Network rendering???

[gmask]

You can in fact network render with Mental Ray but you need to have maya and the plugin installed on all the machiens in question. I thinkt here are some notes on thsi int he docs that come with the new version of the mental ray plugin.

However this will only aloow you to render with Maya and not softimage or any other programs that access MR as the file formats are different but this should not be an issue for you.

[flox]

Great news.Thanks Gmask. I'll check deeper into it now. What do you think of "Smedge"? Someone mentioned this awhile ago? Thanks again for the help,it's appreaciated.

[gmask]

I haven't used Smedge 2 intensively but it does have a free tryout that you can test on two computers

[audun]

Ive been sing Smedge with Maya on 8 machines for over a year now. Works quite nice, it also supports Lightwave, Soft, XSI, Shake, Rayz & Air. Although I have not tried these myself. You can use it to run any comand line batch operation over a network. eg imgcvt. This way I was able to render Inkworks toon renders over the network.

Easy to use, pretty stable and a cheap price... (I'dont get paid to say this, maybe I should? )
http://www.uberware.net/smedge2/index.shtml
Audun

[audun]

My docs states the oposite:
Parallel network rendering allows you to distribute the rendering of each frame in your scene across multiple machines on your network. This requires standalone mental ray rendering licenses and can be invoked while working within Maya (in the Render View) and also from the command line with Maya or the mental ray 3.1 standalone renderer.

The mayarender_with_mr.cmd doesnt have options for setting start & end frame. So in order to create packets/distribution I think you would have to use some pre render mel script. Dont have much knowlegde in MEL, is there a mel guru out there to help us out? I guess one not would be able to use more than one machine per frame using for method.

For me at least the parallell rendering of each would be a very nice feature
Audun

[gmask]

If you read a bit furtther you will see what I stated before..

"If you would like to render from within Maya or batch render with Maya, you will need to install mental ray for Maya on all client machines and mental ray 3.1 standalone on all server machines."

The confusion here is I think that if you wish to have the parallel network rendering on one image that you will need the standalone. However for batch rendering you only need to have maya and mr plugin installed or on computers that do not have maya installed you would need the standalone.

Here's a link to another thread discussing this issue..

http://www.highend3d.com/boards/showthreaded.php?Cat=1,2&Board=MayaMR&Number=126696&page=&view=&sb=&o=[/http://www.highend3d.com/boards/showthread...&sb=&o=

Here's a link to the AW MR Faq

http://www.aliaswavefront.com/en/Tmpl/Maya/html/index.jhtml?page=/en/Community/Support/faqs/mental_ray_faq.htmlhttp://www.aliaswavefront.com/en/Tmpl/Maya...al\_ray\_faq.html

If you had several Maya seats with the plugin installed then you can split your render sequences between those machines.. I see no reason why not but unlike the standalone they will nto work on a single image and you will have to pay for a license on each machine wether it is the standalone or a seat of Maya.

I don't know if for example Smedge could issue batch renders to all the machines on your network that have the MR plugin installed or not but I don't see why not.

I'll look into the mel script unless somebody else has allready written it.. should not be too hard..

http://www.highend3d.com/boards/showthreaded.php?Cat=1,2&Board=MayaMR&Number=125291&page=&view=&sb=&o=[/http://www.highend3d.com/boards/showthread...&sb=&o=

I wonder how much MR standalone's cost?

[gmask]

still looking.....

[audun]

Hello mask
Im trying exaclty what you're suggesting. However all maya does when I add the -s -e -b flags, is to print the list of
Flags:
-v prints the product version and cut number
-batch for batch mode
-prompt for interactive non-gui mode
-nosplash skips the application splash screen
-proj [dir] look for files in the specified project dir
-command [mel command] runs the specified command on startup
-file [file] opens the specified file
-script [file] sources the specified file on startup
-log [file] copies stdout and stderr messages to the specified file
(use complete file name)
-recover recover the last journal file
-render [file] renders the specified file
(use 'Render -help' for more options)
-optimizeRender [file] [outfile]
optimize maya file efficient for rendering
purposes, and put result in outfile
(use 'maya -optimizeRender -help' for more options)
-archive [file] displays a list of files required to archive
the specified scene.
-help prints this message

Not shure what's wrong. What I did notice was that these flags work when using the -render flag. I assume this is because it invokes the maya Render.exe and the flags are sent over to that one. Could be I need to set some env variables or adjust some rayhost file?

[gmask]

Ooops.. I really should have tested that first..

[audun]

Had me fooled there. I'll need to look in to those mel commands. Something around these lines:

Pre render mel
int $start = getAttr mentalrayGlobals.startFrame ;
int $end = getAttr mentalrayGlobals.endFrame ;
int $pack = 5;

post render mel
$start = $start + $pack;
$end = $end + $pack;
//save file

Any progamming minds out there?

[gmask]

I could write this but not right now....

