Maxwell 1.0

[Morbid Angel]

yes I agree about impact. The sky that other programs had before then were physicly not correct. There for cant be judged

[stochastic]

I think Morbid Angel is right, even if he's still drunk on monday mornings

Sure lots of the maxwell stuff is a rehash, like unbiased, mlt, bi directional, bsdf sky system, tone mapping, gamma, etc, but at the beginning, they put it together in a great way. Even the new material model is pretty damn cool. They sure did make an impact this last year, and alot of other renders apps have had to shake up the dust they've been accumulating.

Having said all that, at this point, the sum of maxwell at this point is a colossal train wreck. But it doesnt take away from what positive things they have contributed, and they have done alot, and may continue to do if they every figure out how to fix this thing.

I hope they dont give metroplis light transport / bi directional path tracing a bad name by all the inflated promises and market hype they did without living up to the promises, because mlt is still the highest quality way to go. Maybe Indigo is catching up, its using bi directional, I hope vray will follow up on it too....(nudge nudge wink wink Vlado)

[vlado]

I hope they dont give metroplis light transport / bi directional path tracing a bad name by all the inflated promises and market hype they did without living up to the promises, because mlt is still the highest quality way to go.

I've actually come to the conclusion that MLT is most definitely *not* the way to go. Sure, it can handle some specific "difficult" situations better than standard qmc sampling, but it performs worse for the majority of other situations which are not that "difficult".

If you think logically, for any given algorithm, you can always find another algorithm that knows more about the specific situation and gives better precision in less time. For example, if an algorithm knows in advance that caustics are going to appear in a certain place and will be caused by a certain object, it can calculate them faster than an algorithm that doesn't know it and must find them by trial and error. The absolutely best algorithm is the one that already knows the complete final result before calculating anything, and of course, if this was the case, we wouldn't need to calculate anything at all.

It follows then, that MLT is certainly not the best algorithm out there. Other reasons why Metropolis sampling is not a very good algorithm:

(*) It cannot use standard QMC sampling (this has been proven by theory). It needs pure random numbers. This means that for relatively easy situations (e.g. exterior daylight scenes) normal qmc sampling will be (a lot) better that MLT (less noise for the same calc. time).

(*) MLT is a local adaptive algorithm - meaning that when it finds a difficult place, it "stays" there for while trying to resolve the difficulty. Then it moves on, forgetting all about it, until it happens to come across the situation again. This requires very little memory, but obviously it will be worse than a "global" adaptive algorithm which doesn't forget that easily. Further on, MLT may get stuck in a "difficult" place, ignoring other parts of the result, which are potentially more important.

(*) The MLT algorithm has some parameters for which it is very difficult to find "good" values. These basically control how much time it spends in a difficult place, and by what step the algorithm moves around through the sample space. For some scenes, one set of parameters works well, while for other scenes, they give a worse result. Finding those values which for a given image produce the best results in the shortest times requires some test renders for the particular scene, which is what we are probably trying to avoid in the first place.

(*) This is probably least important, but strictly speaking, MLT is not exactly an unbiased algorithm. It is unbiased only in the limit when you have taken many many samples. Therefore the intermediate solutions may be somewhat different from the final result. This is typically manifested as the image changing its brightness as the calculation progresses.

(*) Last of all, MLT does not fit into "standard" rendering architectures - which is probably one of the reasons you don't see it used much. For most renderers, implementing MLT would require substantial redesign (possibly with an axe ) and will be at odds with the existing image rendering methods. More often than not, it is simply not worth it.

About the only positive thing about MLT is that it is indeed a very elegant algorithm. Unfortunately, it doesn't mean that it is a practical one. As mentioned above, there are other adaptive algorithms that may be expected to perform better on average.

Best regards,
Vlado

[studioDIM]

Sure lots of the maxwell stuff is a rehash, like unbiased, mlt, bi directional, bsdf sky system, tone mapping, gamma, etc, but at the beginning, they put it together in a great way. Even the new material model is pretty damn cool. They sure did make an impact this last year, and alot of other renders apps have had to shake up the dust they've been accumulating.

Having said all that, at this point, the sum of maxwell at this point is a colossal train wreck. But it doesnt take away from what positive things they have contributed, and they have done alot, and may continue to do if they every figure out how to fix this thing.

