[JDHill]

Thanks SJ...that's what I thought.

One other question...how is the lighting set up being approximated between the two scenes? (emitter shape, type, efficiency, etc.)

Thanks,

~JD

[michaelplogue]

Interesting test!

I'm curious.... With the 3ds Max plugin for V1, we still have the old Beta materials available - in addition to the new MXM texture.

In order to compare just the differences in the rendering engines, could you use Beta materials for both engines to see what sort of differences there are? Although obviously the Beta materials are not optomized for V1, at least you could use exactly the same materials in both cases. This would limit any differences in the resulting images to what the rendering engines are doing.

Just a thought......

[JDHill]

...not exactly, Michael...Maxwell always uses .mxm materials...the plugins are responsible for creating the material, even if they do so using a beta-style interface...they're working the same as a wizard. So...unless we know what's actually being created, .mxm-wise, we might be comparing apples and oranges --> really just testing plugins...and not the engine.

~JD

[Frances]

...not exactly, Michael...Maxwell always uses .mxm materials...the plugins are responsible for creating the material, even if they do so using a beta-style interface...they're working the same as a wizard. So...unless we know what's actually being created, .mxm-wise, we might be comparing apples and oranges --> really just testing plugins...and not the engine.

~JD

A simple diffuse material with the same RGB values would not make the difference. If that isn't the case, then the much anticipated new-and-improved Material Wizard will be a waste of time.

Four bounces will only go so far, but gee it gets there much quicker, doesn't it?

[JDHill]

...ok...I'll just stop testing and go home then.

[mverta]

The photo-comparison you can dismiss; there are just a billion billion variables there which we can't compensate for.

The second test is interesting... and don't devalue your experience as a photographer. You're far more experienced in the realm of image perception than the average software programmer. There may be a lot of math going on in there, but the math is supposed to generate images that behave like photographs.

_Mike

[lllab]

mybe there is really something with the v1, hmm...

very good test!
(the 2nd)

cheers
stefan

[Frances]

...ok...I'll just stop testing and go home then.

What - the mat wizard? Test away by all means. When we recieve the engine we were supposed to have, it will be worth quite a lot.

[mverta]

Just so this doesn't get carried away, let's not let half-truths and grinding-axes derail the pursuit of a common goal.

A fact is that v1.0 exhibits different behaviour in some ways than beta, and is vastly more accurate in many respects, to say nothing of the fact that performance improvements, which are ongoing, are increasingly making working with Maxwell a night-and-day experience, favorably, over beta. Those are facts.

Now, the big gray area starts with the question, "What if more accurate doesn't look more accurate?" Personally, I think that truly accurate will look real, but I also won't blindly devalue the idea that beta might have had a "warm-and-fuzzy" quality, which, while not accurate, feels nice next to the skin. On the other hand, simple tests like this sometimes seem to illustrate quantitative differences such that beta has better distribution. But of course, I can counter with equally simple tests that show 1.0 is handling light better.

What I have found frustrating in my own tests is that it's EXTREMELY elusive. Some simple tests make a hands-down argument for 1.0, some for beta. And in truth, I've found it is in the most complex scenes that the differences are most telling. There is no question that the psychological impression given by a render has to be weighed against accurate representation of photographic reality. I, for one, hope that they actually converge at a single point somewhere, and perhaps we're not at that point yet. But as it stands, neither "camp", beta nor 1.0, is in an unchallengable position to declare superiority, since they can both "prove" superiority in different respects. The debate, and the testing, continues.

_Mike

[Voidmonster]

Excellent points, Mike. It's good to have the perspective of someone that's done really significant testing with both versions.

I'd also chime in here and point out that what looks right and what is right do not always converge. Memory and perception do strange things. It makes getting good photographs of people a less-than-straightforward process.

[Frances]

Just so this doesn't get carried away, let's not let half-truths and grinding-axes derail the pursuit of a common goal.

So what is the number of bounces hard-coded to in V1? It shouldn't be a secret, should it? You guys are all dancing as fast as you can to try to think of an explanation that will placate the masses, or at least the more observant. The BS about the new materials being the culprit is just that -BS. That leaves us with either the emitters or the engine. Which one is easier to fix?

Better get busy with that.

[aitraaz]

Sorry to break in here , a silly question brought about by the mention of bounces:

If you add the -rs:0 switch for the mxcl command line flag inside of studio, it appears that mxcl will render with the rs0 preview engine, and thus with bounce control? It seems much faster. Is it using the studio global settings for preview bounces?

Has this already been mentioned? Sorry, I'm having quite the fun night here

[mverta]

So what is the number of bounces hard-coded to in V1? It shouldn't be a secret, should it? You guys are all dancing as fast as you can to try to think of an explanation that will placate the masses, or at least the more observant. The BS about the new materials being the culprit is just that -BS. That leaves us with either the emitters or the engine. Which one is easier to fix?

Better get busy with that.

Frances, it's not about bounces... and yes, the material model has a huge impact on the renders. Whether you are deliberately devaluing or being ignorant of the facts, your oversimplification and resultant slamming is putting you dangerously close to irrelevancy. I'm not here to defend Maxwell, I'm here to clarify rambling, incorrect tantrums like yours so we can all find our way clearly to what - ostensibly - we all want. Which is the most realistic, best looking engine possible. I've already stated, and it can be quanitfiably proven, that both engines have pluses, with 1.0 at the advantage in accuracy in specific areas and performance, while beta seems to posses an "overall" feeling that we like, even if it's hard to define. But tests like this, and a battery of my own, and of others, are comitted to trying to nail down what is truly an elusive issue, and absolutely nowhere near as simple - nor conspiratorial - as you suggest. Just freakin' be a part of the solution, already.

_Mike

[SJ]

Thanks for the comments anyone. I'm glad my test doesn't come along as pointless, at least.

One other question...how is the lighting set up being approximated between the two scenes? (emitter shape, type, efficiency, etc.)

The emitter shape is exactly the same - a rectangular plane on top of the box. Studio tells me it has an intensity of 100000 Lumen, RGB 255,255,255.

"What if more accurate doesn't look more accurate?" Personally, I think that truly accurate will look real, but I also won't blindly devalue the idea that beta might have had a "warm-and-fuzzy" quality, which, while not accurate, feels nice next to the skin. On the other hand, simple tests like this sometimes seem to illustrate quantitative differences such that beta has better distribution. But of course, I can counter with equally simple tests that show 1.0 is handling light better.

I second your statement. But, a good photographic medium has to match the perception of the human brain and has to be "biased" therefore. That was a heavy problem of first digital cameras. The pictures produced by them were looking digital at first glance. The reason was the different gradation curve of film and image sensor. A photographic film that produces "nice" results has a so-called shoulder at its light and a toe in its shadow areas while a digital sensor reproduces light quite linear. The reason for shoulder and toe is the ability of the eye to differentiate finer tonal differences in light and and lesser in shadow areas.
Maybe V1 produces a more linear gradation curve while Beta had toe and shoulder that made it look more like a photo taken on film. Just an idea…
Here’s something about tone curves and their appearance to the eye. I have just read quickly over it, so I can’t guarantee the content matches exactly the problem.
http://www.normankoren.com/digital_tonality.html
http://www.covingtoninnovations.com/dslr/Curves.html
http://www.rags-int-inc.com/PhotoTechStuff/TonesnZones/

I'm just thinking about a scene to find out if there is a difference in material reflectivity between Beta and V1 or if light power falls off differently while simply travelling through "air". Any ideas?

[Thomas An.]

Nice tests SJ
Sound quite rational to me .