Korny Klown2
It's more a feeling, the vibrations I recieve from the community..

Internal assumptions are not external realities.

Korny Klown2
So for 1 min and max 9 it will compare the values at least once or up to 9 times if it doesn't find a “correct answer” for that pixel.

Can you upload you noisy 5+min scene please. We'll start to dissect it!

Thanks!

Ok maybe 5 minutes was a little overstated. This is a 640x480 rendering that took 2:34 with PBR, 1 diffuse bounce.

Post up a scene file aswell .


Rob

I tried to to quickly reproduce something similar in Maya.
The rendering took 29 seconds and is already better, with 3 diffuse bounces.
Yes I know it still is very aliased, but the anti-aliased rendering took 31 seconds, so still quite a bit faster.

Lol- can you post a scene from Maya that has the same qualities as Mantra. that looks like crap.

What do you mean with the same qualities?

Whatever…..this is the scene anti-aliased and the light tweaked a bit.

the Cornell box is meant to represent a real world test - the Maya scene looks worse than an 80's renderman scene.

Edit: you'll see that Houdini is much less tweaking of parameters to get a great result. I bet you're going crazy with the controls to get a decent picture.

You once wrote how it took 5 clicks to set the brightness of the UV background texture image in Houdini, and in Maya it was 1…. so how does render tweaking compare when machine costs is much less than artist cost.

Well, the only thing that's missing is the contact shadow so an AO pass might be needed but appart from that it looks better then the Mantra version. Yes, it looks a bit 80s but that's more because of the artificial constant colors which are very unnatural thus it's looking very sythetic. But in regards of sharpness, noisefreeness and illumination it looks much better for the fraction of the time.
But that's not the point. You said we will disect the Mantra version, so that I can learn something.

I bet you're going crazy with the controls to get a decent picture.

Actually no.

Lucas Montell
Because we already know those settings inside and out, we’ve never run into something that we couldn’t do in MR.

…machine costs is much less than artist cost.

Not for me.

Is that Mental ray? You're using final gather with MR I'm assuming. Try animating a camera and the sphere. You know you'll get flicker. MR is great for stills but then you add Motion Blur, DOF and animation and it crawls unless you spend hours tweaking. The default Mantra PBR settings get you all those flicker free as a baseline but with noise.

If you want to get similar time/results as MR with Final Gather in Houdini, you can try adding a GI Light and go to town messing with that. Unfortunately you'll get some flickering in animation until you spend hours tweaking.

Having used too many Raytracers to name in production, I personally find as an artist, the Autodesk implementations of Mental Ray are atrocious in both Max and Maya.

With the lot of renderers adopting a PBR workflow with the intent of settling on a noise amount, tweaking fewer settings have been comforting. The fully featured IPR renderer in Houdini is a huge time saver.

I will say though, Vray and Arnold are faster than Mantra when going for animated DOF, MoBlur and GI. Mantra is definitely faster than MR and Maxwell for sure in a similar respect. Maxwell is a special case as it goes for total accuracy.

And there are admittedly a lack of tutorials for more general type of work for Houdini as the user base is a lot smaller and more focused on TD. A pitfall IMO. But Mantra sure ain't shit. It will of course be touted when doing work Houdini is known for, but for generalist type work I do I feel it can be slow for a small shop.

I do miss options. Arnold is coming, Vray is doubtful and has been asked for years. Redshift is a gamechanger and I hope it comes. Best tools for the job type of thing.

I was bored. And I can relate. As I have gone through similar growing pains with Mantra.

3 Bounces, fairly noise free, usable in my opinion.

1 minute and 50s render on an aged 4C8T core i7 920 with a GI light (it's not something I'm too familiar with so the settings are… something). Still slow though, I'm not an expert on Mantra and still learning it myself.

3 minutes without the GI Light, but with this you could add DOF, and MoBlur for little to no impact.

You had your area light set up incorrectly in a Cornell box test. You had it double sided inside the cage, which means that not only were you casting light down, it was also going up onto the ceiling. Causing more noise than it should. Did you not notice your buckets taking longer on the light?

I could tweak it further but that was just about 20 minutes. Like I said I can relate, but I'm not so quick to judge. There are tidbits of info scattered across this forum and odforce. [forums.odforce.net]

Korny Klown2
I can now officially say that Mantra is a pretty bad renderer. Very sad.

Your horribly arrogant attitude really doesn't help I these discussions and are rather astonishing considering how fundamental lack of knowledge you present not only in Houdini's concepts but general concepts of current CGI too.

