Is Maxwell a viable option?

[DennisHolland]

Reviewing this thread, I think there's a pretty clear picture...

 

Maxwell is a viable option for production because:

 

- Testrendering is superfast (1 hour to see if you're on track with all settings)

- Quality is fantastic with a smooth learingcurve for a more seasoned archi-visualizer

- At least as stable as the wellknown others in the field

 

Keeping in mind that you'll need:

 

- A nice set of equipment behind it for rendering

- A couple of free hours to grasp the engine

- The organic/human/car models to replace the clipmap and RPC'S

 

The things to do for NL, soonest:

 

- Deal with the noise

- Take care of a few bugs here and there, fix the clipmap

- More control over materialsettings

- Hook up with an online renderfarm for deadlines' sake

 

Not bad for a starter in the field, in fact pretty fine. Like to see some suggestions/opinions/critiques/compliments....

 

Dennis.

[Ernest Burden III]

Dennis

 

Thank you for the clarity of your post. You really raised the debate above my "No, not yet".

 

I do not agree that Maxwell is ready for production in arch-vis for the simple reason that glass doesn't work right yet. We need good glass.

[DennisHolland]

Yes, you're absolutely right Ernest, forgot that tiny bit

 

I was somewhat optimistic but that is my nature. One can not assign it to a simple rectangle > 1 sided, backside blacked out and of course the noise. Must use the shell mod.

 

I also forgot the 4 cpu licence, wihich is costly when you want to render on multiple machines. To be honest, I was kinda hoping on some arguments and probably lots of other things I forgot, of which obviously are more than expected but that's life...

[Christopher Nichols]

- Testrendering is superfast (1 hour to see if you're on track with all settings)

 

Since when did one hour for a test frame mean fast? A test frame should be 1 to 5 mins. Is that just me?

[rivoli]

Is that just me?

 

no, 1 hour for a test rendering doesn't sound fast. at all.

anyway, for what is worth i always manage to get fairly decent test images in about 20 minutes (not fast, but 1 to 5 minutes using maxwell is something you have to forget). the preview really helps, even if it doesn't say anything about materials settings, it gives you a precise idea of what your lighting will look like within a minute.

[Ernest Burden III]

I think I know if I'm on the right track with Maxwell in a minte or so, thanks to the postagestamp preview. Its either right or it isn't. You don't need an hour to find out. At least not for the 'big picture'. Finer points, yes. Longer time to tell.

 

An hour to know if you are going in the right direction means you only have 5 or 6 tests in a day (add a meal, calls, other work matters). I may need to do that many tests by 10AM.

 

Since we're talking about it, I'm tired of all the time I pour into testing. Enough, already. Why do we (I) do so many tests? How about having a system (Maxwell could be that system) that just work like you expect it to so you can get on with your work and go home.

[DennisHolland]

Yes Ernest,

 

Testrendering usually is doing multiple tests (in my case). The initial settings can be examined quite well when doing a 20 minute test on 640x480 in Maxwell, maybe another one for finetuning and off you go with the big one. This could also mean two hours or thirty minutes, it was just an indication...

 

The essence of this is (superfast > 1 hr) that you can do the testrendering (read, multiple rendertests/finetuning etc) pretty fast. In this case I'd like to change the point discussed in: 'Maxwell is (or could be) capable of doing your tests pretty fast'. I hope this picture is (oh, how ironic) clear enough.

[ikerCLoN_]

We're working in my studio since Maxwell went commercial. We're working on animation, so part of our challenge is to get nice images with the smallest render time we can. I have to say that we have had A LOT of problems using Maxwell, not only with render times and noise, but with incompatibility between versions (materials from pre-comercial alpha didn't work with alpha, saved scenes in alpha version do not open well in beta...).

 

In addition, forget about 'layered render'. I mean, no shadow passes, no specular passes, no reflection passes... You have to render 'all in one', and so the render time increases.

 

But as you said, you can preview your final illumination in a very fast way... grainy, but fast And i?m not talking about one hour test: three or five minutes are enough in complex scenes (but not at final resolution). That's one of few advantages that, in my opinion, Maxwell has...

