VrayRT for CUDA

[timdog]

Just a thought...and some stupid questions...

What happens when you use displacement through a gpu renderer?
Does the memory available on the videocard go through the roof? (GTX 480 - 1,5 Gb)

How come theres none that makes "gpu motherboards" so that you can fill it up with memory and stuff yourself?
Is the architecture of the GPUs preventing that ?

[vlado]

Same as what happens when you use it in V-Ray RT currently (i.e. it does not render). If you want displacement, you will have to generate it as actual geometry which of course will increase memory usage.

About the second question, I suppose the reason is that the video RAM is very tightly connected to the GPU for maximum performance, but I don't know enough about these things.

Best regards,
Vlado

[Dariusz Makowski (Dadal)]

HEy

I was wondering if you guys got a chance to test out the new cg from NV(fermi), to see which is better ATI or NV ? Any chance and suggestions ? :s

[vlado]

Haven't tested the ATI cards yet as the ATI OpenCL implementation does not support OpenCL image objects yet (we need them for the textures), but you can see some Fermi results here: http://www.spot3d.com/vray/images/stuff/rt_gpu/

Best regards,
Vlado

[cecofuli]

OMG! That's... fantastic. 26x the CPU version ahaha... And some question (obviously eheh...)

1. I can install two or four VGA in four different PC and run in DR mode? Like VRay-Rt CPU version? Because I don't want to install 4 Fermi in one PC. This VGA is too hot. So I install in different rendernode.
2. Can I install 2 VGA in the same PC and run in DR mode?
4. VRay-RT GPU version support dual GPU card? Like Nvidia 295?
5. Any possibility to use a VRaydisplacement in the future?
6. How many polygons we can render, for example, with 1 Gb VRam with a standard texture resolution (1-2K).
7. There are something like batch render? So I can submit different images during the night? Right now I can't with VRay-RT
8. And, obviously, VRay-RT don't support any external plug-in, like bercon maps, VRayScatter etc... . In the future?


Thanks again

[Dariusz Makowski (Dadal)]

Thats pretty impresive... how about all that extra NV features? All that ''science'' things we were tould about Fermi, double interpre something etc etc, does it benefit in any way or is it to early to tell ?

Also does VrayRT GPU utilize both GPU and CPU or only one? Just curious...


From other hand any render at 2k path from what I've noticed using Vray RT its quite fast and enough to work. Achieving that 2k with 480 its very good for work Having Tesla + that 480 will produce quite amazing renderfarm

[devin]

Those Tesla cards are very expensive compared to the 480's, the biggest difference being the amount of ram on the cards but they also have more processors. If DD works the same way with them that it does with Vray RT then why would anyone spend 7K on a Tesla when they could just buy 3 or 4 480's and get the same performance for around 2K? I guess the memory difference is the obvious answer to that question but just how important will it be?

[Paul Oblomov]

I want i!!!!111
Vlado - I remember you had gtx285 back in a day. How's it compares to new gtx480 ?

[bardo]

One question:

If I have a videocard that isn't supported by CUDA (Quadro FX4500) it means that also it doesn't support OPENCL??

In the official OpenCL website I can't find a supported hardware list.

thanks in advance

[vlado]

For nVidia cards, OpenCL is implemented on top of CUDA. So if your card does not support CUDA, then it cannot run OpenCL either.

Best regards,
Vlado

[vlado]

There is some more info and a video here:

http://www.chaosgroup.com/forums/vbu...RT-GPU-version

Best regards,
Vlado

[Dariusz Makowski (Dadal)]

There is some more info and a video here:

http://www.chaosgroup.com/forums/vbu...RT-GPU-version

Best regards,
Vlado

THanx very much Vlado, ure amazing person !!

Btw, any hint on pricing 2.0 vray ?

[cecofuli]

I think is too soon to ask that

[Intuition]

One MILLION dollars!

[3LP]

What about this Mother Board?
http://www.google.be/search?hl=en&so...=&oq=&gs_rfai=

7 PCIe x16

I know it would be hot in there, but it would be terrific, isn't it ?
I can't imagine to have this configuration [ with 7 GTX480 (or 470 best quality/price I heard) ] in DR.

But I had a question, is the scalability linear, or near linear with VrayRT and those GPU?


Stan
3lp Team