koobas.hobune.stream
Optimizing Performance with tris
Hi,
I'm trying to optimize my 3D-model but I noticed that you could decrease the number of triangles by using a badly modeled object.
Example: 
Left: The two cylinders are connected with a ring of Quads
Right: The cylinders are not connected
Any experienced 3D-Artist would highly recommend the first solution, but when you hide the upper cylinder and convert the polygons into tris... 
...you'll see that the left ring consists of 64 tris and the right ring consists of only 30 tris.
I have two questions:
Does the number of tris even matter? (because the vertex-shader is executed for each vertex and the fragment-shader for each pixel - in this case, the left solution would be more performant since the number of vertices is the same and the area of the screen is smaller - thus decreasing the number of pixels)
Would you recommend the left (more elegant) or the right (possibly more performant) solution?
Thanks for your help :)
This smells like micro / premature optimization and there is more to 3D modelling than $$anonymous$$imizing triangle count and performance.
