research organizations and Fortune 500 companies including Bechtel, Chevron, John Deere, Procter & Gamble, Sanyo, Visteon and Whirlpool. They use CFD simulations to replace costly physical tests during product development, which leads to shorter design times and improved product quality.
With the introduction of this version, industries ranging from automotive, aerospace and defense to consumer goods, bio-medical devices and energy production can reduce CFD simulation times, enabling more and increasingly complex simulations to be carried out by tapping into the parallel CUDA computing architecture.
The product has implemented a hybrid parallel scheme that combines shared and distributed memory parallel processing in a single CFD simulation. The shared memory operations are accelerated on Tesla GPUs using OpenMP standards, while the CPUs manage the distributed memory operations using message passing library standards.
AcuSolve has also demonstrated efficient multi-GPU scalability enabling it to be deployed in systems containing multiple GPUs such as the Tesla S-series and M-series products for datacenter-class HPC installations.
NVIDIA
