I'm doing a fractal with transparant spheres inside transparent spheres ( a
transmit of 0.98 ) - 350000 of them.  add in reflections and specular
reflections and I can't get away with anything less than a max_trace_level
of 25.

Took 10 hours to do 48% of a test render on an Athlon XP 1700+ win 512Meg
RAM at 1024*768.   Youch.



"Yadgar" <yad### [at] tiscalinetde> wrote in message
news:3CA73A99.9189EC05@tiscalinet.de...
> Hi tracers!
>
> (excuse me for the large PNG file - but I thought JPG would have harmed
> the scene too much...)
>
> Perhaps you remember my posting from last week where I complained about
> a mysterious error message indicating "too
> many nested objects". Obviously, the reason for this was that in my old
> scene, there were too many objects within objects
> and again within objects.
>
> Meanwhile, I rendered a new scene (attached here), and though it
> contains no less than 346,201 spheres, I essentially had no
> problems to get it calculated (except for I had to change to Linux as
> Windows occupied too much RAM). Still there are quite
> a lot of spheres rendered unnecessarily twice, but since they are
> coincident, not nested objects, it nevertheless worked fine.
>
> Each sphere is surrounded by three rings of smaller spheres along the
> x-y, x-z and y-z plane, each ring contains 8 spheres,
> but I didn't program exceptions for those spheres at the intersections
> of the rings, so around each larger sphere, 6 spheres
> were rendered twice. With #if ... #end statements sorting them out, I
> could reduce the overall number of spheres to be
> rendered dramatically from
> 1+24+24^2+24^3+24^4=346,201 down to 1+18+18^2+18^3+18^4=111,151!
> Also, this should substantially decrease the rendering time - which now
> was around 16 hours on an AMD K6-II at 380 MHz
> with 256 megs (yes, I know, hilarious antediluvian scrap...), I didn't
> keep accurate records...
>
> And then I also will re-try my old cube fractal scene, probably refining
> it by adding subtractions in the centers of each cube
> side... until I arrive at something like a Sierpinsky box!
>
> Onward to more and more sophisticated 3d fractals - any suggestions
> where to look for general introductions into fractal
> mathematics and modelling?
>
> See you in Khyberspace!
>
> Yadgar
>
> Now playing: Earthrise (Steve Hillage)
>
>
>
>
>
>

Post a reply to this message