I’ve noticed some issues with Mental Ray and the Daylight System. Not the typical “shadow flickering” from Final Gather. The more common problem appears to be a luminance variance on random frames; a split frame sample is attached. The magenta tic marks in the enlarged image show the split line between 2 frames; the text (and scene overall) is lighter in one frame vs. the other. The lighting is fairly basic: one daylight system, FG set to “high” preset, sampling at 1/16. I have tried pre-computing FG and interpolating, still get the luma issue. This has happened in other scenes, if you look at my web site under “recent work”, you’ll spot luma shifts in the solar array animation. ( http://www.firemist.com )

The second issue is perhaps described as “noisy buckets”. Same scene as above, 2nd image shows a couple buckets with dark, slightly noisy output.

This is from network rendering, but not attributable to any specific machines. Seems random. Tried saving in TGA and PNGs, same issues, but different frames.

WinXP32, SP3, /3GB switch on machines with 4+ GB RAM, dual-quad Xeons.
Max 2010 with update.

Any insight appreciated.
Thanks,
Jeff

Max/Composite 2012 (subscription)
Win7-64pro, Intel i7-hex on SuperMicro mobo, 12 GB RAM
nVidia Quadro 5000, render farmette on BB2012

	Attachment

	Attachment

Jeff, are you getting the same issue (Luminance differences) if you render on a single machine?
I had once a similar issue by rendering an animated Daylight cycle using Perez all whether and SKY skylight models, anyway the difference was more pronounce than in image you posted. I didn’t find an explanation and solution, therefore I’m using the default Haze Driven model in animations.
What about the second issue I’m not able to see “noisy buckets” in the posted image.

ivan

Thanks, Ivan. More testing on my part is needed. Looking at the 2nd image again, I think it is also a luminance issue; it looked like noise to me at first because of the texture. If you look at the enlarged part of the “grass”, you should notice 2 slightly darker squares. These appear randomly in the animation, so looked more like noise to me.

-jeff

Max/Composite 2012 (subscription)
Win7-64pro, Intel i7-hex on SuperMicro mobo, 12 GB RAM
nVidia Quadro 5000, render farmette on BB2012

Your lighting difference between frames appears to be because your sun has actually moved. I can see two different shadows from the horizontal part of the yellow object cast onto the diagonal part. Why this would happen, I’m not sure, but you might want to check your daylight system’s time isn’t changing between frames.

The “noisy buckets” problem is odd. Are we actually looking at a finished render is are those buckets still showing the final gather preview?

I think I’ve narrowed down some of the issues… After further, if not extensive, testing what I’ve found is summarized below. For my newest machines (Xeon 5520), I tested some renders by turning OFF “virtualization” and “simultaneous Multithreading” (hyperThreading), noted below by HT on or HT off

I have little experience using the CIE skylight model, so the luma variance in that may be due to FG settings, or operator error ;>)
The darker (previously referred to as ‘noisy’) buckets with hyperthreading seems to definitely be a bug, and I’ll be reporting it as such.

For now, I think I’ll stick to the “haze” model for my daylight systems. My testing when I first installed the new render machines showed about a 10-15% increasing in MR render speed when using hyperthreading (16 cores per machine!), which is called “simultaneous multithreading” in the new BIOS.

Hope it may be of help,
Jeff

-Jeff

Max/Composite 2012 (subscription)
Win7-64pro, Intel i7-hex on SuperMicro mobo, 12 GB RAM
nVidia Quadro 5000, render farmette on BB2012

I have the same issue I’m glad to see I’m not having a random error but a reproducible error.  I’ve already post some time ago a similar problem. Also I found a similar issue when I tried to render CIE sky model in a 8-cores Xeon E5462, but this noisy buckets are very sporadic and almost unnoticeable but still happens

I seemed to fix those issues with turning down the FG sampling, especially in outdoor scenes. (No need for high samples outdoors) and also re-calculating FG every frame. Will this work for you?

"If I see that damn teapot one more time....”

3ds Max 6.5 - Max 2012
Windows 7 Intel Core 2 Quad Q9400 2.13 GHZ
64-Bit, 8 GB RAM

Video Card: Galaxy GeForce 210 (Nvidia)

My Website: http://drmoore3d.com

You can see the problem more clearly this image

	Attachment