Jump to content

Maxwell 1.0 Release is pushed back again


Devin Johnston
 Share

Recommended Posts

Holy xxxx... talk about cool old school! I want to be friends with you. Yost group huh? Or was it even pre Yost Group? I want to be friends with you.... I also want to know what you are doing now. A lot has happened since the mid 80s.

 

I just got involved in computer graphics while studying political science. You could do a university course in 'computer science' and I got addicted to programming. So bought a ZX-81 and wrote hangman. Switched to the amiga and had real fun. When the first ms-dos compatibles came out I bought one as it seemed handy for writing my thesis. Problem was that the graphics card was not completely compatible with the IBM-PC, but it had greater capabilities (more colours) if you adressed the hardware directly.

Couldn't live with Wordstar as I hated the lack of Wysiwyg. So I wrote a text editor with graphic capabilities: it showed proper italics and was blindingly fast. Next was a printer driver to get this in print and by that time my original study was way behind schedule.

 

Doing all this stuff just to write my thesis about the European Union made me even more interested in computer graphics. So I reinvented brezenham's algorithm searching for fast line drawing algorithm. Wrote a clumsy raytracer and fooled around with a fractal generator. Remember no internet in those days so no idea what other guys were doing.

 

Got hired by a company which wanted to produce a really fast renderer for arch viz work. We invented our own algorithms, got the first 386 from IBM 'to see if we could do something with the new instruction set', delivered the goods and sold lots of copies of Render* as it was the only renderer on the market fast enough to produce animations in a reasonable time. However the company had to sell out due to financial problems and the product became part of another application.

 

I then realized that I hadn't seen my wife and kid for two years and hardly remembered what a bed was so I forced myself out of programming. It was too addictive, so I started my own company as a knowledgeable reseller of 3d hard and software.

 

Still active in the 3D community as reseller of several products but mainly as consultant to companies who need advice on 3d.

 

To come back on topic my enthousiasm for 3d diminished over the years. As a professional I was still involved with it, but the feeling that you're on the brink of a revolution left me. Max is just a product since Gary jumped ship and most of the other app's give me the same feeling. We did some work on stereo 3D which was fun again, but even that wears out.

 

Maxwell is the first program in years which really get's me excited as it seems to break away from the 'cheat' approach to CG and it produces very, very convincing images. I realize it's too slow for most of you, but when the Yost group came up with the idea for a modular windows based program, people laughed about it. Most of the professionals didn't even want to try it when it was finally released after years of developing cause it was too slow. But hardware progression is the life saver of every revolutionary approach and I think history will repeat itself.

 

Now that was a long post and this is just my second one :) I'll stop babbling about the past.

Link to comment
Share on other sites

  • Replies 464
  • Created
  • Last Reply

Top Posters In This Topic

And to show I live in the now: a announcement made minutes ago on the maxwell site.

 

"New RC3 uploaded. The render engine and studio have been updated. There are still some issues in the render to be fixed in a next patch. The sky is coming very soon. The C4D plugin has been added.

 

CHANGELOG

 

(12-07-05) RC3

 

· Fixed: A lot of improvements in Render engine.

· Fixed: Preview in viewports more stable and with new features.

· Fixed: Selecting and moving the camera does not crash.

· Fixed: Render options properly saved.

· Fixed: Ortho view settings are not saved in layouts.

 

NEW FEATURES:

 

- A lot of new features in the interface of Maxwell Studio.

- Mxs files supports more than one camera.

- Textures in color channels in plugins now avalaible. THe rest of the textures will be added in next updates.

 

KNOWN ISSUES:

 

- Sky and textures disabled in current version. ( avalaibles in the next update ). Reflectance color texture now avalaible in bsdf's.

- Only custom emitters work in plugins. Wrong temperature mapping.

- Transmitance red color does not work properly."

 

I'm mainly interested in · Fixed: A lot of improvements in Render engine.

Link to comment
Share on other sites

So bought a ZX-81 and wrote hangman.

That was my first computer... had 8k of ram. I tried to write a fly simulator for it (instuments only) to land a plane. A while back I think I found the cassette tape which had all my programs on it. The computer is long gone. I think that 60 mins tape had aboiut 100 programs on it.