[Hazdaz]

I am NOT a Maxwell user (so you can ignore my post because of my lack of experience with this software if you want)... but I see some people saying "NO, NOT YET"... but from the render times I am seeing on some of the posts here (real renderings, not marketing hype by NL), I honestly can't say it will ever be a viable option.

.

I mean I just saw a rendering - that looks great (but still grainy), don't get me wrong - but it took 15 hours for one still. 15 HOURS! How much optimization can NL possibly do with Maxwell that could cut that time down. I wouldn't be looking for an improvement of 10 or even 20%... I would want that time reduced by 90% for it to be really viable. Now unless that rendering was done on a Commodore64, I really don't see how it could possible be sped up enough (through software optimization or faster computers) to be finished in under 2hours.

[Devin Johnston]

There was a significant speed increase from the alpha to the beta, between 2 and 4 times depending on the scene. The RC although totally screwed up was definitely much faster than the beta; don't know how much but at least twice as fast. Victor says in a "few days" (laugh) well see what the RC5 can do so it's probably better to wait until then before saying it's imposable.

[AdamT]

15 hrs. is actually fast for a Maxwell interior render. I've had renders that needed 72 hrs. on a dual Xeon 2.4Ghz. And they were not completely smooth yet!

[smeggy]

I usually get a decent notion of what it'll turn out like within 10 minutes Once you're familiar with how Maxwell progresses through a render it's not so hard to mentally extrapolate how it'll look after a short time.

 

I think even the Beta is production ready if you're familiar with the limitations and willing to work around them. Largely, 'production ready' is dependant on what you want to render. For my purposes I can get by on what Maxwell currently offers. It's by no means production ready for everyone though.

[Hazdaz]

15 hrs. is actually fast for a Maxwell interior render. I've had renders that needed 72 hrs. on a dual Xeon 2.4Ghz. And they were not completely smooth yet!

Dude... that is 3 days!!

There are just no words strong enough in the English language to describe how unacceptable that amount of time is for a still image.

.

You probably can't post the image, but if you can, I would LOVE to see what kind of rendering could possibly take that long.

.

(and what scares me even more is that I have a 2.4 Ghz dual processsor PC too)

[Devin Johnston]

Dude... that is 3 days!!

There are just no words strong enough in the English language to describe how unacceptable that amount of time is for a still image.

.

You probably can't post the image, but if you can, I would LOVE to see what kind of rendering could possibly take that long.

.

(and what scares me even more is that I have a 2.4 Ghz dual processsor PC too)

 

I don't think anyone is going to disagree that 3 days is a long time to wait for one still image. I've always been very disappointed in Maxwell's speed but everyone knows that speed is Maxwell's biggest problem. Remember that every rendering engine your comparing Maxwell to has been around for years and years, that's a lot of time for development. Ask your self was Vray or Final Render as fast when they fist came out as they are today? Anyway the method that Maxwell uses to calculate the GI is known for being slow, that's why it has never been used before.

[AdamT]

Dude... that is 3 days!!

There are just no words strong enough in the English language to describe how unacceptable that amount of time is for a still image.

.

You probably can't post the image, but if you can, I would LOVE to see what kind of rendering could possibly take that long.

.

(and what scares me even more is that I have a 2.4 Ghz dual processsor PC too)

If you check my gallery here: http://www.maxwellrender.com/forum/viewtopic.php?t=2208&highlight=gallery

 

The Ackerberg house ran for at least 3 days and still needed noise reduction. The diamond ring on the box also required 3-4 days.

[Hazdaz]

ADAMT - I'd love to see your renderings, but I am not a member on that site - won't let me in.

[AdamT]

ADAMT - I'd love to see your renderings, but I am not a member on that site - won't let me in.

Ah, sorry, I didn't realize you had to be registered.

