Maxwell Render | Official public forum

There is a big change in benchmark when the leaves are added 374 vs 531. The leaves are a duplicate floor geo where 90% of the surface is invisible (totally clipped)



This is an old issue, I was just hoping it had been adressed, seems a bit screwy no?

eric nixon wrote:The leaves are a duplicate floor geo where 90% of the surface is invisible (totally clipped)

I think this behavior is more or less normal. The floor object fills almost 30% of the screen. If you duplicate it and add a clipmap it means that approximately 30% of the rays will have to impact it, check that there is a clipmap texture, read a pixel value and evaluate if the ray should continue or not. This is is way slower than not doing nothing, obviously So if I am not wrong you will get more or less the same performance if the clipmap has many leafs instead of just a few.

In that case if an increase of 30% in render times is not an option I would try other approaches, like using separated clipmaps for different leafs regions, or even using real geometry for the leafs if there are not many.

Yes, I will just delete the 80% of the poly's which have no leaves / are not needed.. once I'm happy with where the leaves lie. It is worth mentioning this in teh documents as traditional 3d workflow could lead to a heavy slow-down especially in the case of a clipmapped cloud-scape.

I guess my approach of keeping one large map for all leaves was slowing it down (reading that big map each time), best to cut up that map into regions like you said. Not so easy to do clouds in pieces though.

One last question, if a hazy cloudscape was made using the new object opacity option, rather than a material opacity would that be any more efficient?

eric nixon wrote: One last question, if a hazy cloudscape was made using the new object opacity option, rather than a material opacity would that be any more efficient?

It should be more or less the same, but you skip the texture access, that can be significant in some cases.