Hello guys,
here is my last work, i've added 2 macros which i know they aint perfect on sampling but i'll show you anyway







All images have no post production except the main one with a little glow on the window, colors are untouched, dof is bokeh.

Softimage and Mental Ray, lighting consists on 2 portal lights (right and front), neons, mia photo exposure, Final gather only.

a wire



I hope you like it, wish i had more time to fix the sampling on 3d dof but i'm done

Max

Very nice Maximus... did you use the new Unified Sampling? And what was the GI solution used?

Hey, yes i've used unified sampling.
all lights/shaders have sample at 1 (meaning area shadows and glossy samples), Unified sampling settings are Min 1 Max 300 Quality 30 cutoff 0.
Render times for main shot is 8 hours at 1920, macro shots are 5/6 hours.
My machine is a i7 990Xtreme 12 cores clocked at 4ghz. 24gb ram, geforce 570 and ocZ ssd hard disk.
Filter was mitchell set at 4.

Im aware there are some little imperfections. I like unified but i must say that it is a bit hard to control. While at start it seems easier because you just put min/max and control everythin with quality, you always end up havin to launch renders one after another to test the quality, you are litterally forced to do countless tests to see the overall look.
Infact it happened to me when i launched a render overnight after 7 hours i've found out that even if the sampling and parameters was insane, i had ugly artifacts on reflections on metal. And even this render has some after 8 hours. So i cant really say i'm happy.
Everytime you think you've reached a good antialias with a particular setting, you will find out that there is something else thats not working, for example a glossy sample on reflection.
It also have a nasty bug on alpha channel where it produces high artifacts on borders, have no idea how no one noticed this. (it does with any filter you use, from box to gauss to mitchell, so no its not a mitchell problem only).
I think overall is on mental ray style, waste hours to tweak things untill it looks good.

I keep using MR because sadly i've invested too many years learning it and i'm not in the mood to switch over and learn another engine. Everytime i try to switch to something else i go back to MR, not because its better but because i'm tired to learn new stuff. What really annoy me is the complete mess between AD integration and Mental images development.
This will never end and its getting worse every release. This year wont be any different.

Sorry for the rant but after i've spent 40 hours tryin to figure out what setting i should use to have a freakin smooth metal glossy, thats where MR brings you

Anyway i hope you enjoyed the image.

Max


edit: i guess i'll post render settings.

those are the render settings, portal lights and neons have samples at 1, and all shaders are architectural with glossy samples set at 1

Interesting Max... Your rants are completely justified, anyone who deals with MR knows that

I asked because I have tested the Unified Sampler a lot, and have tried to gather information from other peoples experiences also... Overall I think it is a good step forward, but the lack of oficial information on how it works (exactly) and what are the best practices when using it make things shady for users, who at first may think it is actually slower then the ond sampler.

Using 1 local sample for glossiness and shadows, as you described, for example is a great thing! I find that is a great way to get consistent quality all throught the image with low render times, but there is no mentioning of this what so ever in the oficial docs. Shameful. (for those who dont know any thing about this visit this link: http://elementalray.wordpress.com/2012/01/29/unified-sampling-redux/)

At the same time I found a little bug in that workflow. When a lens shader is connected to the pass energy on reflections verys depending on how many local samples you have (check out this thread Ive created at Mental Images forum: http://forum.mentalimages.com/showthread.php?9103-Energy-Inconsistency-with-local-sampling). What is worst, it looks like (from my experiments) that the gamma correction in the render region is implement through some sort of unexposed Lens Shader. So you get different results between the Render Region and the Render Preview when using this technique (assuming you have no Lens Shader in your pass).

The good: When you know all this the Unified Sampler really feels like a great tool. I found myself rendering some (simple) scenes (with not much light bouncing) around in Brute Force mode. If you have tried Brute Forcing GI in MentalRay with the old sampler you know that it just was NOT POSSIBLE. Thin geometry and glossy reflections also have a great improvement in quality.

