I have just really started using the Blender raytracing engine bit by bit since the implementation of AO. I normally use the scanline renderer or Yafray if I need to do any raytracing. My scenes comprise of about four characters with a polycount of 10,000 to 20,000 or so each. I also use environments that include buildings in which only the close up buildings have high detail at the front and low detail on the back and sides. There are also streets, cars, street lights, etc. Again only the objects that are in the camera shot have higher detail.
I don’t think that Blenders rendering engine renders “only” the objects that the in the cameras view. I have even put raytraced objects and lamps on a separate layer from scanline object and lamp layers. I did this at first not for speed but because certain subdivided meshes had black artifacts visible on them with Blenders raytracer but not in Yafray renders. This practice helped to speed up these scene which were relatively lower poly that the ones with characters and buildings. With my current scenes this has not helped that much if any to speed up my raytrace render times.
I don’t know if some of you have seen my other lab renders that I have posted in the past. Most of my work in the past involved a 3d object or two or one character composited over another 3d scene. Now my work is a bit more complex to say the least.
I even have employed Ripstings script and Beast for grass, building and other 3d prop duplication. I have to use these sparingly as they can really slow down a 3d scene render in Blender. When I render one large scene with over 1 million polys with Blender it uses almost 3 gigs of physical ram. I know that this is common for large raytraced and scanline scene renders but I wish that it could be a little faster with the Blender raytracer.
I was wondering if anyone who is using Blender in a production environment like I am has any raytrace rendering optimization tips.
After I get around some of my “business” issues for this project I will have some nice stuff to show out of this. I’m using these new renders for web work to print ads and maybe some character animation work later on… I probably will just continue to use the scanline renderer for my work and compositing because it’s the pro thing to do. My work is 3d still images but I really work the scenes.
With every render job of a different angle I have to show the highest resolution possible. This could be like 20 shots in a production phase to select one or two frame renders. It could be something like 800x1000 render view shots with a final goal of something like 3200x4000 render size. Raytracing is way to slow and crazy to use for this kind of work. For little stuff ray tracing is ok. For complex renders it’s almost out of the question for the way that I work. I have worked out a way to make raytraced and scanline renders fit into the same presentations, I just think that raytracing looks better even if some of my clients don’t really care one way or another. And it is so easy to set up a raytracing scene and let the computer do some of the work for you.
You have to tweak things more to get scanline renders to work right. But in the end even with all the tweaking you still work faster than with raytracing.
Ah, one day raytracing in Blender will be as fast as some other rendering solutions. For now I dream of faster Blender raytracing run test with it. I wish that the Blender coders would add some more technology enhancements to Blenders scanline engine. Theeth’s new “Gradient Ramp” build is doing wonders for my scanline renders.
If anyone has any tips on this that would be great.