Page 1 of 6 123 ... LastLast
Results 1 to 10 of 60

Thread: GPU accelerated GI in 3ds max 2017

  1. #1
    Join Date
    Dec 2004
    Location
    Marina Del Rey, California
    Posts
    4,143

    Default GPU accelerated GI in 3ds max 2017

    We are very interested in this thread to hear your experiences using the new GPU-accelerated Indirect Diffuse (GI) in 3ds max 2017.

    See it in the Global illumination tab of the Render Setup menu. Near the top.
    Barton Gawboy

  2. #2
    Join Date
    Mar 2013
    Posts
    16

    Default

    My opinion is probably the last of importance as my computer is Very ordinary in terms of hardware and I don't render massive scenes but anyway here is my two cents. I've just tested it a couple of times and it seems producing better light distribution than traditional FG solution, maybe not as better as Iray's one but probably close. But I must say now I realize how fast is FG ) after compared it with GI.
    Especially as GI doesn't seem support classical sampling only unified and I still use it quite often for rendering surfaces that have to look very smooth.
    Last edited by ysrostov; May 6th, 2016 at 14:03.

  3. #3
    Join Date
    Mar 2015
    Posts
    11

    Default

    Can GPU acceleration be used with Mental Ray in production render mode? Thanks!

  4. #4
    Join Date
    Jan 2009
    Location
    Berlin/Germany
    Posts
    1,148

    Default

    Quote Originally Posted by murzilka View Post
    Can GPU acceleration be used with Mental Ray in production render mode? Thanks!
    Sure. The GPU acceleration is enabled by default. If you have no capable GPU then mental ray will auto-fallback to CPU mode with identical results but longer render times.

  5. #5
    Join Date
    Mar 2015
    Posts
    11

    Default

    At home PC with i7 4771K 32GB it takes x3 times more time with the Indirect illumination ON - GPU checkbox ON, than the CPU. The graphic is GTX 660Ti2 2GB - average graphics, average processor.
    So Im very serious to buy 2x Xeon machine and forget about GPU things for now... Nobody from my known people uses GPU at all, and nobody uses Iray (maybe 85% Vray, 10% Mental Ray, and 5% something else). I myself really cannot use RT modes by having x300 million polys here and there... Just brutal calculations, no realtime.

  6. #6
    Join Date
    Dec 2004
    Location
    Marina Del Rey, California
    Posts
    4,143

    Default

    Please describe more about your scene. What do you mean by the x in front of x300 million polys?

    Do you see any messages in the Render Message window about it going back to CPU or anything like that?

    This algorithm is unique in that the textures do not get loaded into card memory. But you geometry does need to fit into it. So it is better than a typical GPU renderer.

    I think you should get more information and feedback before making to swift a decision about what kind of machine to match where this is headed.

    We want you to use the best tool for the job. And we believe we have widened potential for mental ray with this hybrid CPU + GPU acceleration approach, especially for the demanding scene requirements of production, and those with rendering flexibility needs.
    Barton Gawboy

  7. #7
    Join Date
    Mar 2015
    Posts
    11

    Default

    OK I will re-make and upload when at home PC. I do not saw any strange messages. The scene was about 2-3M polys, clean geometry with AO node on the diffuse slot of standart material.
    Im not sure what kind of things are accelerated by the GPU, on which material it works, etc. - and because of this I asked in the other thread, where this can be read. I do not need AO pass at all, (I read so much about this), I need to render scenes with tons of polygons, where AO is used as node, mixed with bitmaps and other nodes, to create masks for wearing for example.
    I also have a script from the ZAPs website, downloaded the last year, which says that enables GPU and I get the rendering about twice faster.

    ADDED: I will not stop use MR, just will not lose time to try tips and tricks with the GPU madness... The last weeks I read all GPU render engines are 20 times faster than the others...
    Last edited by murzilka; May 11th, 2016 at 06:11.

  8. #8
    Join Date
    Apr 2009
    Posts
    511

    Default

    Quote Originally Posted by murzilka View Post
    ...GTX 660Ti2 2GB - average graphics
    no its not.
    its low end gpu with just 2 gb of ram. you will have problems with that card in every gpu renderer. you should be save with 4gb in generell.
    Last edited by Kzin; May 11th, 2016 at 11:10.

  9. #9
    Join Date
    Dec 2004
    Location
    Moscow, Russia
    Posts
    682

    Default

    mental ray need ~761mb of GPU memory only for empty scene. So you may render very simple scene with 1GB. You just need to close all applications that may use GPU memory before run 3dmax.
    With 2gb you may render some scenes that not heavy enough.
    Pavel Ledin

  10. #10
    Join Date
    Mar 2015
    Posts
    11

    Default

    Quote Originally Posted by Kzin View Post
    no its not. its low end gpu with just 2 gb of ram.
    The problem is, which card will have better performance than double xeon. I know that 660 is not top, but if it has 1344 CUDA cores and they dont serve for nothing, in my logic some 12GB titan shall not, too. There are no info about this kind of performance, (I asked already - if somebody knows - let me read). In which scenario the CUDA enabled card will serve better, and where is useless? So I can adapt some parts of my pipeline to that posibilities, for example. I tryed to install some demos of Octane, Redshift, Arnold is nor available for MAX, Corona, Indigo not available for max 2016 and above etc. Im not a scientist to learn and compare all this complicated softwares, and in the bulgarian 3D forum, almost all of the artists use Vray and some Mental. As for myself, for now the brute force CPU calculation is a solution, while no detailed info is available.

    Of course, the double xeon will be supplied with a good graphics, just to be sure it can work with massive scenes. The low end 660Ti2 for example, now can model scenes up to 100 millions polys in Max, this is enormous compared to 10 years ago.

    Cheerz!



    P.S. Thanx Puppet, we wrote in the same time hahaha... So the GPU RAM is the most important parameter and not the number of CUDA cores? And multiple GPUs does are (I read not are so much) about to provide much better performance? I cannot buy 10 adapters to compare them between each other, and SLI, or compare them to CPUs in the same conditions.
    This is what I read http://videocardbenchmark.net/high_end_gpus.html and I probably would double this 980Ti because of its low price. But I still dont know where the GPU will be useful more and where less. I do not use ActiveShade, only clean rendering of massive scenes. In the other forum, nobody answered me if Vray can be used with GPU offline (as I see they use the RT module only, with GPU). Is is seen, Im a lamer in this field, but there is no much info to read...
    Last edited by murzilka; May 11th, 2016 at 14:35.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •