Anyone using Indigo?

[Cesar R]

Is anyone using Indigo?

[AJLynn]

It's an MLT engine that's not GPU accelerated. This place is full of ex-Maxwell users who'd warn you on speed.

[Cesar R]

What does MLT stand for? - I Maxwell GPU accelerated?

Edited September 22, 2010 by Cesar R

[Jeff Mottle]

http://en.wikipedia.org/wiki/Metropolis_light_transport

[Cesar R]

Thank you for the link.

[AJLynn]

What Jeff said. The ones linked in the Wikipedia article are all CPU. None are fast - Maxwell users will have learned by now that there are some scenes you just can't get a render of that's not noisy. In the good old days of 2005 there were people posting Maxwell renders that had render times of (I'm not making this up) weeks.

 

The popular GPU renderers (Vray RT for GPU, Octane, Arion, iray) use similar approaches (monte carlo path tracing algorithms) but they're not quite the same as MLT - they're less optimized because the render threads on the many "cores" (I hate using that word because it implies nonexistent parity with CPU cores but I don't have a better word) don't communicate.

 

There are types of optimization you simply can't do in a current generation GPU renderer. (Of course, this is sold as a feature, not a bug. Quoting the iray marketing literature, "Interpolation techniques, which trade final quality, predictability, and simplicity of scene specification for performance, form the core of most current global illumination renderers. Unlike them, iray rendering is based on deterministic and consistent global illumination simulation algorithms that converge without introducing persistent approximation artifacts." This of course is euphemistic phrasing for "This algorithm does not allow for the types of optimization you are used to seeing.") The thing is, this doesn't really matter - the limited set of calculations these renderers implement are ones that are handled so well by GPUs, you can be as wasteful as you want and still come out ahead. (According to a friend who's in i-banking, this is how the major banks used to operate WRT money and many still do.)

 

So, to sum up, MLT is incredibly slow. You dumb it down a bit and rewrite it in CUDA/OpenCL and run it on GPUs and suddenly it's fast. But dumb.

 

What rendering software are you using now, anyway?

[Cesar R]

I use typicalled MentralRay. Back in 2005 I downloaded a Maxwell demo, I was impressed by the concept and the quality of the renders that I saw were produced but yes it was very very slow (in part also due the the available hadrware). I could not get used to the interface. - I knew of the technology but ignored it. Recently I say on youtube some Octane demos and here I am building an new computer as my Quadcore 27" imac does not have a CUDA card (I dislike apple for putting ATI cards!!! - nothing against ATI, I just dont like the fact that I am missing out because of a purchase dession I made before I know about GPU rendering - no ones fault).

 

I also always knew of Fry. Played w the demo a few years back but didn't really dived into anything. I did like the interface a little better. So at the pressent time my quest started as a research based on Octane, which lead on to Arion. However I understand, as I read somewere, that Arion can't do some of the things Fry can... and thus I started to look at Fry-like engines. This is how I arrived to Indigo which I am liking very much. It is simple to use, and their FORUMS and SUB-FORUMS are accesible to everyone unlike Random Control's. So far more pre-sales support and knowledge available to someone shopping.

 

Needless to say there is good information of the completed work and wip on randomcontrol and I do like the examples I have seen.

 

I think I need to look at features list side by side.

[AJLynn]

Don't buy a new computer to get CUDA before settling on software, and don't settle on software before fully understanding the cost/benefit. Don't take this wrong, but based on the questions you've asked lately about various render technologies it doesn't seem that you're fully prepared to make a purchase, but may be buying into some GPU rendering hype, which is something I've been trying to discourage as the technology has not matured, the market hasn't settled and the limitations are not apparent to most customers. (This is the same effect that caused a lot of us to buy into Maxwell when it was in beta and later get burned as it became apparent that the things we thought were minor issues that would be worked out quickly turned out to be limitations inherent to the technology.)

 

Octane is fast, but limited, due largely to the limitations of GPU computing technology - anything a GPU can compute, must be written in a language that was originally designed for processing on-screen 3D graphics. You'll find that Octane is missing some of the shader features you find useful in mental ray, and you may not be able to get some of your images up to "production quality" that you're used to. This was discussed in the video linked here: http://forums.cgarchitect.com/61912-brad-peebler-luxology-talks-about-gpu.html - the president of Luxology decided GPU computing was not at the place it needed to be for them to implement their renderer to their satisfaction.

 

BTW, don't fault Apple for not including CUDA hardware - Apple isn't supporting CUDA and has gone out of their way to make OpenCL an open standard by writing it, then giving it away to the group that manages the OpenGL spec. This is a very good thing for the industry and the market, as CUDA is a proprietary spec that runs only on one company's hardware, and any time a proprietary spec is allowed to become important enough to influence a large percentage of customers it creates a market inefficiency. Multiple GPU companies need to remain equally viable options to foster competition - e.g., if it weren't for ATI's push on power efficiency, the newly released nVidia chips wouldn't be lower power than the last round, and if it weren't for Intel's push on power efficiency, ATI (owned by AMD) wouldn't be doing the same because AMD wouldn't be, but it was AMD who got that going in the first place by shaming the Pentium 4. What needs to happen here is that more companies need to follow Chaosgroup's lead and go into OpenCL.

 

BTW, you mention looking for a rendering engine for your Mac - Modo, Cinema4D and Vray for C4D come to mind.

[Cesar R]

My mac is a PC. So my signature it outdated. I am honestly much happier with it being a PC, and I mean by this is that I have devoted as much of the HD to windows 7 and never boot into OSX. Maybe someday when Autodesk ports Revit and 3dsmax.

 

I know C4D is a great program and so is modo but I like max you know? - maybe I should get into modo and give it a try.

 

Thank you for the advice and it what you are saying makes complate sense. In regards to your asumption, you are correct but at the sime time you're not. I am a fan of the GPU rendering but I am not buying a machine just for that. I Need a second machine, something dedicated just to rendering an heavy work. There are many reasons for the new pc.

 

The reason for my "naive" questons are just lack or knowledge and ignorance. I have been a member for over 10 years and I value that knowledge base you know? thus I turn to this board to properly inform my descissions. I am not buying anything tomorrow (software wise) - I already ordered the parts for my pc, and I feel I need to fully understand my options to make an educated decission. What better place to get feedback from than those who use it.