I was testing face culling via simple cubes drawn on the screen (in theory 3 faces would not be drawn at any time during the rotation). I corrected my vertices so that everything faces the right way and then did glEnable(GL_CULL_FACE). The problem is that I don’t see any change in the frame rate I’m getting (that is, cull_face has not appeared to improve performance at all).
Are my conditions simply not ones where culling helps, and if so, what types of situations does culling have a distinct advantage?
About my test case: I simply draw a bunch of cubes dispersed in a large cube. In increased the number of cubes until my frame rate dropped significantly so I could measure performance improvements. I definitely saw improvements when changing how vertices are handled. But I see no change with face culling.
Rasterizing polygons, especially with simple shaders / lighting, is so fast on modern GPUs that for small polygons you may not be able to notice the difference.
If you want to see the difference more clearly, try using larger polygons or more complicated shader programs to increase the percent of your time being spent on shading.