Search the Community
Showing results for tags 'optimise'.
Found 2 results
I want to scatter some trees as instance .rs files on my ground. I have a trees that contain around 130K points. I have scattered them on a grid, using redshift proxy files and instance. +My apporach: export each tree as .rs import on redshift proxy node. instance proxy scatter on grid the end.... rendertime fuqqen slooow(or is it)? +My comptuer: 32gb ram graphic cards one nvidia 1080 gtx and one nvidia 1070 gtx I have 1000 trees scattered, when i try to render the trees up close, one frames takes around 2 minutes to render, is this slow? When I checked out Atoms video on how to do something like this, he had a much faster render response, but his obj/rs. files was lower on points as mine objects are. Is this the main reason or am I using the redshift instance wrong? I'm not using his workflow, i import each .rs file on separate nodes. Thanks for any help at all I want to optimize my scene, but unsure if i need to reduce my points on each obj/rs or if i can do something else to reduce rendertime. SCENE FILE: redshiftProxy_v01.hiplc
I'm doing a flip sim of a splash. My particle separation is 0.006 which produces 14mil particles so pretty low. However because its quite a big splash the bounding box of the sim gets quite large so by the end of the sim I have 600mil voxels in the surface volume and 115mil voxels in the vel volume. So this is crashing my machine (which has 128GB ram)... My grid scale is the default of 2 but the vast majority of those voxels will be wasted since there are no particles in most of the bounding box. Is there any way to optimise this? The obvious way would be to use sparse volumes in dynamics instead of Houdini's native volumes but I imagine that would be creating a flip solver from scratch.. Or can I increase the grid scale without it negatively affecting the sim too much (i'm going to test this now..) I've attached a flip book of the sim https://drive.google.com/open?id=1k3Kexf-DJ5_ULPLt6dpsfx5f9LtW2OVz