NVIDIA working on hardware acceleration for Mental Ray and ray tracing

[vlado]

@vlado - do you think at some point you will be able to leverage the GPUs found in cards to handle some of the extra loads? How much of an 'acceleration boost' could these people possibly be talking here? As it seems, my graphics card seems to move slowly in standard viewports in complex scenes. But perhaps that's simply an ADesk limitation.

We have looked indepth into current GPU technology before focusing on a pure CPU implementation. Our conclusion was that GPUs are currently not advanced enough for the complexity of the calculations we wanted. They may be powerful enough at some point in the future, but we don't have the time to wait. In contrast, CPU power is readily available and distributed rendering allows you to expand that power very easily.

Best regards,
Vlado

[jujubee]

I suppose with that article then, Nvidia really wants to push/market their Tesla systems. Currently, I think their Tesla systems are mostly useless except for "high-end" computational purposes. Even that I can't speak to as I wouldn't know. I just know they're way overpriced for the average prosumer.

[Neilg]

P.S. BTW - dongle policy still holds warez-mokeys outside

I'm not sure what you mean by this, piracy is no different to what it was before.

Youre actually getting more groups having a go now, because of the challenge and potentially being the first one to release a new working version with better anti-copy protection, haha.

[^Lele^]

@lele - perhaps this just American english - trying to emphasize a point. Then again, perhaps the way I speak American english is simply retarded.

Eheh, neither, Juju.
I was just deliberately twisting the sentence

[^Lele^]

I suppose with that article then, Nvidia really wants to push/market their Tesla systems. Currently, I think their Tesla systems are mostly useless except for "high-end" computational purposes. Even that I can't speak to as I wouldn't know. I just know they're way overpriced for the average prosumer.

http://www.nvidia.com/object/tesla_testimonials.html

you're very right there, they're apparently not aimed at the "prosumer" market.
Still, I think those things (expecially the rack-mount ones) can become VERY handy in the right situations (think Gelato, for instance, which is RMan on gpu).

[instinct]

Or think something like Showcase and alike...multimonitor arrays for realtime Powerwall viewing :P

[jujubee]

Well assuming their project could be off-loaded to the GPU (or distributed GPU in a Tesla situation), that's one less thing a CPU would really need to concern itself with.

GPUs however have enough of a problem handling current Max scenes and an added RT solution on a single GPU could hurt performance. I suppose if they implement SLI within Max, that may be a different situation entirely.

So I suppose two things you will see happen with Nvidia and MR (or the grandpappy):

1) Teslas are marketed towards mainstream professionals. Costs will eventually be lowered or they will keep price astronimcal because the demand is there.

2) Max finally takes advantage of SLI configurations.

[^Lele^]

GPUs and pure raytracing are not very happy a couple atm.
Try Gelato on the most powerful GPU of today, and make it raytrace, and it will be EXTREMELY slow, even compared to a middle-weigth cpu.
But try and do other things with it, like motion blur or DoF, and it is a zillion times faster than the fastest of CPUs doing it through raytracing techniques.
For instance, try to do a DoF preview, or Moblur preview from a max camera.
That's the kind of speed GPUs at the moment can offer, if very selectively.

[jreynolds]

GPUs and pure raytracing are not very happy a couple atm.

It's very painful to do with Gelato allot of things that VRay eats for breakfast, lunch and dinner...

On the other hand it's allot of fun (read technically challenging by me bosses ) to find ways around Gelatos' limitations. In some cases this has to come about with help from coders after an artist describes what is needed. In many other situations it just requires a different thought process that seems to be borrowed form past generations of VFX minds. I have to admit I really enjoy applying some of these techniques back to work I do with VRay, which is always a pleasure to use.

The beauty of Gelato can be appreciated at multiple stages of our pipeline. Our RnD department makes extensive use of it's flexible shader architecture and ease of connectivity. LnR department takes advantage of its render speed and quality with features like fast flexible displacement, fast full range Moblur, fast flicker free SSS and inexpensive image output spatial quality.

Best regards,
--Jon

[stevesideas]

Amazing stuff. This reminds me of the move from a standard slr camera to digital slr. Now we can see instantly what the image result will be. Processes will be faster and alterations quicker. Great stuff !!!

How long do we have to wait ????

[cecofuli]

Cebas seem to be on the case with this as well

LINK

[Morbid Angel]

Cebas seem to be on the case with this as well

LINK

Its not a very good demo at all. It seems slow, and with only 3 teapots? well I think from them we can see a version say 5 years from now?

[cecofuli]

Ah, yes, I agree. 3 teapot and very slow. But this movie show that also Cebas works in this direction. It's not a progressive rendering, like fprime or new Vray engine, but .. works

[stevesideas]

Not really very impressive. I guess its early days. Looks like the guy had an issue with the move tool. heh

[Thomas_Pardun]

Not really very impressive. I guess its early days. Looks like the guy had an issue with the move tool. heh

Morning Guys,

about the fR movie, it wasn't for public eyes. Nobody here know about the background of this move.
i does not need things like "Vray vs. fR" so, i set the video offline now. Have fun

sorry for my poor englisch.

Thomas