OTOY Forums

Oh I'm sure its a bit away (vray 2.0), but regardless of what they're happy/not happy with, you cant deny it's pretty damn fast, can you? like, video doesn't lie in this scenario.


cheers

cp

havensole wrote:Don't hold your breathe waiting for vray. Talking with someone behind the scenes of VrayRT at siggraph, they are still a ways away from release. They also said they really aren't happy with opencl, but have spent so much time investing in it that they're stuck with it now.

cpember wrote:70% of people can buy it for their grommet in its current state ($70k)
with a few changes, you can sell it to the rest or the population that uses a different grommet ($30k)

you're saying you're willing to toss 30k? just like that?

I'm pretty sure that's not what I said. I said that going from $70,000 to $100,000 does not equal a "doubling".

Beyond the fact that we are talking about a fair bit of re-engineering. Its not like you just say "Okay code, use openCL now, not CUDA" and it all just magically transfers over. At some points we are talking hardware level programming which is not as flexible as standard code writing. Its just like programming older, and even some newer, games for consoles. One has to be mindful of everything they are doing (memory management, resources, etc.). In modern computers all of this is handled by the OS, not at the hardware level.

Yes it is fast. My argument for it though is that they state they are using 3 gtx480's. We saw this same kind of performance with our riggs at Siggraph too, using anywhere from 4-6 gtx480's on one system and 2 gtx470's on another. They also aren't doing unbiased rendering which counts for a large speedup on their end. My last point is that the scene they show isn't that complex. a single area light on a 1 mil poly model. Even radiance's lambo scene is more complex than that. Add some GI and that render time goes up fast. One of the biggest things that I have to give them is the integration in Max and Maya and using their shaders and such. I am sure more of that will come down the line as other things get completed in Octane, so I am not that worried.

I'm guessing it's something like porting a game from Directx to Opengl.

okay I guess I'm not really making my point. What I'm trying to say (without putting words in people mouths) is that you would open your market to a far broader audience (by double, you go from *1* card vendor to at least *2* I was using 70/30 as the install base spread, are you really taking everything I say literally?)

I am also well aware of the difficulties in coding (we have a fairly vast R&D dept here, and we worked extensively with nvidia on gelato with Larry Gritz and the bunch) but again, we make ourselves fairly platform agnostic as possible for the simple point of market share.

I'm not trying to start a software war here, but you have to remember, while opencl has it's limitations, dont think for a second cuda doesnt, so why *limit* yourself?


cheers

cp

I think the team never said explicitly that there wont be OpenCL support in Octane, just not in the near future (which I would guess covers the lifespan of 1.0 version) - unless we want to have the final v1.0 in a year or so. Which we don't.

oh i know I can wait for a long time, because at the moment cuda and octane work exactly the way I want it, so I dont really care about opencl.. I was just hoping it wasnt completely off the map is all


cheers

cp

cpember wrote:What I'm trying to say (without putting words in people mouths) is that you would open your market to a far broader audience

And of course I agree with that.
I was merely pointing out that AMD's share of the market is quite a bit smaller than Nvidia's which may give software engineers less of an incentive to immediately go OpenCL when it's currently less mature than CUDA.

Also if your serious about using octane as a production renderer rather than just a bit of a novelty/hobby just using whatever card you might have in your machine, then its likely youll be buying in new cards or the most recent releases so just make it an nvidia one rather than AMD. But at the same time we want nvidia to be competatively priced and not get too cocky so openCL would be good at some point yes!