Radiosity vs mental ray GI


#1

I am confused. Do these 2 things do the same thing? Not often I hear of people using radiosity and even less use of light tracer.

And why does the skylight take so long to render? Can I use omni lights with the light tracer or radiosity?

What does final gather do besides just make GI more accurate ?

Does anyone ever use the Dreamscape SUNLIGHT by itself ? (without other dreamscape objects)

just some random questions I have been wondering


#2

Radiosity, MR? All depends on what render you like to use. Mental Ray, V-Ray, Brazil, are third party renders that calculate and handle real world bouncing light in different ways. You will not know what you like best until you have tested them yourself.

There are ways to make the skylight render faster, but for final results, there ain’t no gettin’ around the waiting.

Final Gather? From my limited knowledge, it takes your lighting samples and works them into every crack and crevace

Got no idea about Dreamscape.


#3

3dsmax radiosity is the same thing as MR global illumination right?

It is just THIER way of doing it?

I never have used radiosity or MR GI much so I am wondering do I need to know them both or just learn MR since that is what people seem to use.

The only other issue seems to be some materials do not have a MR equivalent therefore I must use 3dsmax.

I just have never read anything that basically says THIS is when 3dsmax radiosity is better and THIS is when you should use MR


#4

Radiosity is a lighting solution much like Global Illumination. It wouldn’t hurt to run a couple of radiosity test for knowledge purposes, but I think once you start learning to use Mental Ray, you’ll get hooked on it.

No, you probably won’t find every material you need in MR, but you can use combinations of standard and MR shaders. Again, they calculate things a little differently. Learn to use those MR shaders, you won’t be disappointed.

That last question, I’ll leave for the pros on this forum.

Good luck.


#5

This is what I think to know about Gi (please correct/complete me):
G(lobal) I(llumination) or Indirect Illumination is a general term for lightning simulations that contain bounces.
There are several different methods to archive this.

View independent methods:
-radiosity (is linked to the density of the geometry) - max scanline
-photonmaps (shoots around photons - tricky for sharp corners) - MRay, VRay

View dependant methods:
-QMC (slow and noisy / can deliver great quality) - MRay(called Final gather?), VRay, Brazil
-Irradiance Maps (smooth with very good quality/speed ratio) - VRay
-Lightcache (method with infinite amount of bounces) - VRay

-Progressive Path Tracing (can be used like Maxwell -> constantly refines image) - VRay


#6

hey grau,
as I know QMC is not viewdependent, it shoots rays into the scene in a random pattern and they keep staying there. flickering occurs only when a low sample rate is used and things/cam are moving because it need to shoot new rays. speed is average but with some tweaks and tricks it’s quite fast and ready for production, as you can see in movies from Blur studios (they use Brazil).

Correct my if I am wrong…

Cheers,
Leon


#7

View-dependant means that it doesn’t produce the solution for the whole scene at once, only for what you see. So does QMC, otherwise you wouldn’t need to recompute it when the camera moves.

here’s a link with a general comparison between the the different methods and its pros and cons
http://www.vrayrender.com/stuff/gimethods/gimethods.htm


#8

Radiosity is physically correct method of indirect illumination calculations, based on conservation of light energy. It is CG realization of energy conservation law. But this method is very computationally expensive, and because of this fact popularity of alternative GI methods was increased.
All other algorithms different from Radiosity are fakes(in greater or lesser degree).The most popular alternative algorithms are:
1.Photon mapping (Final Gather in MR terminology) - based on tracing portions of light energy,called photons (CG photons don’t have so much common with physics) from LIGHT SOURCES. It’s main advantage of photon mapping-you can render a scene once,save the photon map to a file and that’s all. You don’t need to calculate GI, if you will decide to change view. Besides pros Photon mapping has some contras: difficulties with corners and meeting-points.
2.Direct computation (Monte Carlo method,“brute force”) is very accurate and computationally expensive. When you use “brute force”, your renderer builds hemisphere for every point of image and samples integrand.The main part of this integrand is diffuse part of BRDF(Bidirectional Reflectance/Refractance Distribution Function)-BRDF calculates all light, which went through the hemisphere from all directions. Sampling of BRDF is chance choosing of one specific value of BRDF. It means that renderer chooses one direction of light spreading and traces light ray at this direction. When this light ray collide with nearest surface, the process repeats - building hemisphere,sampling BRDF and so on.
For example, 1000 rays went through the first hemisphere. Every ray will create one new hemisphere, when it will collide with surface again, and every of this 1000000 rays will create 1000 rays at next point of colliding and so on. It’s like avalanche. And your renderer needs to calculate this huge amount of rays.
I think, i don’t need to explain, why direct computation is used very seldom.
3.Irradiance map is a light-weight variant of brute force.Renderer does not computate every ray - only some of them. Others are interpolated. This algorithm takes less time and keeps excellent quality. But you need to recalculate irradiance map if you change view.

Mental Ray is not appropriate solution for GI-based images. Realization of GI at very good level is implemented into Final Render,Brazil r/s and ESPECIALLY -VRay.

However Radiosity is THE MOST PHYSICALLY CORRECT REALIZATION OF GI.


