For the texture, you need to use an elliptical filter, i dont know how this is implemented inside max, but check elliptical filter with the help docs.
Hello and thanks for a spendid thread.
I have the following problem; i have set up 3 machines with complete install of 3dmax 8. I am using one as master and the other two as satelites to speed up rendering. All machines running XP.
I set up my benchmark scene, and i follow the instructions about using the hilbert bucket and only sending geometry when needed.
I set up hosts and ports, and I use only my local machine for shaders and maps, so that I do not have to rely on shaders and map files on the satelite machines.
I get the following error messages when using the distibuted bucket rendering:
LINK23.n error 191029: failed to link library on host Satelite1
The picture renders ok, and i can read from the progress bar that MR actually appears to be using the satelite machines for rendering - but I can’t measure any acceleration in the rendering (it takes the same amount of time using 2 satelites as when i am rendering on my local machine).
Any help or tips highly appreciated, even the most newbie tips, - I believe I’ve read most of the threads, without luck,
I’ve noticed that distributed bucket rendering for a single image doesn’t seem to save that much time. Unfortunately I can’t really give you any concrete help since I never really got it to work faster (I think in my case that it had something to do with the way the different slaves needed to access the maps required and the performance of the various slaves).
Anyway, the easiest way to know if your render is using the extra cpus is to see how many buckets are being rendered. In my case I had 12 buckets rendering at the same time, quite a sight Problem was that sometimes buckets would get stuck on some areas of the image for a ridiculously long time and thus nullifying the gain from having slaves…
Sorry for lack of help :sad:
Try setting your Bitmap Filtering to Summed Area instead of Pyramidal and reducing the Blur to 0 (or whatever is closest, right click the spinner next to the Blur value). That should sharpen it a little at least. No idea why the rpf files break mr though…
Thanks for the reply Urgaffel.
I’ve investigated a bit furter and found the following observations I would like to share.
First of all, Autodesk states that the distributed bucket rendering is ideal for quick previews of single renderings, according to this article: http://www.virtualpartners.com/PDFfiles/3dsmax_and_mentalray.pdf
However, setting up a DBR renderfarm with only 2 or 3 satellites, like I did, give almost no gain in images that takes less that one minute on your local machine. The network communication eats up all the upside of using the distributed satellites.
However, if you add extra satellites, preferably all 8, and if you are rendering images that take from 5-6 minutes to 5-6 hours, you will really feel the difference.
I’ve experienced that the acceleration is non-linear, almost logarithmic, as a render that takes about 8 hours on my local machine took only 17 minutes when hooked up with 6 satelittes.
However - ALL errormessages in the Mental message window will seriously slow down your render. F.ex. I recieved an error that a certain library could not be found on the satellite machines. After a few reboots the error stopped appearing, and the render time went way down.
Likewise, if f.ex. a HDRI is in a wrong colourformat and thus i handled as a RAW image, the rendertime explodes.
And images render quicker if you distribute the maps and shaders to the individual satellites.
So my advice is; get rid of all errors in the MR prompt, distribute the textures, if you are using bitmap based textures, that is, install more than a few satellites and use it for renders that takes +10 minutes.
Any other advice is appreciated - for instance on how to install the satellite on a new system without having to install an entire 3dMax8.
Hi everybody, thanks for making this post
I’ve been having this problem for a while now and it’s completly stopping me from getting this project any further.
Here is an image to show my problem: [img]http://membres.lycos.fr/cyberman2/coppermine/albums/userpics/10001/ostidcalissdeyeuxamarde.jpg[/img] render is in 3ds max/mental ray left image has no Final Gather while right image has Final gather
as you noticed, the eyes becomes very self illuminated with final gather and i cant figured out why… thank anyone who can help me!
mostly i used the settings from the tutorial with globulous ( [http://www.lamrug.org/resources/doc...in-tutorial.pdf](http://www.lamrug.org/resources/doc/sss-skin-tutorial.pdf) )
i used different scatter group for the face and the eyes (the tutorial says to use the same but if i do it REALLY gives bad results, eyes become even more white and uglier)
I’m having this big problem with Mental Ray, every bucket just stays there without showing what it just renders. The silly thing about it is that it takes as much time as it would if it was looking fine.
Render size 2000, 3000 or 4000 square px.
