Its instancedbuffergeometry (not instancedmesh) better than instancedmesh at performance?

Its instancedbuffergeometry (not instancedmesh) better than instancedmesh at performance? If its better… How to implement it with a world of cubes like this?:

<img src = "cursor.png" id = "cursor">
New Game Load Game Options
Options Save and Quit to Title

Edit 1: the site is this, becuase I cant write code here:

http://xw2.rf.gd/optimizing3/

Edit 2: I found instancedmesh is better than the other, so instancedmesh is definitively the best, because its easy to refer to each one of all instancedmesh voxels.

I’ve opened your link and I barely get 10fps in my computer.
If it helps I’ve checked in chrome profiler that the matrix calculations and the render loop (some bufferData calls) are using all available CPU power, the GPU is barely used in comparison.

As your voxels are in a fixed position and they don’t change you shouldn’t need to be computing all that each frame, I’m sure you can optimize it A LOT and get perfect performance even in low end machines.

There shouldn’t be a performance difference between the two, in most cases. InstancedMesh offers an easier high-level API, I think users will want that. InstancedBufferGeometry gives you a bit more low-level control, but would usually require you to write shaders, … I would probably reach for InstancedBufferGeometry only if the high-level features of InstancedMesh are getting in the way of something you want to do.

2 Likes