I tend to ignore the hierarchical z buffers lately as the z-prepass trumps it in both performance and added abilities.[/quote]Instead of sorting front-to-back to take advantage of the HZB, you can use a ZPP to take advantage of the HZB...
[quote name='AllEightUp' timestamp='1311255734' post='4838439']
[quote name='Hodgman' timestamp='1311212101' post='4838230']
Front-to-back sorting of opaques is designed to take advantage of the hierarchical z-buffer.
However, a common alternative [to front-to-back sorting] is the z-pre-pass, [which] gives you the benefits of the hierarchical z-buffer, without having to use a front-to-back ordering.
...but instead of either sorting or using a ZPP to exploit the HZB, you... use use a ZPP?
[/quote]
I should have probably stated this as more in terms that given my scenes are semi-ordered due to the culling system, a better sort may reduce a bit of overdraw but the CPU cost doesn't ballance out compared to the GPU gain. Again, this is a by product of my pipeline which is optimized for a specific type of scene in a style of game which is not likely going to be a genericly applicable result. But, still, given how fast you can do the ZPP anymore, leaving it with even a lot of overdraw may still be a win given a more simplistic pipeline on the CPU side. (NOTE: I do notice that ZPP on NVidia cards is notably slower than AMD, probably due to having fewer processors, I may find the sort is worth it when I do better testing on my NVidia cards.)