I've recently built a new machine (i7900, 32GB RAM, 2 x Quadro P4000) and the render performance is good although the hardware doesn't break sweat and i think it could be even better if Blender used more resources. Will continue to investigate how to do that.

Anyway, my current issue involves rigid body animations. Once again the hardware is lightly used by Blender - CPU 7% max - and the simulation takes over 2 hours. It involves 1300 spheres falling under gravity.

My question is, can i run the simulation with low poly substitutes (replace UV spheres with cubes or similar?) and then apply the simulation data to the spheres? I think this would really speed things up if there was a way of doing this.

Appreciate any thoughts or ideas.

  • 1
    $\begingroup$ Set the Rigid Body collision type to Sphere, not Convex Hull. The geometry then does not matter. Blender will use just a single core for the calculations. If you want to utilize hardware for simulations better, you will need to do it in another program. $\endgroup$ – Jaroslav Jerryno Novotny Sep 25 '18 at 8:14
  • $\begingroup$ See blender.stackexchange.com/questions/27125/… $\endgroup$ – m.ardito Sep 25 '18 at 8:54
  • $\begingroup$ Thank you Jaroslav for coming back so quickly. I got your comment on alternative software solutions (Houdini) from my related question but my budget is spent and, perhaps more importantly, I've nearly killed myself learning Blender - learning another program will surely finish me!! Interesting point about collision types though - I've always used Mesh because every other setting seems to send everything to hell, particularly Convex Hull. Simulations seem completely uncontrollable with anything other than Mesh. Anyway, i'll try Sphere again and see what that does for me. Thanks again. $\endgroup$ – Stephen Samuels Sep 25 '18 at 9:34
  • $\begingroup$ Thank you m.ardito - that's the first clear explanation of the problem i've seen. Really appreciate your help in pointing it out. $\endgroup$ – Stephen Samuels Sep 25 '18 at 9:34
  • $\begingroup$ Thanks Jaroslav, you solved my problem when i realised collision shapes only work predictably when the origin is centred in the geometry!! Doh! Thank you again. $\endgroup$ – Stephen Samuels Sep 26 '18 at 11:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.