ZX81.jpg

 

Anyway... I sorta disagree with you about Maxwell... yes I am excited about the idea in concept, but by making it an unbias rendering engine you negate all the acceleration techniques and technology that have been developed over the last decade. Conceptually it is interesting and compelling, but in production is still has a LOT of issues. I have some very direct opinions on gi techniques that are basically all monte carlo based, such as lets say Maxwell or hmm... arnold.. I'm not really allowed to talk about them, I just can say that unless you use some bias or acceleration techniques, it just will cause problems in production no matter how fast your core raytracer is, or your hardware is. And I can say this working at a place with over 2000 processors on the farm.

Link to comment
Share on other sites

Anyway... I sorta disagree with you about Maxwell... yes I am excited about the idea in concept, but by making it an unbias rendering engine you negate all the acceleration techniques and technology that have been developed over the last decade. Conceptually it is interesting and compelling, but in production is still has a LOT of issues.

 

What do you think about the VRay Adaptive PPT? I read about somewhere but I don't know if it's available yet... but it looks like it has a lot of potential.

Link to comment
Share on other sites

What do you think about the VRay Adaptive PPT? I read about somewhere but I don't know if it's available yet... but it looks like it has a lot of potential.

 

Well it is out for the latest beta. Has been out for a while I think Vlado released it about a month after Maxwell was announced, but he had been talking about it for a while. It is cool and all... you can even make it unbiased... but it is slow so more of an interesting theory rather then something useful in production... hmmm.. that may be why I see Maxwell in the same way.

Link to comment
Share on other sites

That was my first computer... had 8k of ram. I tried to write a fly simulator for it (instuments only) to land a plane. A while back I think I found the cassette tape which had all my programs on it. The computer is long gone. I think that 60 mins tape had aboiut 100 programs on it.

ZX81.jpg

 

Anyway... I sorta disagree with you about Maxwell... yes I am excited about the idea in concept, but by making it an unbias rendering engine you negate all the acceleration techniques and technology that have been developed over the last decade. Conceptually it is interesting and compelling, but in production is still has a LOT of issues. I have some very direct opinions on gi techniques that are basically all monte carlo based, such as lets say Maxwell or hmm... arnold.. I'm not really allowed to talk about them, I just can say that unless you use some bias or acceleration techniques, it just will cause problems in production no matter how fast your core raytracer is, or your hardware is. And I can say this working at a place with over 2000 processors on the farm.

 

Hangman had to be done in the default 1K, but eventually I bought the 'wobble' block on your picture, but with 16K. After at least 12 hours without saving to my tape recorder it alway's wobbled and crashed.

 

Anyway, I know about production, advise companies how to build their render farms, and 2000 processors is a lot but there are some architectural firms around the world who have almost this capacity in the off hours. However I still believe in this new approach. The basic algorithm behind Maxwell is for some reason kept secret, but I believe it to be MLT based.

 

What is your basic objection against unbiased renderers? You say you are not allowed to discuss it? Why not. If we are talking about released stuff without NDA's we are free to discuss it or not?

 

Look the first raytracers where university born, badly implemented, by the book algorithms. They produced correct images at the price of rendertime.

They were followed by commercially produced smart implementations which cut corners every way the could. 95% of the images still looked cool and the render times went down to 10% or less of the original implementation.

 

Nowadays however most companies are back to a correct, no cheating, implementation of raytracing. They do so because the time saved is no longer interesting compared to the quality lossed. I assume you know renderman (or better PRman) very well and you know that they are hardly sacrificing quality over speed. They just follow the research papers about smarter implementations of the basic approach and code cleverly. Still main parts of PRman are essentially 'unbiased'.

 

(there is the possibility of a great confusion here: I don't mean they are unbiased like maxwell. I mean they follow their chosen paradigm as close as possible and don't cut corners)

 

Could you explain what's your fundamental objection?

Link to comment
Share on other sites

BTW, sorry if you just meant that you can't use it in your workflow. Without programmable custom shaders (Victor promised me these were in the planning), a better SDK and most important some sort of dedicated customer support it's unusable in any real advanced workflow.

Link to comment
Share on other sites

Well it is out for the latest beta. Has been out for a while I think Vlado released it about a month after Maxwell was announced, but he had been talking about it for a while. It is cool and all... you can even make it unbiased... but it is slow so more of an interesting theory rather then something useful in production... hmmm.. that may be why I see Maxwell in the same way.

 

But I thought I saw something about a newer version that was going to be adaptive - or is that what the PPT we already have is? Sorry, I'm confused and the latest beta I've been exposed to is probably pretty old. I've tried PPT and was very impressed with the quality, but it was even slower than Maxwell.

