PDA

View Full Version : Maxwell Render V1.0 Images


Continuumx
04-28-2006, 03:55 AM
The following test was done to see how Maxwell Render handles submerged objects. This is a true test of nondestructive attenuation of submerged objects.

Normally in these kind of renders, you boolean out the volume of the submerged object so that you get a render that looks like one object is within another object like an ice cube and glass of water. The render time was very good for this render. The green glass object in the background is dense glass material. The object in the foreground is the Architectural Glass solution which is an unbiased solution for getting really nice Architectural glass in a short amount of render time. It is not a hack, or shortcut but obeys the rules of unbiased rendering in that what we have is an object with Nd of 1.0 low reflectance and a BSDF layer with higher Nd for reflection. Maxwell Render 1.0 uses real scalable system of Nd.

Once again the render time for this image was very acceptable and the quality is what I was after in this test. You can notice the subtle color bleeding of the submerged object to AGS sphere.

http://img131.imageshack.us/img131/380/agstransmissive3nc.jpg

The next example is of the new Emixer feature of Maxwell Render V1.0. This is a really neat paradigm shift in rendering as there is now an equivalent to mixing for rendering. The emixer is a tool to do serious lightmixing via the Emixer or lightpad as I call it. You can get into a emix session where you set up your render, and then start your render. Maxwell Render V1.0 allows interactive adjustment of lightsources while you render!

So now in 2006 there is this paradigm shift in rendering technology where you can get a number of renderings for just one render. To put it another way, you can now render a complete sequence of Day and Night images all from the time frame it took to render one image in Maxwell Render V1.0.

To put it another way, in this one sequence, I did an example that spans 600 rendered sequenced images all from one render using Maxwell Render. Now that is a time savings that I can appreciate. Where else can you generate an unlimited number of additional live renderings from just one rendered event?

You can view the movie here, scroll down and look for the "Future City Demo" while you are there be sure to check out the other emixer videos as they are all very good to see this feature at work:
http://www.maxwellrender.com/announcement.htm

Now all I did was take these sequences into Sony Vegas, added some titling, did some transitions and added a custom sound track to get you through the footage.
http://fastspace-arch.com/MaxwellRender/share/FutureCity_emitter_strip.jpg

The next image is the newest feature of Maxwell Render V1.0 and that is real camera glare. This feature can work during a render. It is interactive so you can adjust, and change the look all while Maxwell Render does it work. You can also experiement after the render has finished.

http://fastspace-arch.com/MaxwellRender/share/Future-City-Glare_MTM.jpg

Now the real work with Maxwell Render V1.0 begins, I will post more images as I have some concept ideas to do that I can now achieve that used to be done in physical model form.

feel3d
04-28-2006, 03:00 PM
Sorry, but i don't see any?

moka.studio
04-28-2006, 03:12 PM
you need to click on the little white dot :). right?

your site looks quite interesting Continuumx, though it is a bit hard to navigate through, and to go back to original pages.

jddog
04-28-2006, 03:15 PM
not every ones have the signature activated btw

e[dub]
04-28-2006, 03:25 PM
if an image tells a thousand words, this thread tells me nothing.

Srek
04-28-2006, 03:36 PM
Renders from Version 1.
Hy Tyrone,
please try to make your post more readable. The only thing that can be seen is the text i quoted and a hard to find link to your website in the signature.
Thanks
Björn

ooo
04-28-2006, 03:38 PM
']if an image tells a thousand words, this thread tells me nothing.

I think these promised images are still rendering ;)

Continuumx
04-28-2006, 05:57 PM
Hy Tyrone,
please try to make your post more readable. The only thing that can be seen is the text i quoted and a hard to find link to your website in the signature.
Thanks
Björn

Sorry Srek, I accidently erased the images. I will get them back sorry all!

moka.studio
04-28-2006, 06:00 PM
are you trying to pull a 'Maxwell' on us? :)

williamsburroughs
04-28-2006, 06:18 PM
He had a couple images up last night. The one with the spheres looked nice.

