Is Vray better?

[animagic]

As far as lighting goes, vray and mental ray reply on physical algorythms which are similar to each other (vlado can correct me here if Im wrong). Both renderers are capoble of achieving identical lighting results...

Though I am not an expert on the math behind these renders, they may be similar, but they do NOT deal with lighting the "same" way. From an engineering standpoint neither one is physically accurate. That has nothing to do with visually accurate results and IMO Vray does a much better job at distributing indirect daylight than MR does. This does not mean that you can't create results with these renders that aren't very close to one another, identical would probably require an extra day ....

[crazy homeless guy]

Though I am not an expert on the math behind these renders, they may be similar, but they do NOT deal with lighting the "same" way. From an engineering standpoint neither one is physically accurate. That has nothing to do with visually accurate results and IMO Vray does a much better job at distributing indirect daylight than MR does. This does not mean that you can't create results with these renders that aren't very close to one another, identical would probably require an extra day ....

Would say that Photons in MR are useless? ...From my testing, using FG and Photons causes things to start to look cartoony or chaulky. But then again, i am much more experienced with Vray than MR.

[vlado]

From an engineering standpoint neither one is physically accurate.

What makes you think that?

Best regards,
Vlado

[animagic]

What makes you think that?

Best regards,
Vlado

Well first of all there is no way (that I am aware of) to get any analytical results from these programs, unlike an lighting engineering program like Luxicon or even the original lightscape. (before autodesk killed it)

The facts are that a specific energy source (light) with a specific energy distribution (ies) on a specific surface will always yield a specific amount of energy (foot candles/lumens/light). This is NOT subjective, my reason for stating this is only because all other factors that effect this result, are subjective.

Visually how your eye or a camera sees that energy can be very subjective. Some people think that the camera approach is more accurate, but I find that Vrays indirect lighting results are more pleasing to the eye, and yes that be subjective....

Below is a sample to demonstrate my point.. Same 10x10x8 room with the same ies light rendered in Vray (irradiance/LC), Mental Ray (fg) and Lightscape.. Though the Vray and the MR results are similar thay are not the same and the lightscape is very different. BTW we have done extensive testing with lightscape and it has always been very photometrically accurate (with in the margin of error (AKA lamp efficiency)) but NO we don't use it in production anymore...

[RErender]

offhand it looks like the lightscape render is in a different gamma space.

[^Lele^]

BTW we have done extensive testing with lightscape and it has always been very photometrically accurate (with in the margin of error (AKA lamp efficiency)) but NO we don't use it in production anymore...

You do know one of the builtin max GI engines IS the lightscape one, right?
Even, if possible, better than the original, in that it's now adaptive in meshing the surfaces.
But it remains a point-color based GI solution.
How's that more accurate than a DMC, subpixel one, escapes me.
And yeah, the first two look linear, the LS one looks at least a 1.8 gammaed one...

[dlparisi]

Did lightscape have some type of physical camera? It's been a while but I don't recall one. With that in mind, how can you really compare the images to a real light and room (I'm talking about the rendered image, not footcandles, lumens, etc)?

[animagic]

offhand it looks like the lightscape render is in a different gamma space.

The gamma or exposure, is not a issue. This can easily be adjusted or accounted for, look at the actual lighting patterns/distribution, they are not the same. This gets magnified as you add more lights and the scene gets more complicated.

[animagic]

You do know one of the builtin max GI engines IS the lightscape one, right?
Even, if possible, better than the original, in that it's now adaptive in meshing the surfaces.
But it remains a point-color based GI solution.

I have no idea how Autodesk "modified" the Lightscape radiosity engine when they got there hands on it, but I can tell you that the result with the current "built in lightscape engine" is not accurate any longer, I have tested it. They may have made improvements with adaptive refinement, but the have screwed up everything else. If we have to insure the accuracy of our lighting solution, we will revert back to Lightscape 3.2

How's that more accurate than a DMC, subpixel one, escapes me....

That is a bit of a problem as MR and Vray have not tools for measuring the heat energy of a scene file. If said tools existed the accuracy issue could be addressed more clearly. You have to have some kind of starting point for the discussion.

[animagic]

First of all, this discussion is getting a little off topic. The original question was about which render was better, which is a bit of a crazy question anyway, that is like asking Picasso which paint brush is better. My initial response was to explain that in our experience that Vray has served us well for still renderings and animations but MR has provided cleaner animations. Then somebody had to say that X is more accurate than Y. Which is just about as dumb as the initial question. (no offense, it is an understandable question, just way to subjective.)

Did lightscape have some type of physical camera? It's been a while but I don't recall one.

Lightscape does not have a physical camera..

With that in mind, how can you really compare the images to a real light and room (I'm talking about the rendered image, not footcandles, lumens, etc)?

