VrayRT for CUDA

[vlado]

Yep, it looks like it is nearly linear. About the PCIe slots, the spacing between them is important - all the recent GPUs take space for two slots.

Best regards,
Vlado

[andrewjohn81]

@ Paul

I'm running a system with iRay now. There is caustics, and I only have one quadro card and it's really quite fast. Not quite as fast as some of the demos I've seen for some other software ,but then my graphics card isn't that great. I'm sure if I had 4 GTX480's then my scene would look that fast too.

As far as shaders go, I haven't seen SSS yet, but iRay supports it. They haven't given any examples yet.

Noise. That's the bigger concern. If I want a noise free image at 5k x 5k, then I'll be waiting a whole day. But then again, it only took me 5 minutes to set up the render, so there are definitely uses for it. If I set up a vray scene in only 5 minutes of the same thing, I assume I'll be waiting for a whole day to render that as well.

They all have their uses. I wouldn't go bashing them just yet. I can't wait to try out a beta version of RT with CUDA. On my computer with dual Xeon 3.2 wHyperthreading (8 processes) it ends up rendering 4x faster in iRay when I use GPU than my CPU. It makes numeric sence since I have 192 cores @ 602Mhz, so that actually is 4x faster when you multiply cores by speed. I assume I will get the same benefit in RT.

[vlado]

One thing to keep in mind is that V-Ray RT is interactive inside 3ds Max and supports distributed rendering right away (even the GPU versions) - something that can't be really said about any of the other solutions (for 3ds Max).

Best regards,
Vlado

[themaxxer]

will it be compatible with nvidia tesla? I think so, but I am nut sure.

best regards
themaxxer

[vlado]

Yes, of course; it works fine with Tesla cards. Better to get a Fermi card though (2050 or 2070), the 1000 generation are not all that helpful.

Best regards,
Vlado

[davision]

Maybe this has been asked and answered -- but my light "research" of the forums has not come up with anything...

Will Vray RT and/or Vray RT GPU -- be capable of outputting animations? I know this was not RT's original purpose, but since the GPU is so damn quick, it would be nice if it could crank out frames.... knowing of course there will be certain limitations in terms of scene complexity, displacement and plugins.

[GatesB]

Geesh. I was lookin at a Tesla box today. A box with 4 x 1000 generation is $5k, a box with 4 x 2000 generation (Fermi) is $9k - [Both come with dual Xeon]

[scottsch]

I know CUDA/nvidia is popular but is there any word on Radeon 5800 series performance, since RT is done in OpenCL? Seems like Radeon cards can be had for less $ while getting more ram per card.

[Alan Iglesias]

You can use RT to output animation frames right now. You use a script that will let you render a frame by either time or noise threshhold - pretty dang cool! The script is in this thread:

http://www.chaosgroup.com/forums/vbu...9681-Animation

-Alan

[davision]

Thanks Alan -- I do remember an earlier version of that script popping up. However, that's not exactly what I'm getting at. I want to know if we can use RT (the GPU version most likely) in a manner similar to backburner rendering (on compatible scenes). It sounds like that particular script you pointed to (I haven't tried it) will commandeer max while it does it's work. I want something that will work in the background with other computers and not interfere with using max while it's working (like backburner).

[Alan Iglesias]

Ah, understood. Of course you could start the animation from another machine, leaving your workstation's IP out of the RT Server list, but failing that...

From what I understand, RT always uses distributed rendering even when working only on one machine and is always working through a DR render client and render servers, even if it's a local one (RT on one machine). The machine used to actually distribute the rendering information is the Render Client - So to make it work like Backburner you'd need to designate a Remote Render Client (if I'm following). I'd hope that wouldn't be too very hard to code but then again, I'm no programmer. Vlado, is this a fairly simple thing to put together? Or is it non-trivial? I'm thinking that it is probably beyond a script's capability, but I couldn't be sure of course...

-Alan

[vlado]

Thanks Alan -- I do remember an earlier version of that script popping up. However, that's not exactly what I'm getting at. I want to know if we can use RT (the GPU version most likely) in a manner similar to backburner rendering (on compatible scenes). It sounds like that particular script you pointed to (I haven't tried it) will commandeer max while it does it's work. I want something that will work in the background with other computers and not interfere with using max while it's working (like backburner).

The RT render server does not really interfere with 3ds Max on the render slaves in any way.

Best regards,
Vlado

[davision]

Right, Vlado -- I understand the slaves run RT in the background. But I'm talking about being able to freely use MAX on the main user's workstation (meaning the machine where the RT render is initially submitted). It would be great if the animation render could be "sent" to the 'farm' and sequential frames would be saved at a network location (and not necessarily be shown as a live view on the main workstation). The user would be free to continue working while the animation rendered on the farm machines.

As I mentioned, I know this is not the primary use of RT since it was originally designed to give near-instant preview renders BUT in the case of the GPU version since it's so fast and does such high quality, it would be nice to use it as a backburner replacement on suitable jobs.

Does that make sense?

[vlado]

I guess things would be more convenient when the RT can be used as a production renderer.

Best regards,
Vlado

[3LP]

and? may be for vray 2.0 ?
or a updated script or sp3 version of vrayrt before ?

vlado, don't leave us in the dark ;o)

Stan