The bad:
1) I am no developer but it is clear that either the shaders or the lights lack modern optimizations found in other renderers. Explicit lights sample as good (bad) as any emissive geometry which is freaking ridiculous when you work with renderers like Arnold or Vray. The hottest spots of contrast for the sampler are glossy light reflections, if you cant do that right no great adaptive sampler is going to save you.
2) The concept of only 1 quality control for the whole scene is a good. I just found the not linear way it works is very confusing. Example of an extreme situation: try having it at 0,005 your image sucks big time, have it at 0,01 your image looks a whole lot better; try increasing from 0,01 to 0,1 your image looks the same... even though the ratio between the first two values was of 2 and the second was of 10. Plus to me at 1 the image should be perferct, PERIOD. One should not have to find a random number, like 30 in your case, where your image is going to be good enough...

Other considerations about your scene setup:
1) Bucket size is waaaaay to small... this is very time consuming for adaptive samplers. That happends because they have to over sample the edges of the bucket since it does not know how good/bad the contrast is going to be on the samples from other buckets. So if your bucket size is 4x4 that means almost all pixels are being oversampled... Try a very large bucket size and see if differs a lot.
2) Reflection/Refraction depth is way too big... Speacially in very glossy reflections (your case) having more then 1 for trace depth is completaly unnecessary (visually). If I were you I would either bring them down to 1, and then increase it locally (in each material) as needed. Or the other way around...
3) Instead of going crazy with the quality (30) I would do things. First, check if your max samples isnt keeping you from reaching maximum quality... If you have a very high quality already (say, 5) and you still get noise in some materials, just increase the samples of that one material locally and bring quality down (not ideal, but...). Such a high quality may be wasting a lot of samples in areas where they are not needed...

If you have time, and pacience to play with those values, please share your results
And sorry for the long (huge) post...

Hey Gustavo, thanks for the detailed answer.
I follow mental images forum daily even if i dont post there that much due to frustration.
There are a lot of bugs using unified sampling, for example the one you just said about energy inconsistency. I had photographic lens shader, if you simply switch from unified to normal antialias same render region it gives you 2 different outputs in terms of brightness, wich makes no sense so this is another thing to add on top and another problem to debug and to face.

Try to put a simple sphere on a plane, zoom with the camera near the sphere edge possibly focusing on the edge where behind there is the black background, not the grid, render and switch to alpha channel. Enjoy your jagged edge. No matter what you do you can raise to 1milion quality, it wont go away, that is another bug.

There are a lot of problems in mental atm, and its not only in GI its the whole renderer. Now i stop here because i have no idea if it is Mental fault, AD implementation or something else. Of course it is also user fault, i'm not perfect. The real shame is that its been 2 years with almost no improvement in anything on mental ray. 2 Years if not more of nothing, no real update no real gamechange solution, What is more depressing is that i bet there wont be anything on next release, if not some useless thing that wont address the vast amount of problems this engine has.

In my opinion Autodesk completely stopped caring about mental ray integration, 3ds max has Vray as main renderer now, Maya has renderman/3delight/vray as most used engine. Softimage has Arnold/3delight. So why even bother with an old crap engine and why even spend time. Too bad users like me will just get screwed But hey its the life.

Regarding my scene i can assure you ive spent like 40 hours testing and debugging and trying paramters. I'm well aware about the bucket size, but it was the only way to have a speedy preview on some zones since it renders tiny bucket but you can clearly see it faster. The render itself was rendered as Automatic mode.
Agree on reflection/refraction depth, that was 4-4-8 i guess, but wont change that much anyway.

Regarding playin with samples, i would, but when you start reaching some huge rendertime, tryin to figure out the quality/time ratio makes you want to hang urself.
There are way too many parameters with bugs as we said above. I was running test on render regions of like 200x200 pixels and it was takin 10 minutes each time. Its just frustrating and time consuming. That should be dev job, not artist job. its a waste of time and really..frustrating.

I believe that my scene could be optimized again but it was a real nightmare to achieve a smooth metal non interpolated. I dont use optimization->interpolation tab because it really ruine the glossy effect.

