I am always running into the decision whether or not to make a mesh one connected mesh or to add small details as just separate parts to the model. Is there a best practices for this? Does it matter for games either way?
1 Answer
Fewer meshes equate to fewer draw calls, so generally you should try to go for as few as possible.
Split them when:
- one mesh would be so large that you'd need 32-bit index buffers to address (as this is generally too large)
- the sub-mesh uses different textures/shaders/materials
- the sub-mesh is somehow optional and might be detached or attached from the main mesh often (or similar reasons; basically you want to avoid having to update the main mesh's vertex buffer)
-
\$\begingroup\$ And what about the gains of a multi-part mesh in terms of allowing partial occlusion culling? I mean, in case all sub-parts are highpoly, couldn't it be peformant to cull mesh sub-parts that are occluded? \$\endgroup\$– Louis15Mar 18, 2016 at 2:28
-
1\$\begingroup\$ Yes, but LOD systems will also factor in there as well. It's a balancing act. \$\endgroup\$– user1430Mar 18, 2016 at 2:38