2
$\begingroup$

I remember reading an article not so long ago about a node setup to significantly speed up render times in Cycles by telling the engine to only calculate materials based on the portion of the object visible to the camera and not the entire object.

I think it was posted on the official blender page/ blender nation but I cant seem to find this information anymore.

Does anyone know where that article/video is? Does anyone else remember this?

|improve this question
$\endgroup$
1
$\begingroup$

It's not really limiting the render to what's visible to the camera, but you can limit some effects to camera, such as bump maps.

Mixing your bump BW factor with a 50% grey color with light path's "is camera ray" to make your bump map only visible to the camera. Add (math node, clamp checked) "is camera ray" with "is glossy ray" to have your bump visible in glossy reflections too.

enter image description here

|improve this answer
$\endgroup$
  • $\begingroup$ not sure about that but i think it was a Kent Trammel's advice to speed up rendering. $\endgroup$ – Bithur Feb 18 '15 at 11:33
  • 1
    $\begingroup$ This setup messes your bump - its different, also has negligible effect on render time, the time difference is 0.03% in all glass and glossy scene! $\endgroup$ – Jaroslav Jerryno Novotny Feb 18 '15 at 12:15
  • $\begingroup$ Kent Trammel says "Bump maps can significantly hinder render computation. Despite the fact that they are an illusion that don’t actually displace geometry, Cycles allows the illusion to affect light bouncing and global illumination. Therefore, if we exclude our bump maps from being computed by in global illumination we can save significant render time." up to 30% cgcookie.com/blender/2014/07/10/… $\endgroup$ – Bithur Feb 18 '15 at 13:17
  • $\begingroup$ so if Kent Trammel says something its true? Did you test it? In comments there on cgcookie I see people reporting no speed improvements and mixed feelings. It also probabaly differs for CPU and GPU. Also was on an old blender release, things might be optimized since then. The setup you posted clearly doesn't work, because it changes the bump when rendered, try it for yourself on a glass sphere. And measure the speed-difference, far from 30%, you'll be happy to even get 0.1% $\endgroup$ – Jaroslav Jerryno Novotny Feb 18 '15 at 13:25
  • $\begingroup$ if Jerryno says 0.03% it's true? i do trust him more than you. up to 30% doesn't mean always, depending on the scene. and a 50% gray bump means no bump, but can be black. btw, i think the question was about this specific setup. $\endgroup$ – Bithur Feb 18 '15 at 13:32
-2
$\begingroup$

While this method of achieving faster CyclesR rendering time does not include node setups,it does involve scene visibility, particularly tile sizes.

Tiles are the black squares that obscure the scene currently undergoing rendering by Blender. The purpose of tiles are to allow for more manageable rendering by the processor and thus minimizing chances of Blender crashing by limiting how much data the processor renders at a given time, saving memory in the process.

Under the 'Performance' section in the 'Render' Properties window, you will find the 'Tile Size' selection. These dimensions represent the pixel size of tiles, where dimensions in powers of two provide faster render times for processors, 16x16 being the optimal choice.

http://www.blenderguru.com/articles/4-easy-ways-to-speed-up-cycles/#.VLsvs0eorMN It's the third render optimization tip listed in the provided link.

Hope this serves as another useful method aside from what you're seeking.

|improve this answer
$\endgroup$
  • 2
    $\begingroup$ Link only answers are not good practice, please include some of the content here, in case the link brakes. $\endgroup$ – David♦ Jan 18 '15 at 4:11
  • 1
    $\begingroup$ Tile sizes being a power of two typically makes no difference at all. See: code.gregzaal.com/auto-tile-size $\endgroup$ – Greg Zaal Feb 18 '15 at 10:58
  • 2
    $\begingroup$ The optimal tile size is actually dependent on your system and power-of-two has no power here.. $\endgroup$ – Jaroslav Jerryno Novotny Feb 18 '15 at 12:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.