Link to comment
Share on other sites

But I thought I saw something about a newer version that was going to be adaptive - or is that what the PPT we already have is? Sorry, I'm confused and the latest beta I've been exposed to is probably pretty old. I've tried PPT and was very impressed with the quality, but it was even slower than Maxwell.

 

 

Oh yeah... right... Vlado mentioned that too... it is not in any public build yet... but because it is adaptive (which is cool), it will be more bias. I have not had a chance to play with it.

Link to comment
Share on other sites

Hangman had to be done in the default 1K, but eventually I bought the 'wobble' block on your picture, but with 16K. After at least 12 hours without saving to my tape recorder it alway's wobbled and crashed.

Anyway, I know about production, advise companies how to build their render farms, and 2000 processors is a lot but there are some architectural firms around the world who have almost this capacity in the off hours. However I still believe in this new approach. The basic algorithm behind Maxwell is for some reason kept secret, but I believe it to be MLT based.

What is your basic objection against unbiased renderers? You say you are not allowed to discuss it? Why not. If we are talking about released stuff without NDA's we are free to discuss it or not?

Look the first raytracers where university born, badly implemented, by the book algorithms. They produced correct images at the price of rendertime.

They were followed by commercially produced smart implementations which cut corners every way the could. 95% of the images still looked cool and the render times went down to 10% or less of the original implementation.

Nowadays however most companies are back to a correct, no cheating, implementation of raytracing. They do so because the time saved is no longer interesting compared to the quality lossed. I assume you know renderman (or better PRman) very well and you know that they are hardly sacrificing quality over speed. They just follow the research papers about smarter implementations of the basic approach and code cleverly. Still main parts of PRman are essentially 'unbiased'.

(there is the possibility of a great confusion here: I don't mean they are unbiased like maxwell. I mean they follow their chosen paradigm as close as possible and don't cut corners)

Could you explain what's your fundamental objection?

 

I know that some archfirm had the capability. I worked for Gensler. Having access to those computers and actually using them at night is two different stories.

 

The issues I can't talk about are issue with stuff that is in production at my current job. I would not talk about the details of it, but simply explain why I have a strong opinion on the subject.

 

My basic issue is that many clever things have been done to increase speed while sacrificing accuracy (not really quality since that is more subjective). For example, SSS was traditionally done with QMT methods. Took forever. Last week Jensen came by for a lecture at work and talked about how BSSRDF can produce 95% the accuracy of the unbiased QMT method in 1/50 to 1/100 of the time. My question is... why would you not use that? The same is true with things like Irradiance maps, different AA techniques, any adaptive, or caching method.

 

Renderman is very unbias in many ways, it is also essentially a scanline rendering engine. It is fast in some ways but very slow at doing anything like lets say GI.... which is why GI is never used in production. People freak out if you use a raytraced shadow in renderman because it can slow things down so much. Reminds me of using 3dsmax about 7 years ago.... same issues. BTW, I understand what you mean by unbiased in terms of prman. I have a love hate relationship with renderman.... well sorta 75% hate and 25% love, like an old girlfriend.

 

But as you already mentioned, Prman's custon shaders, RIB architecture is its real power. Everyone (vray, brazil, etc...) knows this, which is why Vlado was showing his version of the standalone vray which looked and acted a lot like prman with rib.... shader calls, read archieves... etc... you could even open them up in notepad and edit the shaders. awesome...

Link to comment
Share on other sites

hello dear fellow maxwell-victims,

 

first time i saw this thread and i just read it as a whole while having 1,5 beers. very drama, very amusing ideed ;)

 

i agree the quality maxwell delivers is really interesting, but i found that it can just break your neck in production (at least 2005). once tried with a client on a tiny project, but all those design-changes and 'i changed my mind and want to print a poster tomorrow' can get really bumpy with maxwell to say the least. i was one of those who bought the license very early, i think it was $300 at that time, and i take it as a both interesting and annoying tech-preview.

 

but my guess would be that the tech behind maxwell is disconnected from NL, and one of the other known players will implement something similar which is faster and reliable somewhen in the near/mid future. anyway, good for us.

 

have a nice evening,

aksel.

 

ps: hey chris, how are you doing?!

