A key issue in numerical modelling is the computational speed. Improvements in calculation speed make it possible to execute larger models or to undertake more simulations within a given time frame. This is highly relevant both in research environments and in professional engineering departments.
The use of graphical processing units (GPUs) is a very efficient and cost-effective way to speed up simulations and as of release 2014 we have made it possible to use GPUs with our MIKE 21 FM HD software and with the Flexible Mesh solver with MIKE FLOOD software. We also support the use of multiple GPUs on Windows PCs.
Depending on the model setup and the type of GPU, the speed-up can be a factor 100 or more compared to running the same simulation on a single processor CPU*. Going from one to four GPU cards yields another factor of up to three – leading to a total speed-up in the range of 300. Additionally, the largest speed-up factors kick in when you really need them – for large model meshes of one million elements or more.
For more detailed information about performance, please view the benchmark reports.
- MIKE 21 GPU uses the latest graphical processing units - hardware that is otherwise typically used for speeding up computer games
- Significant increase in speed at a very competitive price
- Available at no extra cost to all owners of corporate licences of MIKE 21 or MIKE FLOOD with valid software maintenance agreement (SMA)
Simulation speed (speed-up factor) for a high resolution Mediterranean model with and without 4 GPU cards (MIKE 21 FM running single precision, first order scheme).
* The speed-up of more than 100 has been achieved using a two core HP workstation with Intel® Core™ i3-2120 Processor (2 core, 3.30 GHZ), 8 GB of RAM and a GeForce GTX TITAN card. The operating system was Windows 7. Results for other GPU cards are available in the benchmark report.