[victor]

Dear friends, it's time to provide the latest information about the status of Maxwell.

First of all, thank you all for your patience. We have been working hard in the last two weeks preparing a stable RC5 version. Our first priority has been to assure the stability and quality of the render engine and, second, integrate with the new materials. Meanwhile, Maxwell Studio has been improved.

We have accomplished these milestones successfully. Our goal was to release RC5 this week, unfortunately we finished the last changes yesterday and didn't want to make a mistake again releasing too soon. There are also some issues that we want to work out before.

We are working very closely to the Maxwell "A-Team" (testing team) which is now taking more responsibility on the testing process and quality assurance. We are now very cautious about releasing something without the A-Team green light.

To summarize, we are very confident about a RC5 release next week, including the Windows and Macintosh versions. I will invite the A-Team members to disclose some information about this progress.

I wish everybody a happy new year 2006 (the "Maxwell strikes back" year)

Victor

[tom]

Hi Maxwellians!

We really tried to provide you RC5 by year's end, and while we could release it as is, we still see a few important issues that need addressing so you're not disappointed again. I understand your frustration over the time frame, but neither myself, as leader of the beta team, nor our very talented and experienced team members from around the world would feel good releasing software that couldn't pass our qualification tests. You deserve the best, and we're going to give you that in early 2006. As always, thanks for your support and your constructive crticisms. Keep the faith!

Happy, healthy and prosperous New Year!

Tom

[Thomas An.]

Further remarks from the A-Team (in addition to what Tom, the team leader, mentioned):

As Victor mentioned, and in view of past problems with the RCs, the A-Team is taking more responsibility for quality control.

In regards to the Dec 22 date:
The developers indeed proposed a release at that time, but the Maxwell testing group resisted the release of RC5 as it did not pass some of our quality control tests. Furthermore, the Maxwell Testing group intents to keep resisting the release of RC5 until it is sufficiently ready for broader use. We are currently working with Cornell boxes and dielectric tests to further calibrate the engine GI and caustics. The goal is to sufficiently calibrate the translation interface between the new material system and the engine core.

In regards to the Dec30th date:
The developers optimistically did issue a new nightly on the 30th, but we need to test further before giving it a green light (or not) and depending on our findings the testing process may repeat.

[victor]

-----------------------------------
ATeam's member Mike Verta
-----------------------------------

Fellow Maxwell Users:

I doubt you guys will be interested in any specifics about the upcoming RC5, but I'll give you some anyway.

First off: Relax. It's way better. Now take a deep breath and keep reading.

A little history for you, so you're clear: The Beta Version had a core engine and a set of materials we were using. We all produced (still do) some seriously cool work with it. But as many of you know, there were some things it had real problems with, that would show up whenever glass materials were used, etc. In some cases it seemed like it would never solve noise or caustics issues, etc.

The current engine is an improved engine, but it also uses an entirely new material set, that solves a ton of the Beta's issues, but also meant basically starting over. So on the one hand, we had to take many steps back, so we could then take many steps forward, past Beta. And this is where we'll be at Maxwell 1.0: Way past Beta. The RC process is about this journey, step by step until we finally reach 1.0.

So RC 5 solves a lot of problems, and has a lot of improvements, but we need to be clear that it's not 1.0, yet. A lot is coming! So here's what we're seeing right now:

Fireflies and Dots. Looks like we've nailed them. I haven't seen them creep into a render for some time.

Caustics, etc. The caustic rendering is vastly improved from Beta. Some very beautiful stuff there. The caustics are more present, show up sooner, and are faster. Very cool.

Speed Can't really tell, yet. It's too early. Some things are faster, but the optimizations won't happen 'til the end, so I wouldn't go looking for huge decreases in time yet.

Studio Huge changes and a list of fixes about a mile long. We've been really focusing on the material editor and how it's going to work, since this will be around in a lot of incarnations, for plugins, etc. Some of our fiercest debates are about the GUI and it's getting there. Generally speaking, it's pretty stable, everything's turned on, etc. Feels like we're hunting down a lot of annoying bugs more than core functionality issues, but again, since it's RC5, a lot of its more advanced features aren't implemented yet. We're just trying to make the RC as stable as it can be with what it does have.

Materials This may change before RC 5 is released, but I'll tell you a few things about the material system.

Textures are enabled, and working, still hunting down some bump issues, and the light distribution model for "low-roughness" to "high-roughness" looks beautiful. Currently, there's no sub-surface scattering, but that's in the pipeline for a future release. Right now, there are a couple of ways of doing things like shiny plastics: for example, you can use 2 BDSF layers, blended with the weighmap function, or you can use a single layer BDSF with a Coating layer. They've both got drawbacks, and they're both temporary solutions. The real goal is a truly "layered shader" approach, where each layer has its own alpha channel for blending with other layers, and you can change the "outer-to-inner" stacking order for different looks. Most of us do this in our current packages anyway, it's just that Maxwell is SO physically accurate, that a lot of other stuff has to be taken into account to get it to perform these cheats in a "real" way. It's do-able; the math is dizzying. But it's a 1.0 thing, not an RC 5 thing.