[audun]

That would be nice feature to have
Had a look around for pricing of standalone
AW sells them for "as low as $1,000 per CPU (quantity purchase required). "

in October 2001 it was:
mental ray for Maya node locked (dual CPU) - $3995 per license
mental ray for Maya floating (dual CPU) - $4795 per license
mental ray Stand Alone floating (1-7 CPUs) - $1950 per CPU
mental ray Stand Alone floating (8 - 47 CPUs) - $1250 per CPU
mental ray Stand Alone floating (over 47 CPUs) - $1000 per CPU

Not shure if this applies anymore.
Also I think there are other distributors of MR stanadalone offering better deals...
Audun

[gmask]

I remeber seeing somewhere more recently that they are selling them for $1000 per cpu as a base.. although I suppose if you bought more that you would get a discount. However in regard sto XSI.. I seem to remeber being told that rendering licenses were $500 per CPU (I could think of at least one person who will correct me on this if I'm wrong).. either way there is a considerable cost to setting up a renderfarm with MR.

[audun]

Listening to Jeremy Birn http://www.3drender.com during a AW Maya render seminar at Siggraph this summer I remember he said something along these lines: Smaller shops (in broadcast) tend to go for solutions with built in radiosity (ie MR, LW etc) whereas large ones tend to use manual solutions matching lights, reflections & color blead (mostly in PRman or propietary renders). Now theres nothing wrong in doing this, youll get a more artistic approach to it, youll have more control of your lights and it would most likely render faster. However without that much manpower, having acess to a render network for global ilumination would be a great asset.
Now if Id only could convince my boss

Id really like to hear from other smaller shops; if they've done the switch, what they're experincing, if expectations were met, things to watch out for, etc.
I guess the most obvious thing Id need to learn is shader coding.....

Audun

[gmask]

You can fake all this with Maya's renderer.. it is a little mor ework ..check out www.pixho.com

[audun]

Hi G
Ive been using Emmanuels scripts for quite some time, and theyre great, Especially GI Joe for diffuse shadow effects. However for colorbleed radiosity style renders I found the faking methods to be somewhat quite timeconsuming. I normally do that with area lights myself. Also I think the maya render has other limitations: motion blur in shadows /reflections, general the motion blur has issues. I know Emmanuel's got fakes for this too, but usally end up doing that in the comp. I guess these issues is why AW are doing the MR for Maya plug in the first place..?
Now I havent tried out his new HDR specularity fake, thats going to happen soon
Id like to thank Emmanuel for sharing his scripts they've been very valuable tools for me.
Audun

[gmask]

Hey well if you need motion blurred shadows and motion blur in reflections and all of these other features all of the time then MR is well worth the money. However some of my clients don't pay enough to get motion blurred shadows and have no idea what global illumination is or radiosity. If you can only afford run MR on a few proccessors I'd say you are in big trouble.. for me it would be like going back to when I only had two Indigo 2's to render with Of course back then I think it cost like $14 grand to buy a render license yikes! I could get MR lics for all my systems for that much.. times have changed.

[audun]

Oh yes those purple sgi boxes. Started out myself on one of those Everything was expensive, had to be bought in moduls, I'm glad things have changed too.

Still I think this price drop for maya actually could leave us with some fundings for extra render software/hardware ?

Ok so maybe motion blurred reflections arent the most obvious, but still caustics and final gather dont seem that slow?
Also MR handle scenes a lot faster than the maya render, did some quick tests:
I ve got a car model (of 3d02.com) the base mesh is 17342 faces. Only phong shaders, 2 directional lights (one with RT shadows)
Maya render globals: render res PAL, Raytraycing: reflections=2, refraction=0 shadows=2
Anti aliasing Prod Quality, changed contrast for 0.100 for RGB

Mental Globals:
Prod Quality, change Max reflection=2 Max Refraction=0 Max Raydepth=20

1st test: No smoothing
Total Elapsed Time For Maya (hh:mm:ss): 00:01:09
http://home.no.net/audunsa1/test/maya1.txt

MR MSG 0.0 info : wallclock 0:00:19.74 total
http://home.no.net/audunsa1/test/mr1.txt

2nd test: Smoothing set to 2
Total Elapsed Time For Maya (hh:mm:ss): 00:01:53
http://home.no.net/audunsa1/test/maya2.txt

MR MSG 0.0 info : wallclock 0:01:02.11 total
http://home.no.net/audunsa1/test/mr2.txt

3rd test: Smoothing set to 2, final gather on
MSG 0.0 info : wallclock 0:05:21.57 total
http://home.no.net/audunsa1/test/mr3fg.txt

Images here (in tif original renders)
Maya: http://home.no.net/audunsa1/test/maya\_img/
MR: http://home.no.net/audunsa1/test/mr\_img/
Ok so the images arent great here but it does illstrate the speed of MR ?

Audun

[gmask]

It is well known that MR has a faster raytracer... the test I want to do is to simulate the GI look and compare times there.

The slight gain in speed shown here is still not compelling enough for me to spend thousands of dollars.

Also with some of my tests I have done are using nurbs models.. MR takes a long time to process them before rendering.. too long..painfully long.

I owuld really like the sub-d's to be compatible and for there to be shared memory space before I take the financial plunge.

[audun]

Also with some of my tests I have done are using nurbs models.. MR takes a long time to process them before rendering.. too long..painfully long.

Thanks G
I generally was looking for info like this. What to look out for, what to work around and where to steer away. You have me convinced, would it be right to say that you think MR integration with Maya is not production ready....

Also Pre/Post Render Mels are out of the question, not supported. I'll split up scenes manually as you suggested and run them through Smedge.

I'll have a look at GI / fake GI when my scenes are ready & post result times, Thanks again.

Audun