Jump to content

3d max/mental ray/trees


mskin
 Share

Recommended Posts

question... am i nuts to think i can populate a scene with 300 or so 3d trees from onyx?

 

i keep getting an "out of memory" error during the render.... mental ray and scanline. Is there a way to avoid this? i have 2gigs installed on my machine.

 

are there any memory handling tactics or tricks that i don't know about (i assume yes).

Link to comment
Share on other sites

so far, i think they are pretty cool. Although my point wasn't that i was using onyx trees, but that i am using 3d trees...

 

i have a scene populated with hundreds of 3d max trees... just for ha has to see how it would render. Not so well. It crashes pretty quickly.

 

I have the same scene with about 1/2 the number of trees - im just not done putting them in yet - and it was crashing too. i changed BSP to large BSP and the file stopped crashing during the render.

 

I tried the same trick on the file with max trees, but it still crashes.

Link to comment
Share on other sites

courtesy of Neil Blevins... these actually make an impact - Thanks Neil!

http://www.neilblevins.com/cg_education/cg_education.htm

 

btw - love that avatar!

 

Scott

 

1) Close all other applications. Seems obvious, but it's surprising how much RAM other small applications can take up. For example, iTunes seems to gobble up about 0.1 of a Gig of RAM, even though common sense seems to suggest it wouldn't take up too much memory.

 

2) Turn off the vfb and render to file. The "Virtual Frame Buffer" (vfb) which is now called the "Rendered Frame Window" takes up a lot of memory. So for the privilege of seeing a preview of your render, you're giving up potentially 200 to 300 meg of RAM, especially if you're rendering a 2k or 3k image. Go to the renderer options and uncheck the Rendered Frame Window Checkbox. If using a 3rd party renderer like Brazil, make sure you turn off it's frame buffer as well. Then set the renderer to save the result as a file in the Render Output section. This will save you tons of RAM.

 

3) Do a netrender. Even for still frames, even if you only have one computer, using backburner or a different netrenderer can save tons of RAM, because it doesn't need to fully load max to render your scene. So I usually set up my home computer as both the manager and the server, I go into max, I set off a netrender, making sure I have the "Initially Suspended" checkbox checked, then close max, then go into the Monitor and initialize the job (if you forget to submit the job uninitialized, your computer will try and load 2 copies of the scene, one in the netrender, and the copy of the scene that's already open in max, and that's a very bad thing). Check the manual for more info on using backburner, or whatever network renderer you choose.

 

4) Command Line Rendering. You can also use commandline rendering to render your file. Just look up Command Line Rendering in the help file.

 

5) Turn off bitmaps displayed in the viewport. This doesn't reduce memory for a netrender, but if you want to do an interactive render inside max, it's quite possible that a bunch of your memory is being eaten up displaying maps in your scene in the viewport (I was surprised the other day when I found out about half a GIG of RAM was being used up displaying a bunch of large maps in the viewport). Just go to the Views Menu and click on the "Deactivate All Maps" option.

Link to comment
Share on other sites

Well I'll respond to my previous post, now that Max is 64bit it won't crash, now that's the good news, I decided to give it a try, rendering 750 high poly trees, the memory is up around 6.8Gigs and its been about 30 minutes now and only 3 buckets have completed, so its gonna be a while, my settings are default, so if I raise them to something like good quality...then it will probably make Maxwell look fast...so the good news is no more crashes, the bad news is waiting forever, vray proxies seems to be the best option still...

Link to comment
Share on other sites

Yea, its one of the few reasons to use Vray.

 

Or use 2D trees for the distant trees, 2.5 (IE X trees) for mid and only a few 3D trees for those right in front of the camera

 

Onxy gives you an option to export low poly count trees, export out several resolution trees, very low for distant and more detailed as the trees get closer.

 

Regardless if you use Vray or not 300+ 3D trees are going to take for ever to render.

 

JHV

Link to comment
Share on other sites

what about proxies?

 

everyone talks about the vray proxy - i find it incredible to believe that mental ray has nothing like this.

 

multiple instancing? placeholders? something.

 

along the same lines... i am under the impression that instancing and referenceing will be of no help in terms of my render - is that true?

Link to comment
Share on other sites

Yea, its one of the few reasons to use Vray.

 

Regardless if you use Vray or not 300+ 3D trees are going to take for ever to render.

 

JHV

 

Ummm...it only took 1.34 minutes to render 750 high poly trees as proxies...with GI by the way...not exactly forever but I guess that's subjective, as I said before 5,000 trees doesn't take that much longer...

Link to comment
Share on other sites

Ummm...it only took 1.34 minutes to render 750 high poly trees as proxies...with GI by the way...not exactly forever but I guess that's subjective, as I said before 5,000 trees doesn't take that much longer...

 

 

well who the hell has 1.34 minutes to spare?!?!?!:D

Link to comment
Share on other sites

Yup, Vproxy give back the speed to computer.

 

...but:

 

  • Trees are identical (you can rotate, load several different)
  • Cant have interaction (wind, "face camera" for bitmaps: totally no-go for speedtree)
  • You dont know where is it really, until you see render(wireframe is simplified, and some mesh disappear in viewport) imagine fly trough animation with 50% time blocked view by leaves

Any thing else: Vproxy its a life saver for me.

 

BTW can you put animated mesh in Vproxy?

and have it animated as proxy?

Link to comment
Share on other sites

Right...that's what I'm talkin about, that's 1.34 minutes of my life I'll never get back...

 

Anyway I didn't mean to turn your post into a Vray vs MRay discussion, but in this area I don't think Vray has much competition, Mental ray has come along way in the user friendly area, but there are still a couple things it needs like this for example...

Link to comment
Share on other sites

"heavy instancing provides for rendering many copies of an object at different transformations without replicating the object data in memory" - from mental images.

 

i guess to answer my own question, i instanced and copied hundreds of trees. no doubt the instanced trees rendered faster (or i should say at all... the other method crashed)... but no where near 1.34 minutes. i set it of an hour ago and its 11.1% through final gather.

 

there has to be a better way with mental ray.

Link to comment
Share on other sites

You are going to wait for ever if you are FGing the trees. Do yourself a favour by hiding the trees during FG, freeze the FGM, unhide the trees and rerender. There is no real benifit in getting the trees to bounce light, especially one in the mid to backround. If there is a tree right in the foreground and its a major part of the scene, then possible include it in the FG.

 

Not to sound to harsh here but I think you need to pause and think through what you are trying to achieve and how you are going to get there. Its going to save you alot of pain later.

 

JHV

 

PS 1.34 minutes at what resolution? at what AA? what GI settings? Polycount? Lighting setup? .... there are alot of variables not taken into consideration. I am impressed that that many trees can be rendered, but not by the time, its meaningless.

Link to comment
Share on other sites

It was at 640 x 480

Adaptive subdivision for the AA

one omni light with vray shadows

Irradiance for primary and QMC for secondary

91,000,000 polygons, give or take

Everything pretty much default settings, if you have any specific requests for a test just let me know...

Link to comment
Share on other sites

Sorry I didn't mean it as a stab at your effort, it was more to highlight the fact that by just giving a time for something to render isn't giving a clear and true reflection solution to the problem.

 

All your example prooved to me was that it is possible to render out 100's of complex trees, Great. However by stating how long it took has clouded the issue.

 

1.43 at 640X480 is avarage, multiply that out to a decent print resolution (4000 longest side) and tell me how long.

 

This is one of my biggest bug bears with most tutorials, They sprout how quick it is, but fail to say they are working at a rediculously low resolutions.

 

JHV

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...