I'm baking a liquid simulation with Blender 2.82a. I have an Nvidia GTX 1060 running in an HP Z600 workstation with twin Xeon CPU's. The baking is only using my CPU's not the GPU, even though I have CUDA set to use both. Is this normal for the Mantaflow Liquid sim or is there another setting I need to tweak?

Many thanks in advance.

  • $\begingroup$ Nope, simple reason: the sim won't fit into the ram of the card hence no gpu support. $\endgroup$
    – brockmann
    Mar 27, 2020 at 14:35
  • $\begingroup$ Ah, ok. That's a shame but thanks $\endgroup$ Mar 27, 2020 at 15:18
  • $\begingroup$ I'm not sure mantaflow even does have CUDA support. They mention it on their installation page but I couldn't find CUDA in their gitlab code though (maybe it's used by their numerical libraries) $\endgroup$
    – xalpha
    May 21, 2020 at 18:40
  • $\begingroup$ "the sim won't fit into the ram of the card hence no gpu support." Any reference to this claim? How much memory would be needed? An RTX 3090 has 24GB. Would that be enough? Even with less RAM, why can't parts of the computations be run on a GPU. I don't understand why you would have to fit the whole simulation at once into the GPU. $\endgroup$
    – Bjorn
    Mar 31, 2021 at 7:21


You must log in to answer this question.

Browse other questions tagged .