OCTANE 2.X Needed Features

[00Ghz]

It would be nice for the devs to say which of those features we are requesting have a chance or will get into 2.X

[Goldorak]

It would be nice for the devs to say which of those features we are requesting have a chance or will get into 2.X

I think many of the suggestions here are sensible. We'll see what we can fit into 2.x after render passes, speed optimizations, cloud and ooc textures are competed, which are the key milestones on the 2.x roadmap.

[00Ghz]

Ok. Not a clear answer but I guess too early for something more precise. How far are render passes and such anyway?

And those texture features, would allow the integrated software procedurals to work with Octane right or is for something else?. Not clear here.

[Goldorak]

ooc textures == out of core textures. In other words textures could be stored in CPU memory. We have this working in Brigade with minimal speed loss. It remains to be seen if we can get Octane to do the same. CPU textures opens a lot of possibilities beyond just increasing texture count.

We're expecting to wrap up 2.x before the end of the year. Render passes will likely be coming out first. They are working already in internal test builds.

[PhilW]

Aahh - Out Of Core! I had been thinking "Out Of Character", "Out of Context", "Object Oriented Crap"...at least now we know!

[Seekerfinder]

ooc textures == out of core textures. In other words textures could be stored in CPU memory. We have this working in Brigade with minimal speed loss. It remains to be seen if we can get Octane to do the same. CPU textures opens a lot of possibilities beyond just increasing texture count.

We're expecting to wrap up 2.x before the end of the year. Render passes will likely be coming out first. They are working already in internal test builds.

Exciting stuff, Goldorak. But I really hope we don't get more slow-downs in Octane with the new features. Or at least either give us the option of OOC textures or have a dynamic solution where textures get transferred to CPU only once they reach a certain limit. That's probably wishful thinking since I know these things are not that simple but I hope we can get some speed increases within the 2.x builds.

Seeker

[00Ghz]

LOL.

Anyway could you tell what kind of benefits would be provided by ooc tex besides number increase?

[Goldorak]

LOL.

Anyway could you tell what kind of benefits would be provided by ooc tex besides number increase?

Here is one example. You could generate huge procedurals in CPU memory, from a new type of plug-in node (similar to Thomas' lua wood script), and Octane could theoretically access that texture directly without having to transfer or fit it on the GPU. in Brigade, we've tested a 64kx64k CPU texture this way, and were still running at 30-60 hz.

That being said, Octane is way, way more complex than Brigade, and it may not work at the same speed as it does in Brigade. Like the faster light traversal we wanted to add in 1.5, which didn't pan out as we hoped (we're still working on it), there is always a risk in sharing R&D features we haven't fully validated.

Regarding speed vs. complexity, we have looked at the idea of turning off some advanced features to get more speed when they aren't needed. It may be possible to get there automatically and invisibly to the artist, but if that doesn't pan out, then a global setting that enables this may be the way to go.

This issue may also be solved at some point in the future through new features in CUDA.

Looking down the line, cloud rendering is going to change a lot of the way people look at rendering time. The cost of an Octane render job in the cloud, even at $1/GPU hr, means that you can render almost any frame in real time (e.g.300m - 1 billion spp/s) for just a few bucks , and output the result in 8K without sweating it. I have found an 8K image with the same sample count as a 2K image looks better. The noise is spread over a much larger area and that makes even fireflies easier to remove.

[Seekerfinder]

Goldorak,
Is Brigade still on the 2014 roadmap? I miss Sam on the forums and presume he's not working on Brigade right now. Is development on track?

Tnx,
Seeker

[Seekerfinder]

Regarding speed vs. complexity, we have looked at the idea of turning off some advanced features to get more speed when they aren't needed. It may be possible to get there automatically and invisibly to the artist, but if that doesn't pan out, then a global setting that enables this may be the way to go.

Glad you guys are considering this.

Best,
Seeker