 

Here are a few of the ones that took an egregiously long time to render:

 

http://www.3danvil.com/Maxwell/mxs003_ps_sm.jpg

http://www.3danvil.com/Ring_Redux.jpg

http://www.3danvil.com/Maxwell/Ackerberg_House_sm.jpg

[Hazdaz]

See those looks great (I espesially like the last one), but I think those should take no more than 2-3 hours, not 2-3 days. There is no way any amount of optimazation is going to cut that time by 24.

.

For instance, I rendered this image years ago (using just scanline) and I can't imagine it have took moer than an hour. (I actually might try to find teh file and see how long it will take on my PC now just for the heck of it).

Now your image obviously has a lot more detail, but still, I don't think 72 times the detail. (obviously I am not trying to compare the images themselves, just talking about the render times)

[Sawyer]

This is rather annoying I would be more inclined to purchase MWR now than 6 months ago. YEAH go ahead I will buy that bridge!! SURE I will buy that underwater island!! At first they seemed so smug it bugged the crap out of me now I want it! Or at least I want it to work. AdamT those are sweet!

 

Now this is just an aside comment but something I have thought about. The real striking quality of MWR is the sunlight/sky. This seems to be the next big thing. Saw it in C4D (what was I thinking not getting that?) with STRATS latest and vray has it in the works. There is a large range of color in the sunlight Adam in the Ackberg house the ambient features look to me simular to almost any current GI package. The direct light has a more dynamic quality. What I haven't really seen is a real sky system with clouds and diffuse sun. If the client wanted a cloudy day or dynamic sunset I assume you need to go back to picking a jpg image that comes close to your sunlight for a bg. And possibly ruining the effect.

[Ernest Burden III]

The real striking quality of MWR is the sunlight/sky. This seems to be the next big thing. Saw it in C4D (what was I thinking not getting that?) with STRATS latest and vray has it in the works.

 

The C4D version is nice--although i found a bug where it goes horribly wrong, maybe that's fixed in the latest update, which doesn't work well with the new Finalrender--which has a great sun/sky. But Lightscape had it in the mid 90's.

 

You can do the same thing very easily with IBL, but the SKY function in C4D also makes clouds, stars, its easy.

[Devin Johnston]

Maxwell can also use HDR images for illumination so you don't have to rely solely on the Physical Sky.

[AdamT]

See those looks great (I espesially like the last one), but I think those should take no more than 2-3 hours, not 2-3 days. There is no way any amount of optimazation is going to cut that time by 24.

.

For instance, I rendered this image years ago (using just scanline) and I can't imagine it have took moer than an hour. (I actually might try to find teh file and see how long it will take on my PC now just for the heck of it).

Now your image obviously has a lot more detail, but still, I don't think 72 times the detail. (obviously I am not trying to compare the images themselves, just talking about the render times)

 

Your image looks nice, but to be fair, it doesn't even have caustics, let alone dispersion. You could get a much faster render time in Maxwell if you disabled those effects--but it would still be slow (maybe 10 hrs?).

[manta]

I think we're all losing sight of whats important here, sure it would be great to get photographic results in a reasonable amount of time, but do we even need that...does the client...probably not.

 

We all know clients change their minds all the time, right up till the last minute,

this rendering took 30 minutes, and is more than adequate to understand how the space will look, thats my opinion anyway, I think we are pushing the envelope too far, just because we can, not because its nessesary...

[AdamT]

I kind of see Maxwell as a beauty-shot-only engine, when the client needs a render for a billboard or promotional material. It's way too slow for day-to-day stuff. Of course it would also be useful if you really want/need an accurate light simulation.

[Hazdaz]

I kind of see Maxwell as a beauty-shot-only engine, when the client needs a render for a billboard or promotional material. It's way too slow for day-to-day stuff. Of course it would also be useful if you really want/need an accurate light simulation.

But wouldn't you then need to have 2 lighting set-ups, 2 sets of material... (I am making the assumption that MAXWELL needs it's own lights and it's own materials)

You couldn't just take your "regular" rendering set-up (that you do your production renderings using another renderer) and then drop it right into MAXWELL to do the billboard stuff, right?