Lots of lights using MR Light Shader
If you are using distributed bucket rendering I’ve noticed that sometimes buckets take ages to render that should take no time at all… What happens if you render it locally (just rendering on one computer)?
we are using the mr_prime script to set up mental ray, i also turned off rebuild for the photon map and the FG. Then I turned on depth of field with a 200unit distance.
after changing the settings i changed our renders started getting wierd squares with random shadows in them(see attatched). Can anyone help us out with this?
that looks like you have your FG set to “Preview (no precalculations)”
Thanks for, that thats exactly what it was.
after solving that we are getting renders we can work with, however our shadows seem to be shaking violently has anyone seen this before, and have an idea how to fix it?
I have fg set at 250 with a radius of 350 and GI set at 1000 photons. We are using backburner to net render the images then compiling them with ram player in max.
here is a link to my video on youtube, you can really see the shadows shaking in the street.
o0n second thought you can only kind of tell the shadows in the street are shaking because of youtubes quality
If you render with final gathering and/or gi, once you’ve built your calculation, make sure it’s saved to file and turn off rebuild and use file is on. That’s always taken alot of the jitters out for me.
I’m working on a scene rendering with Mental Ray and using the Fast SSS, but it doesn’t seem to work when rendering over the network with Backburner. Scene renders fine locally. Other MR shaders render fine on the network. Anyone know why the SSS shader is giving me difficulty?
I don’t have any more details right this moment, but if anything else turns up, or I get the error message again, I’ll post it here.
EDIT: Here’s the exact error message I get when trying to render using Backburner:
“An unexpected exception has occurred in the network renderer and it is terminating.”
Keep in mind that rendering in MR with other MR specific shaders has never been a problem before, only with the SSS Fast shader. Backburner behaves fine otherwise, even on the same scene with the SSS shaders removed. Anyone encountered this or even know what it is?
I am using Mental ray distributed bucket rendering. Currently I have (myComputer) and (2Computer) added to the list and the render works fine using both the computers.
The problem occurs when i try an add (3Computer) it works but (2Computer) stops working. Is it something to do with the ports, cause they all use the default port which was installed during the installation. Is there any way which by which I can change the port.
Also I am able to ping to all computers (2Computer) and (3Computer) from myComputer and the windows firewall is turned off.
I have a setup with 4 machines
(XP) all using the same port.
How are you sure that machine2 stops working? Do you get a warning or error in the log, or…?
If anybody has encountered this error and knows a solution, pls post…i’m using core2duo E6300, 2 gigs RAM, ASUS 975X chipset MOBO on 3ds max 9 and the scene contains characters with hair n fur mod on them.
[b]Mental Ray Fatal Error Message
mental ray has encountered a fatal error and the application will now shut down.The error is:
DB 0.3 fatal 041008: attempt to access null DB tag[/b]
I cant post te file as its almost 40 MB…pls help!
I have no idea what a null DB tag is but have you tried turning off things and/or rendering objects separately? It might help narrowing down what’s causing the crash… For example you could use the mental ray material override and see if it still renders if you use a standard grey material.
it was material editor…there were some materials inthe editor which needed a cleanup…so this error has gone now but a new issue has come up…mental ray crashes on a 720*576 render whereas it renders fine at lower resolutions!!!
starnge but true…any ideas why is this so??
I’m using a E 6300 core 2 duo 1.86 and E 6400 core 2 duo 2.13 both with 2 GB DDr 2 533/667 RAM respectively… Nvidia 6800 and 7200 Graphic carrds respectively…mobo’s are P5WDG2 WS 975X chipset…
Render Elements are not rendering correct … ???
mental ray Sun Sky System
logarithmic exposer control: active, process background, exterior daylight active
all elements render incorrect
just put a sphere (solid glass) on a ground and try it yourself…
The problem seems to be the exterior daylight setup being active
If you diselect exterior daylight and pull down the values for mrSky and mrSun to about 0,01 and put a bitmap (even with ctrl_color_correction) in the environment then it works
but i guess that´s what we call a bug ???
It’s not a bug, you are not using the Logarithmic exposure control properly. Easiest solution is probably Disable Exterior Daylight and set the Brightness to 0. Other option is set the physical scale to 80000 and lower the brightness (about 56~57 has given me good results) till you are happy with the results.
I have seen some talk about the distributed bucket rendering issue, but no answers yet about how to get consistent Final Gather results from multiple machines working together on individual frames in a sequence. Has anyone been able to get a frame sequence to render consistently on multiple machines simultaneously?