Link to comment
Share on other sites

BSSRDF can produce 95% the accuracy of the unbiased QMT method in 1/50 to 1/100 of the time. My question is... why would you not use that? The same is true with things like Irradiance maps, different AA techniques, any adaptive, or caching method.

 

Can't argue with that. Unless you are a hardass about accuracy just for the bragging rights. Lightscape was supposedly an accuracy-based GI-type program, I never used it as such. I put in bulb powers that would have caused an explosion. If it looks right, it is, as far as art is concerned. Science is something else. Is Maxwell meant to be a scientific tool or an artistic one?

Link to comment
Share on other sites

Can't argue with that. Unless you are a hardass about accuracy just for the bragging rights. Lightscape was supposedly an accuracy-based GI-type program, I never used it as such. I put in bulb powers that would have caused an explosion. If it looks right, it is, as far as art is concerned. Science is something else. Is Maxwell meant to be a scientific tool or an artistic one?

 

 

Thank you. I have been really lost on this "well it's more accurate" line of the arguement. Ok great but who cares? I mean really who cares? What is the client that needs this? I know there are some but that's not what I am doing. I am not doing a photometric plot of a site where I need to legally define the light within 1 candle. The developers I know say "I know I said I need it tomorrow but I need it today." This happened yesterday. My compitition is the guy who can DELIVER a good image. Not necessarily the best image. And really accuracy seems a bit misguided. Is maxwell more accurate than my plotter?

 

I have seen really nice images but I don't want that kind of stress.

Link to comment
Share on other sites

Can't argue with that. Unless you are a hardass about accuracy just for the bragging rights. Lightscape was supposedly an accuracy-based GI-type program, I never used it as such. I put in bulb powers that would have caused an explosion. If it looks right, it is, as far as art is concerned. Science is something else. Is Maxwell meant to be a scientific tool or an artistic one?

 

I ,for one, do not believe that it is for scientific purposes. At least not now. In order to due quantitive lighting analysis, one would need some false color and isolux utilities. All it has now, aside from its Maxwell Image Format, is just a picture output.

Link to comment
Share on other sites

I have been really lost on this "well it's more accurate" line of the arguement. Ok great but who cares? I mean really who cares?

 

This is my point of view exactly. If it isn't relatively speaking faster, why bother? ALL of the other solutions out there create really good results (opinion) and do it so that a deadline can be met (fact). This is exactly why the Renderdrives (or an 8-way server) are so compelling. Once you calculate your time into the equation, can you make your deadline, a worthy product, and the rent? :confused:

 

I guess the upside is that the other 4-5 rendering engines out there will have to add features and capability to stay competitive. That is a total upside to the market and the community. Maybe the upside is Autoscreet will have to embed a world-class renderer in the base product.

 

Post 343... Really guys, we (I) need to get a life!

Link to comment
Share on other sites

You know all of you make great points, I admit that Maxwell is way to slow in it's current form to do anyone who wants to use it in production any good. I think we are seeing just how desperate NL has become with the release of these RC's. I also agree that accuracy only for the sake of saying something is accurate is not the smartest thing to do, especially at the cost of speed. But this is really what makes Maxwell different than all of the other engines, and I think if they compromise on this point there is no way for them to survive. I'm sick of the whole drama of this thing and I do feel like I've been scammed, but the only thing I can do is grin and bare it and try and make the best of it until NL straitens it's self out. I really hope that because of Maxwell all of the other engines will be forced to make changes to keep up. It will only help the community out in the long run and in the end it really doesn’t matter which piece of software you use as long as you enjoy what you do and try your best.

Link to comment
Share on other sites

JUST ANNOUNCED:

 

DISCREET, a wholly owned subsidiary of Autodesk, Inc., acquires NEXT LIMIT Technologies, owners of the new world-class Maxwell Rendering engine for an undisclosed sum.**

 

http://usa.autodesk.com/adsk/servlet/pressindex?siteID=123112&id=5070129&linkID=5807594

 

** The above is NOT true, it is a joke, it is to make a point outside fact or reality. The respective names have nothing to do with the poster of this notice. (Insert standard lawyer disclaimer here!)...

 

>>>I get it guys. Follow the money. The reason they don't care about the user is base is that they DON'T CARE ABOUT THE USER BASE.

 

Revit got $140 Million U.S. for creating a new solution to a very old problem. Once Autoscreet figured out that it was truly a powerful techinical tool that needed work, they bought it, polished it up, and we will probably all be using it at some point.

