[Aldaryn]

mverta, I doubt that "a_lizzard" would be Vlado...

Anyway, it would be quite odd... "a_lizzard" with no personal info at all, as Vlado, quoting himself.

[oscarMaxwell]

Hi Vlado,

Nice to know you.

I wouldn't have answered normally since this is not the place for it, but since the comments were directed on me, I will say that I still hold to my original statement - caustics in mirrors or through glass are always going to be very difficult cases for Maxwell - if it stays unbiased. Look at the many posts where people complain that sunlight does not pass through glass windows... in some cases, this may be due to bugs, but in others it is the same limitation of "unbiasedness". Of course, eventually these effects will show up and will be noise-free - if you have the time to wait for them. Often, you won't have it, even if computers or Maxwell suddenly became 10 times faster.

It's not easy, but you know that is possible when dealing with no Dirac functions. It's true that in the limit (by for example user manipulation) you can make your area lights or your mirrors to became almost a dirac function, but it does not mean that there are no ways to compute them in an unbiased and efficient way. They are very complex to compute, but who says all this is easy... in a near future it will not be problematic for Maxwell at all.

The well known sun problem, is an unfinished issue due that is "waiting" for the next sky model that is close to come. It will render OK.

Further on, the noise produced by Maxwell and the "biased" blurring of GI in V-Ray are both residual errors from solving the rendering equation. A noisy rendering is not any closer to the "real" smooth solution than a smooth, but blurred rendering. Replacing blurring by noise and vice versa simply changes how this error looks, but it is essentially the same thing.

Vlado, you know perfectly that error from noise and error from smoothing are completely different error metrics, that altough one can call them both error, it would be like putting in the same level the error in an inaccurate finite element subdivision for a fluid simulation or a radiosity, that the error to an analytical solution due to computers finite precision arithmetics.

This noise error will converge by its own to the solution of the render equation with no user intervention (no parameters) and the benefits are that there is no need for user settings. The smoothing error does not converge by its own and it needs the user to see and try again until he gets with something that he likes.

Best regards.

[j_petrucci]

Hi Vlado,

Nice to know you.

I wouldn't have answered normally since this is not the place for it, but since the comments were directed on me, I will say that I still hold to my original statement - caustics in mirrors or through glass are always going to be very difficult cases for Maxwell - if it stays unbiased. Look at the many posts where people complain that sunlight does not pass through glass windows... in some cases, this may be due to bugs, but in others it is the same limitation of "unbiasedness". Of course, eventually these effects will show up and will be noise-free - if you have the time to wait for them. Often, you won't have it, even if computers or Maxwell suddenly became 10 times faster.

It's not easy, but you know that is possible when dealing with no Dirac functions. It's true that in the limit (by for example user manipulation) you can make your area lights or your mirrors to became almost a dirac function, but it does not mean that there are no ways to compute them in an unbiased and efficient way. They are very complex to compute, but who says all this is easy... in a near future it will not be problematic for Maxwell at all.

The well known sun problem, is an unfinished issue due that is "waiting" for the next sky model that is close to come. It will render OK.

Further on, the noise produced by Maxwell and the "biased" blurring of GI in V-Ray are both residual errors from solving the rendering equation. A noisy rendering is not any closer to the "real" smooth solution than a smooth, but blurred rendering. Replacing blurring by noise and vice versa simply changes how this error looks, but it is essentially the same thing.

Vlado, you know perfectly that error from noise and error from smoothing are completely different error metrics, that altough one can call them both error, it would be like putting in the same level the error in an inaccurate finite element subdivision for a fluid simulation or a radiosity, that the error to an analytical solution due to computers finite precision arithmetics.

This noise error will converge by its own to the solution of the render equation with no user intervention (no parameters) and the benefits are that there is no need for user settings. The smoothing error does not converge by its own and it needs the user to see and try again until he gets with something that he likes.

Best regards.

amen.

