It’s too early to talk about this; I myself do not have a clue yet 
Things will be more clear when we get to user-friendly version.
Best regards,
Vlado
RT for GPU use only Graphic card memory?
Currently, yes - everything is uploaded on the GPU.
Best regards,
Vlado
OK. thanks.
Will you be recompiling it to work with OpenCL? CUDA sounds awfully proprietary to me, which is never a good thing
CUDA → OpenCL is not a mere recompile. Rather a rewrite. But still it was announced to be ported if i recall right.
Kind Regards,
Thorsten
+1 for OpenCL
Its the future imo.
Very good work regardless guys! Could this be used in production renders also or is it only in the working environment? Its my understanding RT isnt really for production but just to speed up feedback in the view port?
You can actually build a system with four 295gtx cards yourself (which gives you 8 GPUs).
Best regards,
Vlado
BTW recently looked at iray… It’s painfully slow with refractions on 4 quadro cards. In demo, called “car” there were some spotlights with direct shadows inside the interior (scene were lit with only one hdri). So the only soft shadows were under the car…
Other demos - well looked like a hardware viewport preview. Very strange materials.
Also there’s no support for importance sampling. Hehe.
No scenes which can demostrate caustics.
P.S. Support only mia material. You can’t write custom shaders. Will be released later in january. Still no connectors to max/maya/xsi
hi Vlado,
In chaosgroup, so you use a 4 GPU 295gtx PC with no problem for the VRay-RT. And it work well?
Paul, can you hare the link to the “car” i-ray video ?
No idea, still waiting for the motherboard to arrive…
Best regards,
Vlado
Hi Vlado,
As an option, we offer 4X GTX 295 systems. They are production systems that work right out of the box, CUDA included. http://www.renderstream.com/HPC.html
We are really interested in GPU render acceleration and we’re willing to offer ChaosGroup any of our 8 GPU systems at cost to help you develop this technology. If you wish to build your own we might also be able to source the neccesary parts and provide them at cost as well.
Let me know if you would be interested.
best,
Joe Pizzini
RenderStream
joe@renderstream.com
Vlado will there be a limit to the number of GPU’s or video cards we can network together or is it unlimited? Is the first release still going to happen this quarter?
V-Ray itself has no limit on the number of GPUs that can be used, but drivers or operating systems may have restrictions. As for the release, I cannot comment on this just yet.
Best regards,
Vlado
waow… RTT / lightmaps baking will soon become completely obsolete for realtime viz… Thanks Chaosgroup !
I want it really badly, will there be a demo?! This would change my workflow and results instantly.
+1 for Vray RT GPU
There is no need to wait for the GPU version to get the workflow of V-Ray RT; the CPU version supports DR so it can be made quite fast as well.
Best regards,
Vlado
yes, for DR I can confirm that I am using Vray RT with 9 render nodes (each quad CPU) and the feedback is really impressive. It’s great to place/tweak lights and materials, basically any tweaks that need doing. the main problem for me right now is the lack of vray proxy and xref support. I wouldn’t wait for the GPU version to come out before buying it… it’s a young product but for it’s price it already has enormous value to me.
(The memory limitation for gpu version seems a big hurdle too…)
In fact I noticed a pretty massive speed up by just adding a single 8-CPU DR slave to my existing 4-CPU VRay-RT setup. It may not be real-time in the way that the GPU-accelerated version is but it’s still damn fast.
Vlado, have you recompiled VRay-RT for OpenCL? If so is there any difference between CUDA and OpenCL?