Any body got early depth test working with a float texture (GL_DEPTH_COMPONENT_32F_NV) attached to an FBO on G80/G92 ?
I have tried it in my application and I can’t get it to work, whereas it works well with GL_DEPTH24_STENCIL8_EXT.
I use 169.21 on a G92 under windows XP.
Here’s what the guidelines say:
Clearing z is close to free and enables early z-cull optimizations
Don’t create triangles with holes in them (that is, avoid alpha test or texkill)
Don’t modify depth (that is, allow the GPU to use the interpolated depth
But since it works for GL_DEPTH24_STENCIL8_EXT I don’t think it’s an issue.
Perhaps it’s because the Z-buffer uses a lossless compression for GL_DEPTH24_STENCIL8_EXT to save bandwidth which has a good mapping for the early-Z while the 32F format is not compressed.