When people talk about network renderdering are they referring to the ability to ship entire scenes off to distributed nodes to be rendered? Or is it the ability to split a single scene among multiple nodes to be rendered?
Or both?
Thanks.
Most applications have some sort of network rendering option to allow you to spread out different frames of animations to different processors on the same software network to speed things along. You can render out single frames by taking advantage of network rendering setups by using camera tiling which some renderers will support. That way it splits up each tile (much like frames) and distributes them to different computers to get you your purty image quickly. Hope that helps.
Yes. Thanks.
A question about rendering frames for animation v.s. stills. When rendering frames do (some) renderers only compute the deltas and then apply them to a base image to produce a frame, or do they tend to be rendered from start to finish independently?
Well, I only have experience network rendering with Cinema 4d with my computers here and here’s how it worked… Say you have a 25 frame animation. If you have 4 computers on your render farm then cinema 4d would split it up so computer A gets frames 1-5, computer B, 6-10, computer C gets 11-15, then computer D gets 12-20. They render each frame in full like any other render but it just makes life easier having more pcs doing the work of one. For stills it’s the same idea so if you have a 12 tile camera setup it will split up each tile by groups acordingly.
There are basically a few ways out there… And the term Network rendering means nothing else that also other boxes than the one you are sitting in front of are working on a project via the network…
One is batch rendering. A central server process or pile of scripts is distributing project data to nodes and then starts render passes on those nodes and then fetches the resulting frames into a repository and from there the postprocessing and compositing takes place. In a simple way think about it as a bunch of PC’s where Marcus Civis wrote a script which copies the ZBrush project file to each PC, then starts ZBrush which loads the scene, renders a picture and saves it. Then the script grabs all thos scenes and voila… a bunch of scenes in the time of one plus the time it took to copy and administer that stuff…
Nodes can be real PC boxes or more common so called ‘blades’ which are procesor boards which represent a complete PC which are build to fit into a rack. Saves cost and space.
The C4D way is working like that as well. The nodes controlled by the C4D renderer are rendering a complete frame each.
XSI and FinalRender also support a mode which is satellite rendering. That works in the viewport/preview window. I have only seen it with bucket renderers which process a little tile of the same frame/picture with each processor or thread if there is only hyperthreading. ON my bog box I have 4 little tiles cruising all over the screen building up the picture in XSI. And when I plop my Laptop into the equation then it’s 6. Fun to look at.
C4D’s advanced renderer renders individual lines and thus I see 4 lines being processed but it does not branch out on the network.
Network rendering is tricky as all sorts of maps and information have to be correctly distributed or the render look funny. Some tiles might be lighter some darker… and other terrible things…
Besides those problems other challenges come with animation/ multiple frames. Sliding textures, illumination changes from frame to frame (many algorithms use statistical methods to solve lighting and thus results can differ from frame to frame). There are ways to work around this. But it’s another science on top of the other 3D related sciences.
Since I started to dig deeper into all this Mental Ray just earns more and more credits with me. Static Stills can be rendered with anything which comes along. But animation or large sizes or many hundred frames without dying… That’s magic!
Lemo
Cool, thanks for the information guys.