Hi, I am very impressed with the Houdini20+KarmaXPU update.
However, when using opacity maps, the rendering speed is much slower than without them.
Is there a feature like "OpacityMode-Clip" in V-ray or "Clip mask" in Clarisse?
I hope there is some workaround or update to this.
Thank you.
Slowdown when using Opacity Map [XPU]
2649 5 2-
- tkasota
- Member
- 1 posts
- Joined: Aug. 2020
- Offline
-
- jsmack
- Member
- 8173 posts
- Joined: Sept. 2011
- Offline
-
- Heileif
- Member
- 259 posts
- Joined: Jan. 2015
- Offline
-
- brians
- Staff
- 581 posts
- Joined: May 2019
- Offline
Its more that opaque shadows in XPU are very fast, rather than semi-trans shadows being slow. Its because there is a very fast hardware path we can take for shadows when we know geometry is opaque. When it is anything else (eg semi-trans, cutout-alpha, etc...) then we need to execute a shader for each hit, which slows down traversal a lot.
I'm curious if you get any speed benefit from emulating your own "clip" opacity mode.
It can be done easily with materialX (see attached scene)
Also, the USDPreviewSurface does "clip" opacity by default, so you could try use that instead.
I'm curious if you get any speed benefit from emulating your own "clip" opacity mode.
It can be done easily with materialX (see attached scene)
Also, the USDPreviewSurface does "clip" opacity by default, so you could try use that instead.
-
- CoreSign
- Member
- 30 posts
- Joined: Aug. 2016
- Offline
-
- Latimerias
- Member
- 38 posts
- Joined: Aug. 2020
- Offline
I did not find clipping the opacity map to make any consistent or noticeable difference, sometimes it was slower than no-clipped opacity and sometimes it was faster but never by more than 5% difference between the two. Using raw geometry on the other hand averaged 4x faster at different grazing angles and object counts. It seems like this is to be expected though, how fast can a ton of opacity maps really be when shooting rays is something i dont know? What are the differences compared to vrays opacity clip, it would be interesting to know about comparable technology performance?
-
- Quick Links