Hi everyone,

I have been really looking for a faster version of POVRay to be able to render
scenes which take about 1 hour with(2 cores and POVRay-3.7 machine) or so in
less than 1 minute (I know it's too ambitious). How do you guys go about
rendering the scenes which take almost a day or even more? There are many
artists here in the forums maybe their expert comments on this would be worth
gold. Are there any possibilities to still make it super quick modifying the
Parallelisable POVRay-3.7? Ofcourse the standard possibilities are

(a) Get a new powerful machine which has more cores.
(b) Run it across over to several machines by splitting the task of rendering.

But has anyone of you ever worked with some publicly available cloud clusters to
aid quick rendering of the images? There is a service provided by Amazon but
it's not free. I'm wondering if anyone tried something of this sort. I'm really
in my desperation to get the rendering of 12K images (each image rendering takes
about 30min-1hour) done within few hours. I'm using condor service at the moment
but I only have access to around ~500 machines. To render 12K images within few
hours (to be specific I'd say 2-3 hours) I'd need more machines and I don't have
them. Any possible suggestions would be most welcome or any possibilities of
changing the POVRay code?

Kind Regards,
Ankur

Post a reply to this message

I don't think there is a "makeItFaster" switch in the code source that's OFF by
default.
If it was "easy" to make POV-Ray faster, I think the devs would have done it
already, at least in the 3.7 branch.
So considering the engine itself is at full capacity, you have only two options
to speed up the rendering.
The first is, as you mentioned, distributing the load. Rendering subset of an
image is easy with just a few INI parameters, same for an animation.
The second is simplifying the scene. There are many potential optimizations to
make, often resulting in a bigger speed boost than an hypothetical source code
improvement, or a PC upgrade.
Slow things usually involve radiosity and/or photons. Try playing with the
parameters here to see if you can get acceptable results in reasonable time...

Post a reply to this message

Am 09.11.2011 07:06, schrieb Gyscos:

> So considering the engine itself is at full capacity, you have only two options
> to speed up the rendering.
> The first is, as you mentioned, distributing the load. Rendering subset of an
> image is easy with just a few INI parameters, same for an animation.

As a matter of fact, the architectural changes in POV-Ray 3.7 that 
allowed for multi-core support were originally implemented for network 
cluster rendering support, and that feature still remains high on the 
agenda. The advent of multi-core processors just has provided an 
unexpected opportunity to benefit from that ongoing development even 
before it is fully completed.

(At least that's the big picture, which doesn't hold for every detail. 
For instance, the current multi-core radiosity implementation is just an 
interim solution, and will have to be changed again for network 
rendering support.)

Post a reply to this message