 

The goal here is to create something cool and get bought out for Millions. I have $50 bucks that says by Christmas 2006 the above referenced joke WILL BE TRUE {It may not be Autoscreet, but someone else}. (This bet can only be redeemed once, by one person, first come first serve. The total liability is $50 to any and all winners. I will be looking at my email next Christmas to see who won!)

 

Sometimes it is the most obvious answer that is the correct answer. How much did Autodesk pay for Lightscape to bury (kill) it? If they really are 6 programmers in Spain this entire exercise will be worth it if the buy out is a mere $6M U.S.

Link to comment
Share on other sites

Can't argue with that. Unless you are a hardass about accuracy just for the bragging rights. Lightscape was supposedly an accuracy-based GI-type program, I never used it as such. I put in bulb powers that would have caused an explosion. If it looks right, it is, as far as art is concerned. Science is something else. Is Maxwell meant to be a scientific tool or an artistic one?

 

Actually Ernest... if you reread my post (it may not have been clear) I make the exact same point you do. The QMT method of doing a SSS will (eventually) converge to a 100% accurate result. The BSSRDF method will only be 95% accurate but will be done much much sooner. So my point is why not use the BSSRDF, the 5% lose in accuracy will be unnoticable (no one will know). What they will see is a noise free result in 2 mins instead of 200 mins. The challenge to scientists like Jensen, is how close to the QMC solution can you get and how much faster can you get. That ratio is the trick.

 

Lightscape was fairly accurate. It did have an adaptive meshing which would make it more bias (less unbias).

 

I, like you, am an artist and want to use to tool to represent the correct look, the physical sun and sky is cool, but I have a lot more control over the look of what I want to get if I use an HDRI and an area light. And if I want to crank that HDRI to get a lot of fill, I will do that. that is a long drawn out debate, and debate for the ages (of CG), and my in fact be the core of this Maxwell debate.

Link to comment
Share on other sites

It seems like this debate of accurate vs. "looks good to me" has been around a long time. As a practicing architect and a renderer (most of my work is rendering) I see the value of both. I was originally drawn into CG by the idea of being able to study how light behaves in particular unbuilt spaces. So, where the film effects house might say "let's add a fill light here" the architect might say "let's revise the layout of the skylights."

 

Both of these approaches are entirely legitimate, and simply reflect different goals. The goal of the effects house is to get the image to look good as efficiently as possible. The goal of the architect is to develop the best design possible. Of course, most of the time architectural renderings are not used to precisely study the behavior of light. And there are plenty of buldings that looked gorgeous as renderings, but atrocious when actually built.

 

Despite my vocal complaints about Next Limit's business practices and basic honesty, I do see a level of realism in Maxwell that I have not seen in CG before. This is really exciting to me as an architect. And as a renderer, I'd like to be able to offer this to clients who appreciate it. (I'm leaving aside for the moment the fact that I can't offer it now due to Maxwell's current state.) However, I think the quality vs. time tradeoff of Maxwell will limit its use to a niche market for the foreseeable future. I would never give up the flexibility of a "biased" solution. But the option for physical accuracy at this level is a new one, and I very much want to have it available to me.

 

Jack

Link to comment
Share on other sites

I might be wrong at this, but all these problems with Maxwell's instability, the bad RC and the delays took out all of the excitement that was created around it for the last few months. You see, people now seem to be taking that NL will simply not deliver what they promised in the near future. It's like that lil' boy who waits for Santa and gets a note saying he wasn't able to make it and will return sometime next year. Now, their customers don't believe them and (worse) are now questioning how necessary a tool like Maxwell really is (time/quality/ease of use/final product). This looks to me like Maxwell is going downhill, and going uphill is always a lot harder.

Link to comment
Share on other sites

Actually Ernest... if you reread my post (it may not have been clear) I make the exact same point you do.

 

Maybe it was I that was not clear--I was agreeing with your statement by saying 'can't argue with that'.

 

The accuracy of Maxwell is an important aspect of the alledged product. I like accuracy. But the real selling point to me was good looking, accurate looking (whether it is or is a cheat) program that was EASY TO SET UP. In Lightscape lights worked in a logical way. In Cinema, and most render engines, they do not. If I never see another self-shadow bias dial it will be too soon.

 

Maxwell seems to be going the wrong way at the moment--Keep It Simple, Stupid

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share


×
×
  • Create New...