Have you checked the memberlist at M~R forum?

[wda]

schlitz. no schlitz? blatz. no blatz? improvise.

 

Best thing I've seen yet associated with anything "maxwell render"

 

Maybe include "no FoxHead 500?" then improvise...seeing they were all brewed are local to me LOL

 

WDA

[Fran]

First of all, let me say that I suck at shaders and shader lists and it is a wonder that I can spell s-h-a-d-e-r.

 

That thin film thing they were talking about - is that a dielectric? Is that something that could replace window glass by itself? Does it have the same issues? We could maybe have soap bubble windows. Big, flat soap bubbles.

 

Stop laughing, you shader know-it-all types.

[_PopArt]

Slight adjustment. It seams that Vlado was saying should be after Vray 1.5.

About halfway down the page.

http://www.chaosgroup.com/forum/phpBB2/viewtopic.php?t=10323&start=75

Actually the whole page is worth reading, because back in Feb 2005 Vlado was saying the problem would be seeing caustics through glass. Which makes me wonder if it is indeed a material problem, or whether it is a problem intrinsic with Bidirectional path tracing.

 

Thanks for the link. Quite interesting!

[whiskey]

the idea isn't bad..just setup a clear coat with 6mm thicknes..voilà...if you can select its abbe and refractive index that is...if it is inherently different to dielectrics that is..otherwise i see the same problems..wait there is no such things as dielectrics in RCs..still the problem prevails:o

 

EDIT:

you know in certain countries where the don't have glass the hang jellyfish into the windows

......

is that the way we'll solve the glass of water problem? model a block of water and render a 4mm clearcoat around it for the glass?

[Devin Johnston]

bricklyne has really analyzed this pretty well so my question to him and all of you is if it turns out that Maxwell can't be a 100% physically accurate render engine is that a deal breaker for you? Let's just say that the only negative impact is that lighting analysis might not be possible or at least not 100% accurate, but the quality that drew us to the software in the first place is still there.

[Ernest Burden III]

bricklyne has really analyzed this pretty well

 

He may have been even drunker than me...

 

if it turns out that Maxwell can't be a 100% physically accurate render engine is that a deal breaker for you?

 

My point was that a goal of perfect accuracy was so unattainable as to be a fool's errand. Even without Heisenburg's damning principle. A model, an image, will ALWAYS be an approximation or else it would BE the thing being modeled. So you cannot hold the concept of 'unbiased render engine' to a very high standard without dooming yourself to failure.

 

So 100% physical accuracy is not required, is not a deal breaker for me. I don't think it is even possible and spending time and money chasing it is a waste. Cheat. I'm not advocating cheat's two closest friends--lie and steal. But yes, cheat, fake, compromise. Just make it all look good and be easy to use and real-world logical.

[AdamT]

Well they've already got lie and steal covered, so why not add cheat for the hat trick?

 

Anyway, a little cheat certainly wouldn't be a dealbreaker for me, so long as the images are *mostly* accurate, i.e., you can still set up a scene using real-world measures and get a pretty realistic-looking outcome.

[bricklyne]

bricklyne has really analyzed this pretty well so my question to him and all of you is if it turns out that Maxwell can't be a 100% physically accurate render engine is that a deal breaker for you? Let's just say that the only negative impact is that lighting analysis might not be possible or at least not 100% accurate, but the quality that drew us to the software in the first place is still there.

 

 

Devin, Maxwell, CAN'T be a 100% physically accurate renderer. Haven't you been paying attention? That was just a tagline they were using to sell and promote their product. Or at least generate the hype that started it all. Heck, even Lightscape, as good as it was, was nowhere near the accuracy needed to make a claim like that ( lack of DOF, SSS etc)

 

A lot of people have thus far realized this, hence the reason, as I pointed out before, why the whole unbiased-ness of their renderer has taken a back seat in the marketing department, to other stuff like the material layering, Studio setup and all that stuff.

 

Personally, I was less won over by the entire unbiased argument, and more so by the simplicity of the setup ( which they've F'd up to some extent with the RC studios) versus the high quality achieved. Let's not kid ourselves, with enough tweaking, setup, test-renders, trial and error,and experience, a Vray user can fake the realism of their image output to nearly or at the quality of Maxwell lighting. But the key here would be the number of tweaks required, or at least the level of experience required to know this. When Maxwell was first released, a lot of people especially at CGtalk, lambasted the whole easy-setup thing with the lame argument that this was going to mean that anybody or even a monkey with a computer would be able to produce high quality renders, without having to go through the same sort of rigamaroles that CGvisualizers have had to go through for years to be considered experts In other words, Maxwell was (for that particular moment in time) threatening to making CGviz easy enough for anybody to do and still seem an expert; assuming NL were, of course, able to get their rendertimes down.

 

When I first heard about the fact that Maxwell employs Bi-directional path-tracing/MLT related methods, I knew that they would have to use cheats at some level or another to at the very least get around the speed issue. Bi-directional path-tracing and MLT methods are essentially brute-force techniques to simulating reality and as such processor hogs. And outside a monumental leap in CPU processor-speed technology, there's no way to get around this fact using software optimizations, or at the very least there's a less-than-generous upper-limit or ceiling to doing so. So, in any case, I was always expecting the would have to cheat vis-a-vis the speed issue; that they now have to do so to get around other issues as well, such as the glass-dielectric issue and perhaps the clipmap stuff, will not be any more a surprise for me than it would be a deal-breaker.

 

I think most people knew this well, even if only at the back of their minds. Which is why when they first released the Beta, there was a greater furore over what appeared to be a loss or drop-off in quality ( i.e the blurriness that they were forced to remove via a patch) than there was regarding the less than considerable and expected speed increase. Most people are attracted to Maxwell because of the lighting quality first and foremost rather than the other stuff; and more so for people who have already done CGViz withother software and who know how had it is to get that quality in other software even with tweaks ad infinitum.

[Devin Johnston]

Devin, Maxwell, CAN'T be a 100% physically accurate renderer. Haven't you been paying attention? That was just a tagline they were using to sell and promote their product. Or at least generate the hype that started it all.

 

 

I wasn't trying to argue that it should be or that it would be, I was only asking whether or not people would consider Maxwell a failure if it was unable to achieve the main feature that it was sold under. I don't care if Maxwell is biased; my only concern is being able to use it effectively in production. I think at the very least NL should give users of Maxwell the option of switching from a unbiased to a biased setup. This would answer both sides to the equation, speed for those who need it and accuracy for those who are willing to wait for it.

[AJLynn]

Repost of something I put on the NL forums. I included PPT because it's the "in" thing over there

 

I ran Vlado's test. Following are on a P4-2.4 laptop:

 

Vray normal:

 

11 minute render including calculating caustics

 

 

Vray PPT:

 

8 minute render using saved caustics map from the last one. PPT settings were 750 subdivs and 0.02 sample size.

 

 

Maxwell beta:

 

I let this go 2 hours, which was more than it needed but I wanted to see if anything would happen.

[Devin Johnston]

why is it so dark?

[AJLynn]

why is it so dark?

 

I took a guess on the camera settings, went away while it rendered and it turned out I was wrong - but it still illustrates the point.

[dave davidson]

bricklyne: i agree i bought it for the lighting and feel to the images i saw comming from the renderer not the materials and def not the studio (which i really dont want)

my 2 cents