VrayRT for CUDA

[saudade]

THE TITLE IS: POWER CONSUMPTION

mental images, bunkspeed, randomcontrol announced their GPU compatible rendering engines
I hope that CHAOS will do better, I want to be surprised...
I'm an happy VRAY+RT user but RT can't substitute VRAY because doesn't support proxy, (all my scenes now use proxies), and lacks a lot of VRAY features.
The question is:
we will see a third production rendering engine based on GPU+CPU?
will Vray be fused with RT?
will I substitute my 10 workstations with one Tesla or 3-4 slied Nvidia cards in one single workstation?
Today the power consumption is the most important problem to resolve because 3d artists give all their money to pay the electricity bill.

Best regards

Gabriele

[vlado]

The next update for V-Ray RT includes support for VRayProxy objects. It also includes a light-weight version of the light cache, which helps to clean the noise a lot faster. We do have a GPU version as well (see the demos from Siggraph 2009), although I cannot say yet when it will be commercially available.

Best regards,
Vlado

[walee]

The next update for V-Ray RT includes support for VRayProxy objects. It also includes a light-weight version of the light cache, which helps to clean the noise a lot faster.

i hope the next update will be available pretty soon

[cecofuli]

The next update should come out shortly after the official release of 3ds Max 2011.

[Dariusz Makowski (Dadal)]

Hey

I got 1 question tho. If VrayRTG works better on OpenCL then we dont really need nvidia cards right? Untill you get fully suported C++ for Cuda but thats duno when its gona happen.

[VI_MOVE]

Is there any Information about the new Geforce 480GTX (released yesterday) already concerning speed and scaling of the new chip architecture in RT-GPU compared to the 285/295GTX? I know it's probably a bit early so I'm just asking out of curiosity. Maybe Nvidia was generous and sent you a sample?

Can't wait to see next version of RT with it's optimizations. I'd rather have the better light calculation vlado pointed out before (lightweight lightcache) than getting the GPU part now. While the GPU version promises to give a great performance generally as it comes to complex scenes I guess we're getting into many hardware based limitations such as memory amount. In other words it could be a nice feature to play around with but not as important as long there's no workaraound over hardware restrictions. And that seems to require a pretty substantial redesign of the vray architecture itself as discussed on other threads.

Anyway pretty exciting to see the next big step sooner or later!

[Dariusz Makowski (Dadal)]

Heya

I sure doubt if Chaos Group got gtx 480 directly from Nvidia. After all Nvidia owns Mental ray don't they.

Off the records, this card is disaster from point of watts/noise/performance... Hope VrayRT wont need NV and ATI will do as good as NV cgi.

[VI_MOVE]

Oh you're right on that. Don't they also have their own render engine/language? Anyway if I was them I'd try to make as much money as possible and therefor I'd try to sell as many cards as possible. Obviously they are in trouble big time already in belong of having not much to offer (if theres just something about the rumors concerning TSMC) and also an outdated license for developing processor chipsets. But that's another story.

Well I agree that the 480 GTX seems not to be an optimal performer in what it should be - as a gamer card. Too expensive, way high power consumption and bad availability. But looking at it as a GPGPU the needs are a bit diffrent. And from this point of view the architecture could really have it's benefits (a big pack of shader units connected to an high speed cache for example). For myself I wouldn't mind the bad things if it was a big step performance wise. We'll see all I know so far is based on assumptions. Maybe someone's gonna go to change the board layout to 3GB memory. That would make things easier...

[timdog]

According to this article it seems to be quite prominent for cluster computing http://www.anandtech.com/video/showdoc.aspx?i=3783.
Have no idea though if it actually have any substantional gains for realtime pathtracing/ Vray RT gpu and the likes?
I read somewhere that Nvidia cut the double floating point "option" for the gamercards.
Does this have any negative effect for a vray GPU version?
Would be great if one could utilize it for rendering.
The biggest problem is really the crazy energy usage.

The GTX480 is sucking 480 W on load!!! (furmark), thats seriously affecting the electrical bill :s
Thats like 3 flatpanel 46" HDTV on at the same time :s

[vlado]

I hope to get a GTX 480 soon, so we'll see how that goes As for the double-precision numbers, we are not currently using them for the OpenCL version.

Best regards,
Vlado

[timdog]

I hope to get a GTX 480 soon, so we'll see how that goes As for the double-precision numbers, we are not currently using them for the OpenCL version.

Best regards,
Vlado

Oh, thats great , so we dont have to spend all our savings to get a Tesla card, (in case theres no other stuff thats missing)

[Dariusz Makowski (Dadal)]

I hope to get a GTX 480 soon, so we'll see how that goes As for the double-precision numbers, we are not currently using them for the OpenCL version.

Best regards,
Vlado

U dont currently... but that is possible? Its better to get the one with it or shall we ignore it for next 2 years?

[timdog]

I guess a renderer that could do without the Tesla for 1000s of $ would be far more attractive for the
average consumers.
How about ATIs cards?
Do they have any cut down features? Are they any good for GPU rendering?
They seem to be alot more appealing in terms of energy usage ,heat and noise levels.
I guess its still too early to tell....

[StateofArt]

We do have a GPU version as well (see the demos from Siggraph 2009), although I cannot say yet when it will be commercially available.

Hi Vlado,
don't you think that it should be better if the gpu version comes out as soon as possible? I'm a fan of Vray but seriously I'm thinking to move toward Arion by Random Control. The platform works in combination with cpu and gpu and it has great results. The realtime is not an illusion anymore but now we need to optimize all the process and to squeeze all the power of the hardware.
Hope you will consider that.

Best regards.

Gianpiero Monopoli
Director
State of Art Architectural Visualisation Studio
http://www.stateofartstudio.com

[vlado]

It will do you no good if we rush it out; it must be working smoothly so that it does not disrupt your workflow in 3ds Max. Speaking of which, V-Ray RT on the CPU already gives you such an interactive workflow, and it works very well. GPU only adds some speed to that.

Best regards,
Vlado