I'm at a loss for words: am i right interpreting these words as "sure the stuff was around, but they added package and hype, and i like that? Empty package or thereabout, sure, but nevertheless..."

Lightwave's Skytracer does beautiful physically correct, hdr skies.
From the dawn of cg.
Maxwell says they use "spectral" calculus (as in what, quantum sized information packets while things like vray use simple float numbers?I have to see it)and we all go "yeah, that's physically correct?".

If they said they used the ethereal technique that employed ultrafrabulators most people would have just said "woah, that's the way to go!" and jump on the bandwagon.
I honestly see that experiment as a marketing one, rather than a technical one which has to converge to a sellable product.
Indigo's doing a much better job at that, with a fraction of the BS...

I should just chill, as the poor guys made some effort, but man, i was sold as well.
It's the bitter at the back of the throat that stings, here

Lele

[trick]

...I think you guys are missing the point, it is a fact that Maxwell had a impact on V-Ray's developement.

The point is that I use software based on my clients demand. They have always been very satisfied with Lightscape, that have been even more satisfied since I used fRender and they were even more happy since I used VRay to provide them with beautiful images and animations in a relatively short period of time. As an other alternative I'm offering Maxwell output: until now I haven't seen any happy client after I tell them they have to wait a few days for the 300dpi A3 render only or a few years for a 2min. HD animation; even if this would be the most realistic output available. In 90% of my work lighting has to be (post)tuned, preferably in an interactive way, in ways that would not be possible in Maxwell. They are more interested in interactive concepting in Redviewer...while this viewer doesn't use GI/Radiosity at all. The fact that Maxwell had an impact on other renderers has more to do with emotions and the slow progress of current renderers and hardware then common sense...Unless you think people buying 30 and more licenses of a renderer in alpha-stage is a smart thing to do...

[Da_elf]

[quote="vlado"]

About the only positive thing about MLT is that it is indeed a very elegant algorithm. Unfortunately, it doesn't mean that it is a practical one. As mentioned above, there are other adaptive algorithms that may be expected to perform better on average.

kinda like saying that a ballerina is elegant but put her in a wrestling match and she lacks the power

[Paul Oblomov]

MAxwell - beautiful tests, maybe some of them are commercials (I mean you are paid for this)
VRay - just a "production ready" tool

BTW - Vlado, you said about PPT and MTl a lot. What about those Maxwell spectral, not integer color calculations ? Does in worth trying to implement in Vray ?

[vlado]

BTW - Vlado, you said about PPT and MTl a lot. What about those Maxwell spectral, not integer color calculations ? Does in worth trying to implement in Vray ?

Probably not at this time, since the effects are really subtle and the slowdown that will be added by the more-than-3 components for colors is probably not justified. For those cases where dispersion is required, it can be faked as needed. Of course, this doesn't mean that spectral calculations will not be considered at some point in the future.

For clarification, V-Ray does not use "integer" colors. It uses floating-point colors, which have three components (red/green/blue).

Best regards,
Vlado

[studioDIM]

About spectral calculations: i know nothing of the match involved, but knowing physics i can't help wondering what's the wavelenght resolution of a system that implements it.
It's all fine when a beam of white light creates a rainbow pattern of dispersed light, with the 7 hues.
But what if i split a very low frequency beam? Would the renderer be able to perceive the shadings between infrared and red, and the micrometric wavelenghts differences? To what degree of precision/recursion?
And most importantly of all, if that is the case, what's the real impact on most lighting situations of such a feat, when it's got to be converted to a visible-range RGB image?
Anyone who could point some whitepaper / info about this, even if not about maxwell, would be welcome...

Some images from the maxwell forums do exhibit a very high degree of realism, but i think it's mostly due to the lighting from the sky (exteriors) and the shadows quality (both exteriors and interiors/design).
The materials look good indeed, but i haven't seen anything yet that cannot be faked to render in a fraction of the time and look as good...

Lele

EDIT: Vlado posted as i was writing this... Like not said :P

[cpnichols]

Well I agree with Lele... and I have used this analogy before. "The Emperor has no cloths..."

