RealFlow Glossary: Iterations in Dyverso

December 23, 2016 Dyverso, Technology, Tips 0 Comments

In the last post you learned essential tips about RealFlow’s substeps and how they work. This time we want to go ahead with iterations. Here I am going to show you their influence on the new “Granular” material type. Before we start, let’s see where the “Iterations” parameter can be found:

Click on “Options…”, open the “Dyverso” tab and go to the “Solver” section. When you go through the available solver types you will see that only some of the entires have an “Iterations” parameter, for example “Granular”:

For the moment we want to keep the default value and take a look at the project. What I have created here is a typical parameter set for sand or snow.

In order to make the particles rest on the objects, their “DY Friction” value has been set to 1.


What we will be doing here is to apply different “Iterations” and evaluate the parameter’s influence on the simulation. The following video has been made with RealFlow’s standard of 20 iterations:

In the next clip, a value of 40 has been used:

You might have noticed that the snow appears more rigid now and that simulation time has increased by roughly 75%. With 80 iterations, we get a completely different behaviour:

With 300 iterations, the material behaves similar to compressed powder or ash, and shows many more small cracks and fissures:


Higher “Iterations” make the material more rigid. The effect is comparable to cloth solvers with an “Iterations” parameter. There, the constraints between the cloth’s vertices become more and more rigid with higher settings. With Dyverso’s parameter it is very similar and you can simulate very different materials just by changing “Iterations”. The range goes from sand to snow, mud or clay, and highly.compressed ash.

If you want to achieve even better results it is also an option to slightly increase substeps to values around 4 or 5, because “Iterations” are applied with every substep. Additionally, higher substeps will care for a better overall result and avoid problems with the collision objects. This, of course, will also mean higher simulation times, but with a powerful GPU it is possible to cut down these times drastically.

The last video shows a quick render of wet sand. The domain values are exactly the same as shown in the image above, but I have been playing with the objects’ “Dyverso – Particles Interaction” parameters instead to get a more vivid behaviour. Solver settings are: MIN substeps = 4, MAX substeps = 4, Iterations = 40

All simulations have been made entirely on the CPU (2x Xeon E2370 v2 with 2x 6 cores @ 2.27 GHz, 24 threads, 600 K particles, average CPU load 87%):