Some Radiosity Related Questions

[MegaPixel]

The following questions are Radiosity related and I'm using MAX ver. 7.5.

 

1) Adaptive Subdivision Feature:

 

Has anyone used the new adaptive subdivision feature on anything relatively complex, like a fully furnished interior shot, with decent or better results then global meshing would acheive? I have a project which I'm trying to compare the results of both methods. My globally meshed scene down to 3 inches @ 80% quality process's in about 1.5 hours on my dual 3.0 xeon box w/2 Gigs memory. My adaptive subdivision settings on the same scene are at 12" max - 3" min - 6" Initial and 25% contrast (same quality proccesing). I've noticed not only does the processing take about ten times longer, but I can't even complete the calculations without running out of memory for some reason. How does the "Shoot Lights" selection affect the calculation? Which lights should I use in the Shoot Lights phase, Area's, Linears, Skylights and could these selctions be adding the calc. time? Just for your reference, the globably meshed scene file gets well over 300 megs in file size after the Rad Solution but never runs out of memory.

 

2) Backburner performance with large radiosity files:

 

I'm a little concerned that my 15 PC render farm will be useless to me if the scene file mentioned above threatens to run my workstation out of memory (which has 2 Gigs of ram). Most of the PC's on my farm have about 1 Gig of memory. Does the backburner server reduce alot of overhead that an opened program of MAX would normally need to work with? I was going to try meshing a really high quality solution overnight so that I could reuse direct illumination for faster render times. But if the super huge solution file proves too much for the farm, I may have to switch to regathering which will add 10 times more render time per frame with a certain degree of noticable splotchiness. Any thoughts?

 

3) Large Resolution Regather Render Settings (Minimal Splotchiness):

 

I've been trying to nail this one down for ages but I have never really been able to get rid of the noticable splochy's that regathering creates. It also seems that the settings for good regather results are different depending on what resolution your rendering will be and how complex your scene is. What I mean is, the same settings used for a 640x480 test render usually don't work for a 5000x3750 render. So my questions are:

--- Does Rad. mesh quality/density/refinement affect the quality or amount of splotchyness created at render time using regather?

--- High filter radius values usually get rid of the splotchy's but tend to smear away the details. Is there a rule of thumb or maximum F.Rad. value that should never be used?

--- When do you know how high a Rays value to use? Is 1000 enough? is 5000 too much? Obviously render times go through the roof at this point. I've never done a high resolution Radiosity render before and due to the amount of time it will take to complete, I'd really like to get the quality settings "spot-on" the first time around.

 

it's alot to chew on I know. Thanks for any input in advance - Mega

[MegaPixel]

"Pokes the carcus with a stick to see if there are any signs of life"...

 

I'm still very interested in learning more about the use of Adaptive Subdivision. You can skip the last 2 questions but throw me a bone on the first eh?

 

-Mega

[abicalho]

My adaptive subdivision settings on the same scene are at 12" max - 3" min - 6" Initial and 25% contrast (same quality proccesing). I've noticed not only does the processing take about ten times longer, but I can't even complete the calculations without running out of memory for some reason

 

The reason is simple: It's a too small mesh. You need to follow the same rules than Radiosity (without adaptive Subd): Mesh only what's needed and make sure the meshing size is good enough for that object - in your case, 3" for the whole scene is too small. Either exclude some objects from meshing, exclude them from Adaptive Meshing, or choose a higher Min Mesh size.

 

The usual candidates are the sky, ground plane, things that do not need to be meshed, but in the end are eating all our time and memory.

 

Hope it helps,

[MegaPixel]

Ya I usually exclude my Sky/Ground Plane objects from Adv. Lighting all together.

 

The reason I was shooting for such a low minimum mesh size, was because I was told somewhere here in these forums a while back that I would loose shadow detail on the floor/walls where small objects would cast (legs of chairs & tables on floor, etc..) unless I tightened up the mesh in thos areas. I've always had trouble getting those subtle shadow details to pop.

 

I was hoping that Adaptive Subdivision would keep me from having to micro manage the different mesh sizes in my scene which would save me time consuming guess work. Wrong impression?

 

The attached image is a test render I did a while back on the scene I'm working on. There is also a complete Kitchen behind the camera. It's changed a bit since then, but it was globaly meshed at 6in. - solved for 20% quality and regathered. Judging by the level of complexity of this scene, how might you have meshed it?

 

Thanks - Mega