[Maxer]

I didn't understand a word of that but it sounded really good!

[vkiuru]

Oh come on already. I doubt Vlado would post under such a clumsy nick. Also what PiledotNet said about that one statement, no way.

[tom]

So, this also proves "everything is possible"

[Renato Lemus]

and if you look at recent Maxwell renderings you will see that they have started to look almost like V-Ray renderings...

Best regards,
Vlado

I don't know, this doesn't look like a Vlado statement.

But if Vlado, the leader of V-Ray Render, really wrote this, unquestionable unprofessional position.

yeap, that's not fair. Maxwell is a very different approach. Also remember that vray introduced the "Progressive path tracing " method , inspired in maxwellrender, and because of vray users were begging for a method and a sun system similar to maxwell's.

[Aldaryn]

Oscar, those are some promising words...

Renato, believe me, it was neither NL, nor the Chaos guys who discovered the major rendering methods. But they do know how to make an existing bare theory into a kickass rendering enigine.

[Maximus3D]

Oo.. this is interesting! it makes me wonder where the guys from Cebas are on the forum too and the producers of mental ray. They surely are hiding on the forum as well..
But a very interesting discussion this one.....

/ Max

[Renato Lemus]

Oscar, those are some promising words...

Renato, believe me, it was neither NL, nor the Chaos guys who discovered the major rendering methods. But they do know how to make an existing bare theory into a kickass rendering enigine.

I agree,

Math rules!!

[mtripoli]

I'm going to take major "shtuff" for this question (and perhaps I should!) BUT, ok, get ready...

Does it matter? Here's why I say this: the "purists" will say Maxwell is unbiased and calculates x,y,z - no faking. HOWEVER, if the NL guys are clever enough to give an "unbiased" result that's been "flavored" to GIVE US THE IMAGE WE EXPECT, do we really need to question it. I'm speaking strictly about the resulting rendered image, not neccessarily "how it got there".

ON THE OTHER HAND, and I think I understand that this is why this question (and thread) was raised, from a purely mathematical point of view, well, if something ain't right, give'em hell...

Mike Tripoli

[iandavis]

anyone out there a able to present a more reasonable assumption... please do!

however, my understanding is that the blurring on a 100% transparent substance AND the darkening of light being removed or deflected from what SHOULD have been a perfectly paralell course of refraction/reflection is due to the resolving power of the rendering equation used. Simply, it's not accurate enough in one manner or another to resolve to 100%... in fact, it looks from the level of blurring and darkening in a clear substance that it's only at 90% or less. That means that there are a substantial (if not most) number of rays that will never get to 100%, or said another way; a large percentage of the light is inaccurately bent by a factor or +- 5% which would mean that any surface setting could not be predicted with an accuracy greater then 90%. The irregular rays will all diverge from the accurate solution creating a very smooth blur from the zero point to the limit of the maximum divergence from the solution, in this case +-5%

I'd be interested to know if this hypothisis is full of crap, or if that's what is happening.. anyone from NL? Shoot me fulla holes?

I wanted to add that I have faith that these issues will be dealt with. I trust that NL has no intention of releasing a final product which is incapable of recognizing and rendering accurately: glass, metal, wood, cement, etc. If mathimatical shortcuts must be taken to combat the amazing undertaking of a real-world light simulator and camera in order to iron out these "5%" errors... I trust something will be done, and I look forward to it. I enjoy using this product. I enjoy seeing it grow and being a member of the thousands of gardeners here.

[morbid angel]

imho path tracing as quasi monte cralo and as any other algorythm was inveted way before any render came out. So to say that one thing is inspired by the other is not to say anything.

[Mihai]

Best regards,

oh come on Filipe, why did you edit your post.....

[PRIAD]

But us, the users, should love and respect eachother too. If Vray adapts some of the principles of Maxwell and vice versa, it only helps us on both sides.

You are right Chris!!
Keep up the good sense