Heat diffusion on multiple GPUs

May 05, 2015

The implementation of the heat diffusion equation on multiple GPUs is explained in the talk (in german).

The heat simulation is implemented in the frameworks CUDA, OpenCL and C++ AMP.

There are two benchmarks

  • The pure performance of the GPUs without disturbance by the OpenGL engine, i. e. a console application. The results are given in the talk.
  • The performance as an OpenGL app. See the following video.