On the other hand I can think of things 10 to 12 years ago. If someone told you there you were going to calculate Diffuse GI via raytracing, people would have laughed at you. But 5 years ago, the timing was right, and that is when everything hit. Vlado makes good points, and I always tend to agree with him, but I also talk to Marcos Fajardo at work (original creator of Arnold), and he also has some great points about unbiased rendering.

[joconnell]

If you see Marcoss any time soon give him a punch in the arm from me

[studioDIM]

"The Emperor has no cloths..."

Still laughing thinking of the Grimms' story :P

On the other hand I can think of things 10 to 12 years ago. If someone told you there you were going to calculate Diffuse GI via raytracing, people would have laughed at you. But 5 years ago, the timing was right, and that is when everything hit.

This is so very true...
I still toy with POVRay,a nd when i'm down on my (by now) crappy pc, i pull up one of the demo scenes i used to render with 1.x and watch it go.
What took 24 hours back then now is the blink of an eye.

The only issue i have with this is that someone thought of marketing it, with the results we all saw.
I'd rather follow indigo, where the author at least keeps it free (while at the present incarnation, engine wise is well in competition with the commercial one), than have to sift thrugh a billion hate posts to find anything useful.

Lele

[percydaman]

I hope they dont give metroplis light transport / bi directional path tracing a bad name by all the inflated promises and market hype they did without living up to the promises, because mlt is still the highest quality way to go.

I've actually come to the conclusion that MLT is most definitely *not* the way to go. Sure, it can handle some specific "difficult" situations better than standard qmc sampling, but it performs worse for the majority of other situations which are not that "difficult".

If you think logically, for any given algorithm, you can always find another algorithm that knows more about the specific situation and gives better precision in less time. For example, if an algorithm knows in advance that caustics are going to appear in a certain place and will be caused by a certain object, it can calculate them faster than an algorithm that doesn't know it and must find them by trial and error. The absolutely best algorithm is the one that already knows the complete final result before calculating anything, and of course, if this was the case, we wouldn't need to calculate anything at all.

It follows then, that MLT is certainly not the best algorithm out there. Other reasons why Metropolis sampling is not a very good algorithm:

(*) It cannot use standard QMC sampling (this has been proven by theory). It needs pure random numbers. This means that for relatively easy situations (e.g. exterior daylight scenes) normal qmc sampling will be (a lot) better that MLT (less noise for the same calc. time).

(*) MLT is a local adaptive algorithm - meaning that when it finds a difficult place, it "stays" there for while trying to resolve the difficulty. Then it moves on, forgetting all about it, until it happens to come across the situation again. This requires very little memory, but obviously it will be worse than a "global" adaptive algorithm which doesn't forget that easily. Further on, MLT may get stuck in a "difficult" place, ignoring other parts of the result, which are potentially more important.

(*) The MLT algorithm has some parameters for which it is very difficult to find "good" values. These basically control how much time it spends in a difficult place, and by what step the algorithm moves around through the sample space. For some scenes, one set of parameters works well, while for other scenes, they give a worse result. Finding those values which for a given image produce the best results in the shortest times requires some test renders for the particular scene, which is what we are probably trying to avoid in the first place.

(*) This is probably least important, but strictly speaking, MLT is not exactly an unbiased algorithm. It is unbiased only in the limit when you have taken many many samples. Therefore the intermediate solutions may be somewhat different from the final result. This is typically manifested as the image changing its brightness as the calculation progresses.

(*) Last of all, MLT does not fit into "standard" rendering architectures - which is probably one of the reasons you don't see it used much. For most renderers, implementing MLT would require substantial redesign (possibly with an axe ) and will be at odds with the existing image rendering methods. More often than not, it is simply not worth it.

About the only positive thing about MLT is that it is indeed a very elegant algorithm. Unfortunately, it doesn't mean that it is a practical one. As mentioned above, there are other adaptive algorithms that may be expected to perform better on average.

Best regards,
Vlado

I wonder if vlado sees algorithms like Neo sees the matrix.....

[Morbid Angel]

I dont doubt it !

[Da_elf]

give me a little more than a couple of white rum and sprites and ill see algorythms any way anyone wants me to see them. Wasnt that how you saw things on monday?