I might try to use ctrl.ghost and render with irradiance particles and importons but my spiderman senses already can see its gonna be a total failure. But hey tomorrow is sunday:)

Thanks again for the feedback

Max

I hear you man. I totally hope all this Nvidia ARC thing gives users some results... but honestly, I dont believe it untill I see it, far too frustrated with MR already

I think that's not possible for a render engine because it doesn't know at what point an image is visually flawless.
Even unbiased renderers need to be told when to stop. And the required number of samples per pixels can be VERY different from scenario to scenario.

the unified thing, again.
ok, let me say some more general things about how to setup rendering because i read again my advice to use only 1 sample which david adept for this article as alternative. i wrote this already on elementalray.com why i used only 1 sample.
at the meantime, i dont use 1 sample anymore because i dont use dof rendering at the moment.
to clarify things on this topic, my samplingvalues for 1 glossi and 1 area shadow settingswere:
min 1
max 600
quality 10

you can now imagine that these settings would give smooth dof renders and also automaticly smooth wide glossis. but normally 1 glossi will not be enough for more normal renders simply you cant reach the samplecount for smooth wide glossis you will need. david gave a good example on how you can calculate the sample you need for unified compared to adaptive. so with this in mind, you can see that max300 and 1 glossi cant give you smooth results.
so the awnser to render cleaner glossis is to use 4 local glossi rays for example. i do this all the time at the moment and it works perfect.

so my advice for your scene (render without dof) would be to try these settings:
min 1
max 250
quality 8-10
cutoff to 0.03 and 4 local glossi samples for all wide glossi shaders and 2 local samples for all shades with glossi values of 0.8 for example. glossi values below need 4, below 0.4 try 6 or 8 local glossi samples.
the cutoff will help to reduce oversampling in areas which dont need that high quality/max sample values.

for smooth dof renders it looks a bit different because you need more samples for stronger dof renderings.
here i would go with:
min 1
max 600
quality 8-10
cutoff 0.03

take these settings as a starting point to find the right settings for your renderings. it took a while to understand unified because its really different from adaptive. for example i did tests for 2 weeks and some help from people like david to understand what unified does. you have to understand what it does, its the same like for final gathering, you have to know what you do.

mi is working internally on all the things you want. the whole unified will also include shaders and lights. importance sampling is on the way, examples are the ibl and the bsdf shaders, booth are not implemented by ad. you have new sss shaders and more ibl shaders in mr 3.10 that will simplify the workflow alot. but it depends on ad to implement them in si2013. you have progressive rendering in mr, very usefull stuff, but its not implemented, not to mention iray. i can understand the frustration, i myself cannot use the ibl because of the geo shader limitation inside xsi.

the whole documentation thing is another problem. but you now have the mi forum, use it, you will get an awnser very fast. go over ad and get in touch with the developers.

@max:
i would not use mitchell. the filter sharpens the image by default, so you will introduce more aa problems.
suggestion is to use gaussian shader with 2:2 for sharp renderings.

it would be great if you could share parts of your scene which causes the shading problems.

for scenes with alot of lights i would suggest david's post on elementalray regarding the light treshold. the physical light can be found under the mental ray lights shader in the rendertree.

and perhaps post your renderings in the mi forum with a list of your problems.



for renderer comparisions we should clarify what types of renderings you need, to all users here. it makes no sense to talk about gi and how good it is in vray if you only need one image to render. i have to render some arch animations at the moment for example in vray and i rely on brute force because vray is not capable of render flickerfree gi without insane values or alot of direct lights. lights in vray are not that fast like its often mentioned. it you need alot of them, all areas for example, you also have high rendertimes. only arnold renders them really fast, because of its importance sampling, that is completly missing in vray, the domelight for example. try to render a noisefree image with it and you will be suprised how long this can take with a more complex scene. not to mention the coarse envlighting representation.

what i want to say, its hard to optimize the render in all renderer and its never easy to render complex scenes with more complex lighting.
look how many techs are out there for different kind of problems and every tech which has an advantage over another but has minimum one point that is hard to archieve.
this does not mean i disagree for more mr optimizations.

again, if you have a problem, open a thread in the mi forum or on this board, because it makes no sense to sit down 40 hours to try to solve problem that another user had some time before and found the solution.