Page 2 of 12

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 5:56 pm
by Lutze
Why was there no word about ChaosGroup/Vray also doing gpu-rendering?

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 6:14 pm
by timbarnes
I think that these kinds of comparisons are useful if they are clearly documented. My own experience as a modo and Octane user is that the speed is not really the greatest asset of Octane. I get pretty fast renders (generally) from modo, and I know that experts who know what to tweak get even better renders. By linking my Mac Pro, my Windows 7 machine, and stealing :shock: my wife's iMac, I can get 14 cores working on renders simultaneously with a single license.

My experience is that I like the feel of the Octane renders better (at least the way I'm using them), and I like to have alternatives. I have seen many cases where Octane needs to run for a couple of hours (GTX 470) before the noise clears up—perhaps because I have a ton of glass in my models. Of course this may change with the new version.

There have certainly been some cases where modo renders have taken me days and days (to do a set of archviz renders on my 8-core Mac Pro), and that's still with abstract materials. I'm hoping that I can get more speed out of Octane as I learn more about it.

I think that one reason for the video is that a lot of people in the modo community have been wondering and asking when Luxology will implement GPU rendering: the company has a good reputation and produces a great product (I have no financial motivation for saying that: I'm just a satisfied user and customer of both Luxology and Refractive).

Finally, Luxology and Refractive are not really in competition, as you can't model in Octane. I suppose someone might decide to invest in Octane rather than upgrading to modo 501 when it comes out, but I doubt it: Lux usually provides a lot of functionality in a new release and makes it worth the upgrade.

And finally finally :? the cost of Octane if you factor in a couple of 460s or 480s is not as low as the software price would suggest. So as in all things, there's no free lunch. I am a huge fan of Octane, and of modo. They help me to design and to communicate design ideas, and that's immensely valuable.

tim

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 6:28 pm
by Lutze
As a Modo and Octane user myself this is exactly what I feel. I use both about 50:50 but I usualy get a decent looking image-setup faster in Octane than in Modo. It feels as if there is more freedom to experiment. Both are very good programs and I agree that there is no reason to think of them as adversaries. Although I like Brad a lot I think the presentation looks a bit like a intel-inside-ad. Comparing a future version of Modo against a beta-version of Octane is also not very usefull to come to a real conclusion about gpu v. cpu.

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 6:40 pm
by radiance
what GPU was used ? what CPUs were used ?

how many bounces ? that star trek head will look the same with 2 GI bounces and 1024 as there is nothing to bounce off.
In fact, i'd be suprised if the modo render on the left has any GI at all.

also, a finely tuned irradiance cache against a spectral path tracer is always going to give you different results.
and lastly, i don't think those were octane renders, both images are so alike in terms of gamma/colours, i think they both came out of the same engine, and the left image was just left rendering longer.

The very fact that they are producing this video without any fair information to try to convince the less informed user means they fear GPU based rendering.
This is just another handicapped / unfair comparison, but i'm suprised it's delivered by a professional company instead of the usual flamewar thread.

Radiance

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 7:00 pm
by radiance
here's the magic bullet :)
I'd bet money that is an octane render, the reason is on the image.
fireflies.jpg
Radiance

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 7:12 pm
by Chris
What do you mean Radiance?

These are my observations:

The "octane" image has noise obvious, the ball with the chrome material is faceted (no normal smoothing turned on?), Different caustics and "higher" exposure settings (eg the image is a bit brighter)


Chris.

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 7:16 pm
by radiance
Chris wrote:What do you mean Radiance?

These are my observations:

The "octane" image has noise obvious, the ball with the chrome material is faceted (no normal smoothing turned on?), Different caustics and "higher" exposure settings (eg the image is a bit brighter)


Chris.
That image should have fireflies on it and it doesn't. it looks more like an MLT render to me. it's got glass, very shiny materials etc...

Radiance

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 7:17 pm
by tchoa
me too Radiance!
That is detestable. They behave like (i don't know, perhaps cause i'm french).
I could'nt see all the video. Technically speaking it's... hmmm how to say?... I lost my words.
How could they do such very bad comparative publicity?
Is it to convince their new so-called-bid-customers-in-the-movie-industry to use their renderer?
Let me laugh...
Are they sick as the guy talking seems to be?
What do they have in their head to do that?!!
Ridiculous.

F

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 7:32 pm
by Lutze
Radiance,

I have a question about the image you choose. Im am not shure, but it seems that on both images the material settings are identical, and I mean identical down to the last specular higlight. Is it possible to render the picture in two different renderers and to achieve such a close resemblance, even when using the same HDRI to light the scene: I am just curious and maybe its possible. I often switch programs when working on a project, just for the fun of it and I seem never to get this level of resemblance (which is the reason I shift programs in the first place - the slightly different look).

Re: GPU rendering is not the magic bullet it was promised to be

Posted: Wed Aug 11, 2010 7:37 pm
by radiance
Lutze wrote:Radiance,

I have a question about the image you choose. Im am not shure, but it seems that on both images the material settings are identical, and I mean identical down to the last specular higlight. Is it possible to render the picture in two different renderers and to achieve such a close resemblance, even when using the same HDRI to light the scene: I am just curious and maybe its possible. I often switch programs when working on a project, just for the fun of it and I seem never to get this level of resemblance (which is the reason I shift programs in the first place - the slightly different look).
I don't know. I've never managed to get a render look %100 the same between lux and indigo in the past, and those are identical in algorithms used.
Indigo multiplies all it's colours by .8 by default, while lux does'nt etc...

I don't think the image posted on the right is an octane render, it's got no fireflies.

Radiance