Continuumx
04-29-2006, 02:51 AM
Images reposted!

moka.studio
04-29-2006, 07:36 AM
Thanks for sharing, the exemples look quite nice.
What are the render times, and on what kind of system?
Keep posting your experiments with Maxwell, I am very interested in seeing more.
jp

avocade
04-29-2006, 02:14 PM
I remember reading of the new lighting system at Pixar that they showed off at SIGGRAPH last year. They've used it for The Incredibles and the upcoming Cars to dynamically adjust and test the lighting. Could you elaborate on what this new "emixer" feature means practically, and perhaps compare it to Pixar's lpics?

http://www.vidimce.org/publications/lpics/

AdamT
04-29-2006, 02:47 PM
Looks good Tyrone. As far as AGS being unbiased, I can't say I'm buying that. Is there a real-world example of a material that has different ND numbers for refraction and reflection? Regarding render times, can you tell us what they were specifically? "Very good" in the Maxwell dimension can be "unacceptably slow" in the rest of the time space continuum. :)

ooo
04-29-2006, 03:22 PM
Very nice renders indeed! Thanks for posting!

odo

BeccoUK
04-29-2006, 03:35 PM
Nice tests with a difficult program. At the outset Next Limit always maintained that Maxwell would be physically accurate. Their Architectural Glass Solution (AGS) is a bodge job by Next Limit because they still can't get sunlight to go through glass. Since Beta came to a standstill last July (2005) Maxwell has steadily declined.

Also, it may be worth mentioning that the post render light mixer feature will not work with Maxwell's glare. Instant crash if you try to apply glare.

I would like to see some nice 'diamond' images now that dispersion is working again. Acurate dispersion was one of Maxwells' original strengths not all the new fangled (and partly working)cheating that can be done with most other render engines.

AdamT
04-29-2006, 03:58 PM
As I understand it the glare feature only works in the preview render; is that correct?

Continuumx
04-29-2006, 04:32 PM
Looks good Tyrone. As far as AGS being unbiased, I can't say I'm buying that. Is there a real-world example of a material that has different ND numbers for refraction and reflection? Regarding render times, can you tell us what they were specifically? "Very good" in the Maxwell dimension can be "unacceptably slow" in the rest of the time space continuum. :)

Hello Adam,

The AGS glass is unbiased in that it is possible for a material to have this kind of Nd values, with Maxwell this material in particular can be produced but it would required a controlled environment. In this material there is no refraction, there is only reflection and transmittance. I will do some research to see if there is any information about a material that has these physical properties.

The time frames, in the attenuation test, the image was good (clean) in about 4 hours. The AGS material is very fast, there is literally no wait for noise to clear up, and it allows for caustics. The emixer example was allowed to run for about 24 hours because I wanted a very clean render. A satisfactory result could be reach in less than 12 hours. The generation of additional 600 frames took less than a minute.

Continuumx
04-29-2006, 04:39 PM
Nice tests with a difficult program. At the outset Next Limit always maintained that Maxwell would be physically accurate. Their Architectural Glass Solution (AGS) is a bodge job by Next Limit because they still can't get sunlight to go through glass. Since Beta came to a standstill last July (2005) Maxwell has steadily declined.

Also, it may be worth mentioning that the post render light mixer feature will not work with Maxwell's glare. Instant crash if you try to apply glare.

I would like to see some nice 'diamond' images now that dispersion is working again. Acurate dispersion was one of Maxwells' original strengths not all the new fangled (and partly working)cheating that can be done with most other render engines.

Becco, AGS is not a botched-job created by Next Limit, it is a real working and tested solution, and it allows sunlight through architectural glass which is by definition a very thin and lightweight glass with very little or no refraction.

Dispersion works, it has been tested as well. Check the website as there is an example of dispersion for V1.0.

There have been tests (many) conducted to compare. It was very easy to see how the material quality is much better. There are some images in the laboratory that directly compare the images produced with the Beta to the V1. I can tell you that V1 exceeds the Beta. The 'beta experience' is no longer something to think that you cannot get and exceed in V1.

Continuumx
04-29-2006, 04:40 PM
As I understand it the glare feature only works in the preview render; is that correct?

Adam, the glare feature works in the larger view. You have to hit the refresh button, or adjust iso, shutter a small amount to get it to refresh the larger image.

BeccoUK
05-01-2006, 04:15 PM
Whichever way AGS is looked at it is a 'fake' material introduced at the last minute because Next Limit couldn't get sunlight to go through glass correctly.

