View Full Version : MR - Primary Renderer
07 July 2011, 02:34 AM
I have been using Mental Ray for a little while now and have come across a few posts of people recommending using other renderers than Scanline. If anyone could help explain ro direct me to information to me why? or when? you would use Raysterizing or Raytracing renderer I tried using Raysterizer and compared it to a render with Scanline that achieved a similar quality and the Raysterizer took much longer and had slightly changed the lighting particularly in the darker shadow areas.
If anyone could shed some light on this that would be greatly appreciated
07 July 2011, 10:36 PM
Different algorithms have different advantages. Over time this is becoming less important overall as machines get more powerful and faster.
Generally speaking, most things are going the way of raytracing. This means pure raytracing where your primary rays are raytraced and not scanline (Scanline Off or for mental ray in Maya, select "Raytracing"). Most modern scenes will render much faster with Raytracing and even save you memory in the process (renderers are "on demand" engines where objects are built/traced when necessary and not all at once for most software rendering) Raytracers shade pixels by sampling the pixel.
Rasterization is a more modern kind of Scanline where you are shading polygons that are then mapped to pixels. This means it is shading polygons rather than pixels and has a few benefits as such as a trade-off. Your anti-aliasing and shading quality are now separate and not tied together. Motion blur and DOF can be very fast because it saves the shading information and drags it across the screen. Hair and fur can achieve fine details without requiring a lot of shading overhead.
But the rasterization still slows down overall as triangle density increases. Renderers try to mitigate this with other tricks but the performance gain is decreasing.
Adding raytraced effects to rasterization also begins to eat away the benefits of rasterization. This is why a lot of REYES renderers advocate the baking of raytraced data for rendering. This is an extra step but means faster renders.
My advice (since it's obvious I'm leaning that way) is that you use Raytracing as your primary renderer for both primary and secondary rays. This is the default for other renderers as well like Vray and Arnold. Techniques will overlap in some cases.
If you have Maya 2012 I would combine this with Unified sampling. So now you will have faster render times even when using motion blur or depth of field. As well as simplified controls.
Rasterization still has its place but the tools in mental ray for rasterization are not as developed as something like Renderman or 3Delight. Nvidia's emphasis for this is raytracing.
07 July 2011, 07:53 AM
Thank you! Bitter that was most helpful I had to look up a few things you mentioned eg what REYES is... Ive mostly been rendering rather high poly scenes im guessing thats what Raysterizing i started doing a test comparing render times for raytracing and scanline but had to stop it early.. I will have another go later. Im guessing you would turn of raytracing for renderlayers with surface shader masks and AO etc? But I can see the advantages for beauty passes.
It was actually a post in another thread you had that got me onto this question so glad you could answer it.
07 July 2011, 01:20 AM
I would keep raytracing primary rays on for everything.
If you need a fast render with a lot of geometry but without secondary effects (reflections, final gathering, etc.) You can disable those in shaders and globally and still have a faster render time overall.
07 July 2011, 01:20 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.