is there a blender - yafray benchmark out there or should one simply use the sample scenes of the first release? the cornell box and exposure test scenes both seem to work allright, but they are not especially heavy - wich is what you are looking for when you are planning on running it in a cluster enviroment. especially compairing results from machine to machine would become alot simpler.
do i make any sense or are these words again the ramblings of a madman? :Z
you can do one yourself.
tip: a scene with a lot of glassball. or cube. and to automate the creation of the cube.try to script it. the use to test brasil with enormous infinit scene of simple object.
i know that i can do one myself - but. the point was that with a unified benchmark people with different hardware and software configurations can compare the results of the hardware they are running the benchmark on. in cluster configurations it can tell you if it does scale well or if you have to optimise the scene. ram impact an so on. not to mention - a widely spread unified benchmark will show preformance gains in yafray devellopment very well. ie - before 115,2 sec, after 87,5 etc. and lets say that on another configuration that optimisation didn’t fare as well - well then one can at least see it. i think it has more good points than bad at least.
the new (3.5) pov ray benchmark for instance - is very complex and uses several different techniques. and the reason for that is to squeeze the most out of the machine. a similair one cannot be that hard to work out. how about adding a fun twist to it - lets make it a contest 