Arion: a Fryrender RT engine + GPU support. Coming soon

[cecofuli]

I know vlado, but see this image

http://img146.imageshack.us/img146/968/set2s.jpg

from my friend. 10 min with I7 965 @ 3.2 + 280gtx

10 min! It's a miracle. No noise in the glossy floor.
For catalogue Arion is a must! I think more company will buy Arion (or fry) instead VRay.
I know, for huge project, like complex exterior with 20+ million poly, 4k-6k size etc... we can't use Arion or GPU renderer, but in some work
I think that we can use it in production without 10+ CPU render node.

[Morbid Angel]

10 minutes eh...

[instinct]

10 minutes eh...

Quoted for agreement :P

[cecofuli]

Yes, 10-15 or 20 min it's very, very good rendertime IMO. No noise in all glossy shader and shadows. In VRay it's very hard obtain that in only 10 min and in this resolution.

[Morbid Angel]

and is it correct to assume its all brute force gi?

[instinct]

Yes, 10-15 or 20 min it's very, very good rendertime IMO. No noise in all glossy shader and shadows. In VRay it's very hard obtain that in only 10 min and in this resolution.

Erm...10 or 20 minutes is a pretty steep difference if you ask me...and that does make a big difference. Just read a little around...i won't dive into the buzz-wording hate again (The first hybrid unbiased renderer in the world...erm...yeah sure. :P)
But i gotta admit that 10 mins for the image you posted seems at least "doubtable" given what they show themselves as a 10 min render:

http://www.randomcontrol.com/images/.../classroom.jpg

That seems like a huge difference (in size, noise-levels, pretty much all around)...

[Morbid Angel]

only time will tell....think soon we will have fermi, and a year or two after a new card thats 4 times as powerful...realtime rendering is inevitable. I said this 10 years ago, when vray was a young alpha...there will be only 1 app, which will do 3d/2d comp all in one in realtime. And look where we are headed, integrated toxic, bridges between apps...realtime.
Too bad world is suppose to end in 2012.

[cecofuli]

Hi instinct,

yes, but a Fry render CPU take about 3-4 hour (+-) for the same noise free image. So, 10 or 20 it's good time. it's In conclusion Arion is pretty fast for simply interior or design catalog render
Obviously you can't render billion poly scene, like complex exterior, or character animation or in complex pipeline production. But for some works, I think it's a good product.

[Frances]

and is it correct to assume its all brute force gi?

There is no need to assume, since it is stated clearly on the Arion web site.

Unbiased rendering isn't everyone's cup of tea, but for those who want it, like it and use it, Arion is a great thing.

[voltron7]

All these newer technologies makes me think about having to buy newer equipment - and it makes me cry...

Lol, same here...

[tammo]

I always wonder how they manage memory.
There is only very little RAM on videocards.
Do they have to constantly swap the objects and textures from the main memory to the videocard?

So maybe this system is only fast if you dont have to render real production scenes with highres textures and several detailed objects.

[Morbid Angel]

this is one of the problems with gpu rendering. You would not be able to process high polycounts and large textures because the present cards do not have over 2gb limit (as far as I know).

[Morbid Angel]

Francis I see a lot of your work examples on the arion website. What were the avarage times for those images?

[cecofuli]

I study a bit Octane
It's in beta stage and there aren't a lot of features and, for me, is not usable in production. See here

http://www.refractivesoftware.com/features.html

And Octane take about 100 MB of VGA Ram for 1 milion poly.

From Radiance:

we're currently able to push about about 1 million polies per 100MB of video ram,
and we're working on improving this. adaptive rendertime / micropoly displacement is currently not supported,
and it will probably take a while before it's implemented as this is something that's very difficult to implement on GPUs,
however it should be possible given more research. Micropoly displacement is not planned for v1.0.
It's something that's incredibly difficult to adapt to a GPU,
so it's not something that we will have available in a few weeks.
I'm added regular pre-computed / stored triangle displacement fairly soon,
but for adaptive raytraced subdivision you will have to be a bit patient.
hardware tesselation is only useable if it's exposed to cuda, which it is'nt yet,
and i'm not %100 sure it will ever be available in cuda,
nor if it will be useable in a raytracer...
we'll have to wait and see what nvidia does.

[vlado]

I know vlado, but see this image

http://img146.imageshack.us/img146/968/set2s.jpg

from my friend. 10 min with I7 965 @ 3.2 + 280gtx

Well, the old dog has a few more tricks... here is a scene rendered for 10 min with V-Ray RT 1.50.SP1a and with the new SP2. This is on one i7 machine, no GPU. Only sun & sky, no other light sources.

Best regards,
Vlado