Maxwell Render | Official public forum

I did this test scene to see how fast I could set up a Final Render and Maxwell scene and how they compared to each other. The Final Render scene took me about an hour to get the correct lighting settings. It uses a Sky Light as the light source and a simple diffuse material. The Maxwell image took about 5 minutes to set up and uses Physical Sky as a light source and a diffuse material. Each image rendered for 36 minutes.


Final Render


Maxwell

There's not much point to this other than to say that when Maxwell does finally work I'm going to save my self a whole lot of time.

But if you look at the difference just quickly on that simple testscene, sure the FR one is finished quickly but it does look like hell. All blotchy, dark and it looks pale. Also the Maxwell rendering sucks just as much due to it's massive amounts of noise which is hard to get away from, it's however properly lit and has no big blotches and other strange patterns.

Comparisons like this isn't quite fair for either engine, now you see both of them producing crappy results which sure is fun to see but it says me nothing other than to maybe use the good old scanline renderengine again hehe

/ Max

Yea, neither one of these will win any beauty contests but I think it illustrates the potential Maxwell has. You can spend 5 or 10 minutes quickly throwing a scene together and create an image that is 10 times more realistic and accurate than any other render engine can produce. Adam is absolutely right, unless NL finds a way to squeeze 10 or 20 times the speed out of Maxwell cooperative rendering is the only answer.

We've subjected Maxwell to a number of pipeline tests...

We normally render a color/diffuse pass, a specular pass, a reflection pass, a shadow pass, an ambient occlusion pass, an ambient lighitng pass, and a motion vector pass. When you combine the total time of those renders plus hours and hours and hours of lighting setup and God knows how many hours setting up, tweaking and optimizing GI and FG, there's no question that Maxwell is actually faster than our current solution, and it's far more accurate.

It doesn't FEEL like it, and the scenario may scale differently for one-off renders, but for us, the real-world production numbers don't lie. We were as surprised as anyone.

_Mike

There´s one thing though I´ve been wanting to ask you mverta, won´t you still need all those passes rendered out separately? I mean if it turns out Maxwell can´t separate a reflection pass for example, is it still usable to you?

For me it would certainly still be useful, but I wonder what tv/film people think about this situation?

I support the idea of being able to do passes, because if it's technically possible, why limit people? You never know what might come up in production.

But it's important to remember that the reason we started using passes in the first place was to speed up the process of tweaking the composite. And the reason we have to tweak the composite so much is that we were faking and approximating reality to start with, and we couldn't get it quite right.

In my shop, we even use Channel Lighting, where you light your scene with lights colored pure R,G,and B and this allows you to adjust the shadows, density and fill in nearly real-time in the compositor. But again, we need to have all this fast tweaking power, because we can't trust the render to spit out truly accurate values for any of this.

In the end, I'd say 99% of the visual fx work I've done is trying to simulate, or integrate with photographed reality. And so far Maxwell gets us to that crucial 70%-of-the-way-there point almost instantly. And that's a tremendous time savings. Plus, we don't need to keep going back and forth looking at reference photos trying to guess what's going on with the light and making adjustments, or doing all the guesswork and test renders, or all that optimization work to stop the GI from being splotchy, etc.

If on top of Maxwell's accuracy we ALSO had layers, I guess that would be useful, but we really only needed them in the beginning because we didn't HAVE Maxwell's accuracy. So if you have layers, I guess you have the best of both worlds.

_Mike

Right now, the only things limiting us from putting Maxwell into real production are the shaders (soon to change), the motion blur (soon to change) and the noise (soon to change). But we're already planning to adopt it for some smaller projects beginning in January.

Interesting that you say you mostly use layers so you can quickly adjust an object to reality, I thought in many cases they would also be used for artistic purposes, for example someone wants less reflection on a particular object, or stronger blue shadows etc.....

I suppose they are used for that purpose also, but I guess then not as often as for matching to reality...

This simple test illustrates what I love so much about Maxwell. I don't want to sit around for hours figuring out the light settings in order to get the right look. Its challenging enough to get the materials to look good. With Maxwell, all of the light work is essentially figured out for me. Sure, it renders slow, but assuming that things speed up, this is gold!!

Thanks for the tests. Also you could try to render the Maxwells scene 55 min longer (the time of the Final Render setup). So, we could see, what solution we get after the same time, like in a 1:36 hour speed render contest.

Or what is, if you render the scene in double size + denoise + downscale. Most I get good a noisefree and sharp results on this way.

An other test could show the advantages of Maxwell. Make a noisefree image in Maxwell and render the scene with the same time in a biased renderer. My expirience is, that for difficult light scenes Maxwell can be faster than a biased renderer in same quality.

maxwell is beautifull but try to render an interior animation at a decent risolution.. with mental ray ( for instance) u can , in maxwell u can need ages ...
maxwell is the future. probably, but it need the processore of the future to work in an operative manner

cooperative rendering

in render what counts is ratio between money (time) and risults ...
i can buy the castle of edimburg perhaps and surely is better than my home(who knows).. but how much does it cost?
it is too easy to spend and incredible amount of money (time) to get a beqatifull risult ... u have to respect a reasonable ratio

giacob, you're recycling an age-old argument: You can produce inferior results faster with another engine. So what?

You can render for 11 billion hours with mental ray but it still can't do what Maxwell does, so what are we even talking about? I've been using mental ray forever and it's a great engine, but it's no Maxwell when it comes to realism. And frankly, I'm not going to miss all those hours trying to optimize mental ray so the GI and FG isn't blotchy. And you virtually can't use FG or GI in mental ray in an animation, anyway.

And if all that wasn't enough, Maxwell's only going to get faster; the core's not even been optimized yet, and processors are only getting faster. So if you want to tie yourself to yesterday's limited engines, knock yourself out, but the bell has already tolled for them.

I've said it before and I'll say it again: Maxwell represents the END of traditional rendering. It's over. It's like digital photography and film, which is like digital audio and analog tape: it's only a matter of time.

_Mike

P.S. How about you send your clients to me, as well; then you can render the work in mental ray and I'll use Maxwell. We'll see who makes the client happier at the deadline. Care to put your money where your mouth is? (Answer: Not if you want to keep your clients.)

mawell is going to be faster... everybody will be glad... up to now, at least for interior, there is not reasonable ratio time-risults ( that was the argoment not " You can produce inferior results faster with another engine" as u umproperly)simplified
mverta, please, do not act as the prophet of render
p.s i wonder if u would manidge to respect the deadline though, my dear

DELETED