Over the last two decades LS-DYNA models gained tremendously in terms of size and complexity. This evolution went hand in hand with the buildup in hardware infrastructure onsite customers. The optimal parallelization of high complex models on more and more compute cores is an extreme challenge for the hardware, the solver and the user.
The LS-DYNA out-of-the box performance is already on a great level. However, with respect to a particular application the performance can be increased.
The ultimate goal is – and always has been – to keep the runtimes down to a minimum and, hence maximize the simulation throughput.
The tuning knobs to improve the throughput performance of a system run from hardware settings to model design. Through our long time experience in working with LS-DYNA in different environments onsite our customers, we are able to provide deep knowledge throughout this whole range.
For our customers, in typical car crash simulations, we were able to reduce turnaround times from 10% up to over 30%. In a given example of a 22 hours model runtime, the model could be computed in about 15 hours. With this performance increase, a one day solution turns into an overnight solution. This is a significant improvement to the entire CAE workflow.
In case are you interested in reducing the runtimes of your models we are happy to help you. In a first step we are analyzing your current hardware configuration, your workflow and your model setup. Once we gathered all this information, we will identify potential bottlenecks and propose solutions to eliminate these.
Did we stimulate your interest? Please contact us for further information.