[iandavis]

I haven't been able to find a way. it's not too difficult tho to get a good render.

some tricks to to 'test strips' for a scene:
(based on Lightwave, but I'm assuming these abilities are also present in other apps)

- Create a low-poly version of your scene, it's very easy if you are using SubD, otherwise run a poly reducing plug... Then get the exposure correct and record your settings for the final render with the full scene. From my observations, after about 10 minutes (or 7-10 iterations) the exposure is pretty solid. Darken it down a bit at this point and let the render get to 12 iterations. Record settings... this should do it. (Maxwell is much quicker with lower geometry...go figure)

- For textures such as stainless, or marble flooring, duplicate your scene and load only a test tile, or sphere... get the texture perfect, then save it as a surface... Do the same for emitters and dialectric objects. (must be loaded as independant objects. (btw)

- Try to leave the shutterspeed at the default 1/60 and change the other parameters which can be changed before hitting the render button. This way you can choose "-hd" for larger renders and walk away. The feedback display doesn't work properly during a hard drive render and you need to set the exposure before render in this case. AND... if you cant get at the display adjustments, you cant change the shutterspeed...

so far I've found it relatively easy to end up with a good exposure... and though maxwell has some issues, no fooling, I have found myself actually using it for various paying work, since it can render radiosity/caustic/global illum quality and honestly doesn't take too long... (that is with physical sky, with emitters it's a different story).

It must also be recognized that it IS a beta, and there are still some serious issues, for example, you cannot render glass correctly, it casts a dark shadow and blurs objects behind it... but I'm confident all will be made right by V1... and personally I don't regret buying it for ONE SECOND.

mmmmmmmmmmaxwellrender.

cheers.

[Mihai]

Geometry shouldn't be an issue, but the materials CAN be can be an issue.

When you are watching the render progress in the mxi viewer, you will not see an accurate representation of light levels until it gets to around 6-7 sampling levels. Then the light will change to show you an accurate view.

No need to make test scenes with less polys or no materials, just make a tiny test render and watch the light levels.

If you're on Max you can also use the render region option, but this will only speed up renderings if you're not using caustics. Caustics have to be calculated for the entire scene, even if you're using the render region, so you get no speed improvements in this case. But it can be very useful if you want to test bumpmaps/diffuse maps on a full size render.

[iandavis]

Mihai,

I have noticed quite a link between geometry complexity and rendertime. Mind you, we are talking 500,000 triangles or above.

One guy on the forum had a render go from 40 hours to 40 minutes for the same quality by reducing the geometry from 1.5million polys to 600,000...

So, assuming someone is doing an interior render, a scene with 500,000 polys will take quite a bit longer to render then it's 'teststrip' scene reduced to a few thousand polys. In other words, you save more time in rendering then you spend creating the 'teststrip' geometry.

I will do some tests when I get home, I now want to see how dramatic this actually is... in hard numbers so to speak.

cheers

[Mihai]

That's not normal behavior, try rendering a few cubes with 100 polys each, then subdivide them so they are 1 million polys, the time should be almost the same (ofcourse the voxelisation time will increase a bit, but not much).

What could have happened when he made a low poly version of an interior for example is that the geometry changed so much that the room perhaps had 'holes' where light could reach in from different directions, making the calculations a lot faster (as if you had removed a wall).

Second case if you have geometry like leaves and so on, which ofcourse makes lots of light bounces. If you delete a lot of those leaves than render time could decrease a lot.

But in all other cases, if polycount doesn't influence number of light bounces, it will have almost no impact on rendertime.

[DELETED]

DELETED