Anyway after it was confirmed dispersion is working I decided to give this a test with a simple Cinema scene on my Pentium 4 based system. The 500*500 (small) render has been chugging happily away now for more than 15 hours and still lots of dispersion related noise exists throughout the scene, with dispersed caustics only recently starting to noticably resolve on the ground plane.

Maybe a further 15+ hours and it will be nice. That's my definition of Maxwell reality, not AGS.

AdamT
05-01-2006, 04:26 PM
Hello Adam,

The AGS glass is unbiased in that it is possible for a material to have this kind of Nd values, with Maxwell this material in particular can be produced but it would required a controlled environment. In this material there is no refraction, there is only reflection and transmittance. I will do some research to see if there is any information about a material that has these physical properties.
Any luck with that? :)

rob rhodes
05-02-2006, 11:33 AM
I actually think AGS works well and was so relieved when I saw they had a glass solution in V.1. I know it is obviously an expected feature in a modern render engine but Im please they found a way to get a glass solution without the render time hit. If they had included a physicly accurate glass shader that took twice as long to render as well I would deffinately still be using AGS as I dont really care if its a fake as long as it looks good!

As for animation I don't think this render engine (RS1) could really be used to produce that many frames needed for animation - it takes ages for 1 so people would just be saying it is unusable anyway! Was it in the beta? Thinking back im sure i've seen some animation in earlier times.

The plugin is actually better than id hoped! My expectations had got so low i guess. But i just use the external material editor to create my mats. There are issues with sweep caps and sometimes errors saying not enough space on the polygon or something which is annoying. I was supprised to see that phong tags were supported though. But in general its quite useable IMO.

seco7
05-02-2006, 01:11 PM
Sorry for thread jacking, but I'm trying to spend as little time as possible at the MW forums ...

I have been getting an "oily" look with AGS from certain angles. Has anyone else run into this or has anyone discovered a workaround? I'm not at work now, but I'll try and post an image later.

Thanks,

Steve

lllab
05-02-2006, 05:15 PM
seco,
play with the roughness and also with the strenght and falloff of the reflection layer, i use about 3-5% roughness.

cheers
stefan

seco7
05-02-2006, 06:32 PM
Thanks Stefan, I'll give it a try.

Continuumx
05-03-2006, 03:02 AM
Some of my images are not able to be shown because of webhosting problems. I will get them back up once this has passed.

For the meantime, here is an SSS test that actually turned into something quite good. Pardon the SSS objects, I plan to put something else in their place.

Enjoy! Sorry for the image size! This one was ready within 3 hours at this size (no noise, and very smooth color.) I am allowing it to continue rendering because I want to reach a certain sample level. But this is a good test of SSS as they are quite a number of them and this kind of thing goes a long way to see how the algorithms hold up.

http://img142.imageshack.us/img142/8427/v1ssslarge5ec.jpg

Ernest Burden
05-03-2006, 03:42 AM
here is an SSS test that actually turned into something quite good

It sure did! Looks great!

Are you doing this with the C4D plugin or Studio?

AdamT
05-03-2006, 04:02 AM
Nice one! I've heard mixed reports about SSS but haven't tried it yet myself. Glad to see those render times.

Matariki
05-03-2006, 04:39 AM
Tyrone,
that look very good. Can you give us an indication what sample level was reached? I have a rather simple scene and I had it in the oven for 5 hours and I still got considerable noise in the shadows. I have noticed that the shadow areas are usually noisy as very little light comes from these and it takes considerable longer to calculate. This is the unbiased renderer for you. Do you know if NL has considered to to weight the render effort in order to give problem materials such as glass etc and low light areas more CPU time? I know that this could stuff up the noise distribution function and that might result in noticeable artifacts in the final render but it would be interesting to know.

AdamT
05-03-2006, 04:54 AM
That looks really good Matariki. I think the noise is fine enough that NeatImage or Noise Ninja would do a good job.

Regarding weighting, I seriously doubt that NL would consider it. They're very tied up in the whole unbiased thing.

Continuumx
05-03-2006, 04:56 AM
It sure did! Looks great!

Are you doing this with the C4D plugin or Studio?

Ernest,

This scene was composed in C4D. I set the mapping in C4D and the materials. I then rendered out the mxs to Studio because I needed to work on the SSS material and fine tune the ground material. I am using 2 BSDF layers for the ground layer to get that 'wet look'.