#9

Ups my fault, please excuse my wrong statement…


#10

so you mean 3dsmax RADIOSITY is the BEST so the quality is better than mental ray GI? Just takes longer?

I have the impression that besides 3dsmax scanline default rendering that about

1% people here use light tracer
4% people here use radiosity
95% people here use MR,vrya,brazil

something like that?

So basically I guess it is best for me to ignore light tracer and radiosity and for high quality single frame rederings concentrate on MR since that is what everyone else seems to be doing


#11

so when everybody else jumps out of the window, you’ll too NO, to be serious just use what suits your needs … you don’t need to use MR only because everybody uses it.


#12

-max’s radiosity is not the best or only correct way, all solutions are based on a physical model
-the methods with the best quality/physical correctness are qmc and path-tracing
-the other methods use approximations and therefore are less exact
-just try out the options you got, can’t hurt to know them all


#13

I want to do something beyond standerd scanline render. and since there does not seem to be alot of talk on rendering with light tracer I don’t want to use that.

It is confusing because all thet many books I read they always mention radiosity and MR briefly but never seem to explain which one is better for what.

I see a whole forum dedicated to MR so ill stick with that for now


#14

to punkouter
Yes. Radiosity is the most correct method of calculating GI. The most people use VRay for interior rendering, because VRay works much faster, and obtained results look realistic - almost like Radiosity render. But 99% companies,spicialized in interior rendering, use Radiosity - because they have enough resources to carry out these huge expensive renderings.But for individual use the best solution is VRay with its ability to provide very good results at acceptable amount of time(i.m.h.o.,of course).

to grau
I have slightly different opinion. QMC is good GI solution, but it is only approximation also. To calculate diffuse lighting absolutely correctly, it’s necessary to solve integrals of illumination analitically. But in most cases it’s impossible, and in order to solve these integrals QMC method is used. QMC gives opportunity to solve these integrals as sum of LITTLE quantity of ACCIDENTALLY chosen functions. QMC is method of numerically solving integrals of illumination (not analitically). It means, that QMC method is approximation also.
Besides QMC has one BIG disadvantage: to increase quality of calculations as two times much we need increase size of it (calculations) and rendering time accordingly at four times much. If we want to remove noise, we must pay very high price.All I said was concerned with brute force QMC. Irradiance map method is modification of QMC,it means that irradiance map method is approximation of approximation.It cannot be physically correct (at least like Radiosity). It gives good results, takes less time - but ISN’T AS ACCURATE AS RADIOSITY.

I mean that when we deal with necessity of REALLY good results, QMC isn’t better solution. Of course Radiosity has own disadvantages, but its quality is inimitable.

I’m not impeccable, but results of Radiosity is much better than results of QMC (when you use settings of QMC, which our computers can “swallow”).


#15

I rendered this using Dreamscape Sunlight. It’d pretty good for exteriors. The fog, water, sky are all from dreamscape. The bridge is a free download from 3d cafe.


#16

@Cuprum
You’re right about QMC being an approximation as well, but in the end all that can be computed in less than an unlimited amount of time is an approximation if the resolution of reality is infinte. What I meant is that QMC is unbiased whilst the quality of max’s radiosity depends on the density of the scene geometry. As an Example a low quality QMC image will give you a high amount of noise but a correct simulation. While a low quality radiosity solution will smooth out the result delivering an incorrect result.

@punkouter
sorry for hijacking your thread


#17

to grau

Look at my post at previous page. I edited it, and my position would be more clear to you, if you read it.

to punkouter

Light Tracer was developed to provide quick rendering of brightly-lit outdoor scenes.Unlike radiosity, the Light Tracer does not attempt to create a physically accurate model, and can be easier to set up. Main advantage of Light Tracer is SPEED.It provides very quick rendering for outdoor scenes with usage of Skylight. It’s the worst solution for indoor scenes. Interior rendering isn’t for Light Tracer.


#18

so how does light tracer/skylight compare to dream scapes sunlight?

So if you want a outdoor scene with something quicker than MR and better than a simple omni light then than is when light tracer is really used?


#19

from my experience,FG, apart from making the calculations more accurate (and more render-intensive) it provides the bounces with the capability of casting shadows.

here is an example. The render on the right has FG disabled and rendered in 9 seconds, while the one on the left HAS FG enabled and rendered in 1:53 (quite a difference, see? the samples for the FG were set at 100, which is not really too high). You will also notice that the bounces do look very different. the only light in the scene is a directional omni aimed at the wall thtrough the hole on the ceiling set to raytraced shadows.

hope this sheds some light on your doubts. i might post the scen if you need it


#20

For me the whole question should not be about
which model is the most physically accurate, but
rather which model produces what is needed for
the particular piece I am working on.
For vis and architectural work Mental Ray along
with its GI/FG combo has got to win. For outdoor
work Light tracer can produce some great results,
if you are doing a light study the scanlines Radiosity
engine can give you spot on results.
The end result of each of the lighting models available
depends to a large extent on how well you understand
and know how to use the options provided for each.
Safe to say just switching on FG in MR and rendering
will NOT produce the results you want. Experimentation
seems to be the only way to get to understand which
model works best in what situation. IMHO. :stuck_out_tongue:

Geonome