New era of rendering in front of us
The nature of look development has been always a tough one due to the limitations imposed by the shading and illumination stages along with the rendering process that needs to be kickstarted to check your changes. So far all sorts of trickery have been used to get things easier, delay decision making by sending layers to the compositing, precomputing of very intense processes as reusable maps and what not.
But, now we are starting to see the new generation of distributed rendering, or more accurately, massively distributed rendering projects. Mental Images has already done a lot of work on this field with the Reality Server but now a further step is cloud computing, and although I am aware of all the still remaining obstacles I can imagine the power of using 100.000 machines to crunch 100.000 frames. Furthermore, nVidia new Tesla GPU capabilities to build a cluster of GPUs for the desktop also promises the ability to interactive rendering
So am I talking about Amazon Web Services? Not really, I am talking about Condor, and specially the new innitiative, the world wide grid (aka gEclipse project).
But what about licenses? well, I am sure a new business model will make sense so you pay by day of rendering no matter how many licenses you use, or a model based on your scene size. In any case, there are good reasons so change the actual situation.
So why do we need a new rendering approach? well, on the massively distributed rendering what is clear is that physically correct lighting and rendering (unbiased one specially) is going to get more popular and we can’t afford in advertising and film some render times like 60 hours per frame with a limited pool of computers (aka renderfarm). This does suddenly make a lot of sense if you consider a infinite scale solution is affordable and the other technical limitations (transfer of scenes and recollection of images) are solved, but P2P seems to be the perfect tool so is just a matter of time till this is on the table.
Will this service be provided by someone that uses Mental Ray renderer? or Maxwell renderer? clearly the times of old-hacky style (aka Renderman) are gone and although we use it the end is, at least to me, very clear.
Have a look at the next links.
Leave a Reply