Magasul Posted November 23, 2014 Share Posted November 23, 2014 Hey everyone! I am trying to create a hailstorm effect with raining huge blocks of ice (hail). I set up a simple simulation where there is a grid and each frame a random point is selected and a hail is copied onto it, then with a sop solver it gets merged in. It all works, but after a few pieces, the simulation slows down drastically. They are packed and the DOP uses autofreeze, but still it's slow... Please somebody with experience look at it and tell me why? Thanks! Jegeso_001.hip Quote Link to comment Share on other sites More sharing options...
mawi Posted November 23, 2014 Share Posted November 23, 2014 Why do you need 1750 polys for a sphere? Just put a normal polysphere there and you´ll get speed. Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 23, 2014 Author Share Posted November 23, 2014 Not really working. I have tried to replace it with a simple box and it still slows down... Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 23, 2014 Share Posted November 23, 2014 Please somebody with experience look at it and tell me why? Thanks! Anytime someone asks a 'why' question, Feynman's answer comes to mind No time to do some tests; did you run the Performance monitor and see the slowest bit; how about simulating a lightweight particle system and attaching a the geo to it, or, running the sim and baking it out perhaps Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 I did tests. I replaced the ice geometry with a simple cube, but it is still slow. I just don't get it, when I see simulations running smoothly with hundreds of simple object, why does only a few slow this down? I ran performance monitor and the rigidbodysolver is red, but even when tweaking it to less substeps it is still slow.... I need these hails (ice) to drop to the ground and bounce realistically, but I need maybe hundreds of them. I am using bullet but it's slow as hell.... Anytime someone asks a 'why' question, Feynman's answer comes to mind No time to do some tests; did you run the Performance monitor and see the slowest bit; how about simulating a lightweight particle system and attaching a the geo to it, or, running the sim and baking it out perhaps Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 Trying to do this: http://forums.odforce.net/topic/21300-help-with-emitting-rigid-bodies/#entry127071 ((Can't open the hiplc files in it.) Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 24, 2014 Share Posted November 24, 2014 Just tested your file and it's faster than realtime - i.e. moving to frame 1000 takes 4 sec!, when you connect a BoxSop to xform3. Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 (edited) I have an i7-2700K CPU with 3,8GHz 16 GB DDR3 RAM and an NVIDIA GeForce GTX 560 video card. I don't think that's a bad PC but even after connecting a Box SOP to xform 3 it still slows down. Here is a screenshot of my performance monitor at frame 300: Edited November 24, 2014 by Magasul Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 24, 2014 Share Posted November 24, 2014 oh - time it with a stopwatch, the Performance Monitor has a known bug with Dops. Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 (edited) First hundred frames takes about 20 seconds.... (after clearing cache and reseting simulation) Do you have a hip file lying around where you emit rigid bodies like this and you know it works so I can test it? Thanks! (Or could you tell me the best solution for making a hailstorm with irregularly shaped ice spheres that bounce around and hit eachother after they land, then after a small roll stop as if they landed on grass?) Edited November 24, 2014 by Magasul Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 24, 2014 Share Posted November 24, 2014 Yep - attached - this scene runs >120fps on playback, takes 4 sec after clicking on frame 1000 OsX 10.10.1, Amd 7950, H13.0.582, 2 x X5680 @ 3.33GHz Jegeso_001-2.hip Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 Well that's a bit faster, but still not a hail storm. Any ides on how to achieve a good few hundred instances with good speed? Or I'm dreaming...? Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 24, 2014 Share Posted November 24, 2014 what are you benchmarking against? What is considered good speed for you? Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 I really don't know what to expect to be good as for simulation speed goes. What is the best solution to simulate hundreds of these hails falling on the ground? How would you do it? Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 Maybe the solution is to do it with a particle system and just copy the hail shapes onto the points? But how do i make the shapes roll and behave like their size? Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 24, 2014 Share Posted November 24, 2014 You would normally have hero pieces up close to camera - with the most detail and better simulations, then lesser versions background. It could be a great big mixture of RBD, bullet, particles When you're doing sims and it's running 120fps I'm laughing - get into detailed water sims, or fluids where it'll take many many minutes, to do one frame. Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 24, 2014 Author Share Posted November 24, 2014 Yeah, I know, but it's strange, because with Modo I did this hailstorm effect and it was much faster. :\ Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 24, 2014 Share Posted November 24, 2014 Everything I've read on the Modo forums suggest that the simulations are quite bad in stability- is the quality the same? Quote Link to comment Share on other sites More sharing options...
Magasul Posted November 25, 2014 Author Share Posted November 25, 2014 Well it wasn't today, did it this summer at a studio, but it was satisfying. As far as I can recall, yes, the stability. Like as if they can't find a rest position, everything mostly jittering, but it can be fixed. Quote Link to comment Share on other sites More sharing options...
Guest tar Posted November 25, 2014 Share Posted November 25, 2014 Modo is probably fine for that sort of simulation, what Houdini can do is give you a great workflow to optimise, or mix in other simulations too. You should still be able to get great speed out of Houdini but the main part is it offers a lot more than sim speed Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.