I did make test renders out of the plugin to see where I stand.

Continuumx
05-03-2006, 05:00 AM
Tyrone,
that look very good. Can you give us an indication what sample level was reached? I have a rather simple scene and I had it in the oven for 5 hours and I still got considerable noise in the shadows. I have noticed that the shadow areas are usually noisy as very little light comes from these and it takes considerable longer to calculate. This is the unbiased renderer for you. Do you know if NL has considered to to weight the render effort in order to give problem materials such as glass etc and low light areas more CPU time? I know that this could stuff up the noise distribution function and that might result in noticeable artifacts in the final render but it would be interesting to know.

Hello Matariki,

What you are looking at is about 17.6 S.L. For most scenes a good range is about 12-15. For glass materials, S.L. 25 may be a good baseline. It really depends. The thing you want to do for those low-light scenes is render them 2x the size you want so you can just reduce the image down. This will usually get rid of noise in glass and other problem areas. For this image, this is the size it rendered, no reduction or manipulation. 100% Maxwell Render.

Matariki
05-03-2006, 05:16 AM
That looks really good Matariki. I think the noise is fine enough that NeatImage or Noise Ninja would do a good job.

Regarding weighting, I seriously doubt that NL would consider it. They're very tied up in the whole unbiased thing.

You are right , Adam, they have enough on their plate already.

Tyrone,
17.6 is quite high. I got 13 when I remember right on a 2kX1k image in 5 hours. To reach 17.6 I guess it would take another 5 hours. But as you suggested I can scale down or use Neatimage as Adam suggested.

Thanks guys.

lllab
05-03-2006, 08:20 AM
Nice image Tyron!,

but regarding your suggestion rendering at 2x size:

how can i render it at 2x size if a nromal image here for archviz print is 3500-5000px wide? that would be 7000+- i never manged to get it render that big, even i use the 3gb switch.
whta is the max size you get on a normal arch szene?

cheers
stefan

moka.studio
05-03-2006, 08:44 AM
nice tests Tyron -
3 hours on a single machine? What kind of system? just curious,

Magnus3D
05-03-2006, 09:42 AM
Here's my 1st test with 1.0, it took 16 hours and reached SL 16 on my P4 3ghz with 1 gig ram. I know the floor and wood textures and tiling suck but i had no other textures to use :/ sorry.

http://forums.cgsociety.org/showthread.php?t=352689

/ Magnus

Matariki
05-03-2006, 10:32 AM
Here's my 1st test with 1.0, it took 16 hours and reached SL 16 on my P4 3ghz with 1 gig ram. I know the floor and wood textures and tiling suck but i had no other textures to use :/ sorry.

http://forums.cgsociety.org/showthread.php?t=352689

/ Magnus

That's nice Magnus. I can forgive you regarding the texture ;-). 16 hours thats quite a bit of time and the handle is still a bit noisy even so you pushed it all the way to SL16. Did you use NeatImage or a similar tool. Is this the original size you rendered, I just want to get a feeling for the time it takes to render transparent objects. What material did you use? Sorry for bombarding you with questions.
Cheers

Magnus3D
05-03-2006, 10:56 AM
Thanks :D

Good you too noticed the textures were crappy hehe, no news there. But yep i agree 16 hours is quite alot of time for something like this, although i should say i had it running on a lowpriority thread while i were working on other stuff so that probably slowed it down a bit, but about the polyurethane handle (or however it's spelled) that ofcourse boosts the time needed to render it clear, personally i don't mind how long it takes as long as i get a nice looking image in the end :)

I haven't use any type of noisereduction on this, it's straight outta Maxwell well with the tiny frame and my signature added to it but that's all.

And this is also the original size, 1024x768

The materials used for this i did myself using the materialeditor in Maxwell Studio, the polyurethane material looks like this.. the texture there is to roughen up the speculars and to give the material a slight bump to it.

http://img426.imageshack.us/img426/6689/shadersettings4cs.jpg

Btw, i don't mind answering questions hehe :) ask away..

/ Magnus

AdamT
05-03-2006, 01:07 PM
nice tests Tyron -
3 hours on a single machine? What kind of system? just curious,
Tyrone is using a 128 core Opteron box. ;)

CGTalk Moderation
05-03-2006, 01:07 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.