This is more the point, how an Image is effected (exposed, gamma corrected or filtered) is mostly an artistic exercise, so when someone says this is more accurate than that is generally bull. This is why I introduce the engineering factor, a solution that is based on
a heat entropy which is the standard for engineering analysis is the only thing that you can even begin to talk about accuracy.

[^Lele^]

This is why I introduce the engineering factor, a solution that is based on
a heat entropy which is the standard for engineering analysis is the only thing that you can even begin to talk about accuracy.

And how is HE (heat entropy) different from V (Value) in a GI computation?
As an engineer, you'll appreciate you have a monodimensional represesentation of energy which decays based on the accepted set of laws of Physics in BOTH cases.
Now that you have fully 32bit per channel imagery, you can also measure those values a LOT more accurately than you ever could with lightscape (unable to save at higher than 8 bits depth, if I recall correctly).
Lightscape was a world-space solution, very interpolated, with some definite advantages at the time (viewport display one of many), but i still fail to appreciate how that can compare to a Brute force + LC for accuracy and precision when it comes to rendering GI.

[animagic]

And how is HE (heat entropy) different from V (Value) in a GI computation?

The biggest difference between HE and V is the HE can be related back to E (energy) in the form of cd/m2 or fc's, which are quantitative. You really are not looking at the images pixels for the accuracy or the pixels depth you are looking at the Luminance or Illuminance of the actual surface. Currently beyond engineering programs, only Max's built-in Radiosity "lightscape" can provide this type of analysis. (but don't trust it though)

Sorry, for departure on this thread, I just bothers me when some says that
"x's" render is more accurate than "y" when none of them can prove there accuracy anyway, so just don't say it. These are artist tools and the most important thing is to produce beautiful images/films...Mental Ray and Vray are both fine renders and can produce spectacular results in the hands or a good artist!

(BTW lightscape can produce high bit depth image 24/48 bit tiff, but does NOT have any kind of linear work flow or physical camera Just the magic "brightness" and "contrast" button

[^Lele^]

The biggest difference between HE and V is the HE can be related back to E (energy) in the form of cd/m2 or fc's, which are quantitative.

V is VERY quantitative too, brother.
I measure it pretty accurately with my exposimeter, and do all sorts of simple math and logic with the quantity.
V can be related back to energy as well.
Known the surface albedo, one can exactly measure the intensity of the incident light, from an rgb image (or read values directly from a raw lighting pass, or divide a lighting pass by a diffuse filter and measuting the result, and on and on...)
I don't see how a different representation of the same thing has to do with precision, accuracy and the such.
If you spoke of light distribution, and could show a sample of a renderer getting it way wrong, in a very perceptible way, compared to "engineering" programs, i'd be with you.
My experience of this stuff tells me well otherwise, though.
There just as much, if not more, physicaly correct math in a raytracer as there is in AGI or equivalent.
The rest is a case of setup and data readout, imho.

[gunny263]

Wow - fun topic (one that can go on ad infinitum).

I've had drinks with both Darin and guys from Frantic (before you joined them, Lele, so i'll have to catch up with you sometime in the future - so I feel I can add something to the thread and attempt to bring it back to the original post at the same time.

For the accuracy bit, I’ve looked at this subject quite a bit over the last couple decades extensively and both Lele and Darin are correct (as far as my experience tells me). When it comes down to brass tacks, you can interpret the representation of energy with a rendering produced by Vray (or any other rendering engine) using the pixel sampling method described by Lele (some really good relighting papers on the subject) and while that may be fine for matching backplates or set lighting for shot integration, it won’t stand up in court (not in the US anyway) as a scientifically accurate and verifiable method regardless of how nice the image looks and it will take you a whole lot more time than it does with LumenMicro, AGI32 or similar and time equals money so I’ll wager that the original poster’s bosses don’t really care if the rendering is scientificly accurate or not – that’s why they pay the EE staff. With all of the color mapping options, display limitations and post tools so widely available these days, scientific accuracy is rather moot in both VFX and Architectural Illustration since it all comes down to the human perception anyway.

For the original post, it’s more than likely going to come down to the ROI. If you are Max/Revit based, then MR will have the advantage every time simply due to cost and scene setup time. The speed with which one render engine cranks out pixels over another rests in the hands of the artist more than the underlying code and the quality is far too subjective of a subject so unless Vlado (Vray) and Bart (Mental Images) are in your office, it’s going to come down to whether the 8 guys can produce good work faster or you by yourself and their Revit scene already has materials and basic lighting in place so you better get busy.

Rmejia: your avatar is evil btw - guess you hear that a lot

[rmejia]

Rmejia: your avatar is evil btw - guess you hear that a lot

hehe, it's been mentioned before