Mental Ray, final gather and bucket artifacts

[AJLynn]

I'm working on some light studies with 3DSMax7 and mental ray, I've used Max quite a bit but I'm pretty new to mental ray and I'm using this to get some experience. The model is an import from Sketchup of an interior, the materials palette is using mainly architectural and basic blinns, and the lights are mental ray area and spot.

 

I did some renders using global illumination and wasn't too happy with the light distribution so I turned on final gather and ran a test render, was immediately much happier with the lighting, but it introduced artifacting - values changing from bucket to bucket, so the whole thing comes out with a blocky gridding effect that shows up on the blinn surfaces (just basic material with color white, for background walls where I didn't need anything fancy). I increased filter which seemed to make it go away, but when I increased resolution it was back. Thought it might be a sample size issue so I tried more samples and smaller radius, which didn't help.

 

Can anybody tell me what causes this kind of artifact and what I can do about it?

[Ras]

Hi AJ

 

If I understand you right I think the problem might be that you are using clipping planes with the camera?!? Can you attach a rendering to the post for us?

[AJLynn]

Yes, it has a clipping plane. (It's an interior section-perspective and I was being lazy.) If I change the model so I don't need the clipping plane, would that help?

 

I've lost the render that had it, but if I se it again I'll post an image. (Re-creating that render would take a long time. It was 6 megapixels.)

[BrianKitts]

6 megapixels???? are you rendering with your kodak camera?

[AJLynn]

6 megapixels???? are you rendering with your kodak camera?

 

No, I'm rendering for press quality

[AJLynn]

 

I recreated it using a low-res render with a small bucket size. It's worse at higher resolutions, I think the strength of the effect is proportional to the image size / bucket size ratio. You can see it along the top of the wall in back in the left half, and all through the right half.

 

(Also, I'm not going for photorealistic here, I think I've got the effect I'm looking for, but if somebody can explain why the back left corner is so well lit it would put my mind at ease.)

[abicalho]

You need higher number of samples on those buckets. Another option is to play with Radius both for GI and Final Gather.

 

Those are the only ways to solve it and they will likely increase your render time.

[Tim Fairbanks]

If you are saving and/or re-using your Photon Maps (pmap) or Final Gather map/solutions, you need to re-process them to acommodate the higher resolution final render. The size of the Pmap and Fg maps are directly proportional to the size of your final render.

 

The blockiness you're getting may be the result of using low-rez, test render Pmaps/Fg maps for your high-rez final.

 

Basically, it's like Shadow Maps: small maps are fine for small renderings, but the shadows get blocky at higher resolution because the shadow maps are too small.

 

TimF

[AJLynn]

If you are saving and/or re-using your Photon Maps (pmap) or Final Gather map/solutions, you need to re-process them to acommodate the higher resolution final render. The size of the Pmap and Fg maps are directly proportional to the size of your final render.

 

The blockiness you're getting may be the result of using low-rez, test render Pmaps/Fg maps for your high-rez final.

 

Basically, it's like Shadow Maps: small maps are fine for small renderings, but the shadows get blocky at higher resolution because the shadow maps are too small.

 

TimF

 

I don't think it was that - I wiped the files and ran it at low resolution with a smaller bucket and got the problem anyway.

 

I did fix it though. Well, not "fix" per se, I finally broke down and installed the 7.5 update extensions, and I'm not getting that problem any more - they've made some changes to the final gather algorithms, it's a lot smoother and a heck of a lot slower.

[samov]

hmmm it's the smell mr. Anderson......

 

no just joking ... it's the thickness of the wall or anything else.... I noticed when i wanted to make a box zero height it does that.....it's like the photons sometimes get through the surface..... so use the shell modifier when necesary.....

 

do not make things in max that can't be build in reality like 0 height/width things.... even e sheet of paper doesn't have 0 zero height.....

 

and in max 7.5 USE HALF THE SAMPLES U WOULD USE..... it will still be better then mentalray 3.3 .....

 

 

hope this helps.... bye

[AJLynn]

hmmm it's the smell mr. Anderson......

 

no just joking ... it's the thickness of the wall or anything else.... I noticed when i wanted to make a box zero height it does that.....it's like the photons sometimes get through the surface..... so use the shell modifier when necesary.....

 

do not make things in max that can't be build in reality like 0 height/width things.... even e sheet of paper doesn't have 0 zero height.....

 

and in max 7.5 USE HALF THE SAMPLES U WOULD USE..... it will still be better then mentalray 3.3 .....

 

 

hope this helps.... bye

 

I thought of that, but the walls were realistically thick - maybe something to do with the way Sketchup->DWG->MAX geometry works out, I'm not so fluent in that area yet.

 

But yeah, even with 300 samples (down from 1000) mr 3.4 is way better than mr 3.3. I like it so much I went and got a used Precision on Ebay to use as a DBR render node. (You would not believe how cheap those things are these days, compared to a year or two ago - dual Xeon 2GHz with all the stuff, standard price is $400! I could get 4, and with DBR even distributing the photon map and final gather work, I'd have a full-blown render farm with 8 CPUs, which is the max on one license, for $1600.)