If you have mostly outdoor scenes, then forward rendering is the better approach i think. But what if you are also doing a lot of post process effects such as atmoshperic scattering, SSAO, etc..?
For example i'm implementing Bruneton's precomputed atmospheric scattering and i find myself torn between either calculating the scattering results for every visible object or just do it as post-process and be done with it. If i choose the latter, then i will need depth and normals... Might as well just use a gbuffer, no?
But then there's the headache of implementing anti-aliasing and handling alpha transparency...
Edited by Waaayoff, 10 July 2013 - 07:15 AM.