Noise Noise is very closely tied to the material set, so with the material improvements, noise is greatly reduced - or even possible to solve for - in ways Beta couldn't.

General Thoughts

I'm trying to imagine what it'll be like for you when you first get your hands on RC 5. I think if you're a Beta user, you'll generally feel you've stepped up, especially if you do caustic stuff a lot. But with the right set of project needs, you might fall right into the places where RC5 is still "RC" and not "1.0". It really depends on the project. I'll bet that the ratio of people who stick to Beta versus RC5'ers will be about 50/50. Just a guess. There's a couple of import/export and facetting issues that could really change that balance, and since they're in progress, I can't say for sure.

But the RC is so much better in a couple of places, that you might find yourself where I am, which is basically not able to go back. I've already taken some of its new power for granted, and I don't want to feel like I have my hands tied.

You will absolutely not be able to miss the tremendous improvments in RC5 over RC1-4. Some of you will view it as Beta-superior. Some of you will hate it even if God delivers it to you on a shaft of golden light, because you're beyond frustrated, and maybe rightfully so.

What I'm hoping though, is that working with it, you'll have the same feeling I do: even though some things aren't turned on, or polished, you can actually see the finish line. You can see how it's going to work, and you can tell that it's better. For me, that brings with it a lot of confidence and patience. Once I know things are moving forward, actually forward day by day, I'm cool with the journey. It's the not knowing that drives us crazy. .

I hope this has been of some help to you guys; I know it's not much, and perhaps if you have some other specific questions, you can ask them and I'll see about getting permission to answer.

Keep in mind, though, that I'm a tester, and my first job should be to keep on the testing-line and get more features to you sooner. Other A-Team guys may be able to chime in occaisionally during renders

About screenshots and renders. We set up scenes deliberately to break Maxwell and find its flaws. And we only let them run long enough to see the issues. Lots of times, we'll render little 200x200 thumbnail versions because it's enough to see, and we can already tell what's going on at Sample Level 2, when it's mostly noise. Lord, I can only imagine what would be said if we posted those! We just don't want to be sitting through 16 hour renders of things we know are going to work... that's what you get to do! So while we have a number of rendered images, they're usually in really rough shape, and we'd have to release them with a paragraph of explanation. Plus they're not sexy. When this is over, I'm going to offer the world's worst coffee-table book: 6000 renders of Cornell Boxes with Various Unpleasant Artifacts.

Well, okay.. here's one good one:


Maxwell's light distribution at work.



Thanks, and Happy New Year,

_Mike Verta

[tom]

6h45m @ 1024 x 1024 @ P4 2.8HT

[tom]

[tom]

[tom]

Courtesy of Benjamin Brosdau [Tora_2097 / A-Team]

[tom]

DO NOT TRY THIS AT HOME WITH BETA

RC5.7 // 14h17m : SL 15.64 : 1 emitter : ISO 100 : ss 40 : fstop 8 @ 900 x 1200 @ P4 2.8HT

Full Indirect GI with 1 emitter / 1 lambert reflector
Single Material with no maps.
3 basic BSDF layer (lambertian black, violet glaze, sandblasted amber glass) + 50nm coating layer
550K-poly scanned Ajax model by Torolf (jotero.com)

[tom]

A few material tests, sorry about the misplaced ribbon, I didn't have time to finish the rest of the scene. All pics rendered for 30 min at 640x480.




In this pic the golden ones are 3 bsdf layers, all with slightly different gold color and roughness. This way you can get lots of subtle specular variation, useful for many materials, not least car paint. The one in focus has a glass material and if you look carefully, something that many have wanted to do with glass for a long time

Not so visible, so here is a closeup:







This one shows anisotropic reflections on the violet/red ones and on the gold one you can see two bsdf layers, one has a weightmap applied (the white one). The mixing of layers to create one material not only gives you great variation, but also makes texturing a lot easier. Example with this material:

If I wanted to make something like this previously in Maxwell (and other renderers), I would first have had to create a color map with the golden/white parts in it, then a corresponding bump and specular map. I would then have to adjust each map individually until I was satisfied the reflection and bumps were ok. What if I then wanted to change the shape from the cat (yes it's a cat) to something else? I would have to redo the color, bump, and spec maps...

Here, I have one bsdf layer for the gold, and one layer for the cats, with the weightmap of the cat shape applied. The bump is just a noisemap applied to this layer and no shapes necessary in the bumpmap since the influence of the layer is driven by the weightmap (the bump will only show on the cats layer). So if I want to change the shape, I just change one map, the weightmap. It is also much faster to make adjustments, since I can control the gold and white material separately, not dependant on any spec,refl, or colormaps.


This last one shows again glass with etchings on it. Very simple and quick to set up this material. Two layers, again one has a weightmap.