Answering in such thread is ultimate waste of time me thinks, but for a sake of others' people interest whose time you might waste: the main difference in those scenes is that you force Mantra cornell box to use specular rays, while mental ray only diffuse. Now, if you were even close the position that gives you the right to render such harsh statements about any rendering engine, you would know what difference this makes, specially in light of path tracing nature of Mantra.

Not to mention using final gather in comparison to brute force path tracing is pretty much meaningless, but how would you know that, right?

The thing is this setup shows precisely why Mantra is light years ahead of your favorite render engine, you simple can't see that, as an expert from openmindn(l)ess perhaps. Any decent path tracer out there can render cornell box scene much faster and better quality than mental ray, specially if you help it with final gather like optimization, which can be done in Houdini with a single click, but you don't know that either, right?

Ask your Maya friends why they spend tens of 1000s of dollars for Arnold before making any further assumptions.

Then compare render times of equal quality images by mental ray, mantra and Arnold.

Agree Symek.

These threads kind of feel like reading a chapter about ‘smash the four olds’
http://en.wikipedia.org/wiki/Four_Olds [en.wikipedia.org]

and we know how successful that turned out…

@SYmek If that was directed at me (it's not clear and just making sure): I'll be the first to admit I have a lot to learn in Houdini still. And my experience is only from small generalist shops (1-3 CG artists) where the idea of a render farm is just co-workers machines.

And I totally forgot matching the Maya scene and turning off the reflective properties of the default shader and the 2 mantra surface shaders set up. Changing them to simple surface shaders with no spec, the render time dropped to about a minute on my same home machine with the GI Light in the scene I posted. 1:10 without the GI Light.

I still think it will be slower than Vray with Brute Force as Primary and Light Cache as secondary which I find gives me no flicker, usable moblur and dof for animation. Which is what I care about. How the renderer gets to that is secondary. I'll run a test to show on Monday when I'm back at work.

With the reflective properties in the shader, KK2's initial scene would have 10 (defaulted) reflective traces. In my scene I dropped that down to 4.

Peter Quints [vimeo.com] amazing free tutorials is the only set I've found that is the next step after SideFX's learning path for new users, and I recommend it to everyone. You get into FXPHD, CMIVFX and you jump straight into TD VFX territory.

No Gyroscope, that was directed to me

Listen Symek,
you call me arrogant?
Yes, maybe I'm not such a superhuman as you guys are and probably I'm not as expirienced as you guys but I want to be and thus I've started this thread, to ask for knowledge. But after even the the experts can't tell me any more about my concern then to look at the documentation and the documentation just says analogous “One should always use PBR” and “There are only two controlls for quality”, I have to assume that this is all. The problem is that you guys can't put yourself in my position. You're already familiar with Mantra. Maybe you've never known this position, maybe you don't know it anymore but you don't know how confusing it is to see Mantra in front of yourself and not knowing what do and yet you can't put yourself in my place you do judge about me and judging about other people, that is in fact arrogant.

Here, the number one gym rule:
http://37.media.tumblr.com/4c9bf4632a6e7eee53eb67e804cd6448/tumblr_mxh4de7jd41t0vx74o1_500.jpg [37.media.tumblr.com]


MartybNz,
throughout this entire thread I haven't heard one single critic about Mantra of you guys. All I hear/read is that Mantra is on the same level as Arnold and other A grade VFX tools, Mantra is so much better you're just lacking in knowledge, if you don't know how to use it it's because you are used to such aweful renderers like Mental Ray and V-Ray. Do you still wonder why I feel the vibe of a glorified Houdini/Mantra? Gyroscope was the first here who contributed an objective statement. When you read all the posts carefully you will see that saying

"MartybNz
Internal assumtions are not external realties.

that is in fact arrogant.

MartybNz
These threads kind of feel like reading a chapter about ‘smash the four olds’
http://en.wikipedia.org/wiki/Four_Olds [en.wikipedia.org]

What?!
Except there is no death and distruction in my threads.


I started this thread to learn a thing or two but all you do is putting me down for not being as expirienced as you guys.

Gyroscope, thanks for the link but I already know about Peter Quints tutorials about global illumination and PBR but espacially the one about gi is highly confusing not least because he's using the terms global illumination, full irradiance, photon map and final gathering interchangeably which is wrong in oppinion because they don't mean the same and since he's jumping back and foth between these terms I never really know which method he's actually talking about. In addition to that these tuts are recorded with version 10 or earlier, so now, 3 versions later, I don't know if the same workflows still work.

Gyroscope
1 minute and 50s render on an aged 4C8T core i7 920 with a GI light

I opend your scene and started a rendering without any changes and it took 03:01 on my AMD 4C8T (AMD FX-8320) 12 GB RAM.

hi
here is a render of mine
https://vimeo.com/69748935 [vimeo.com]
and here is a scene
http://forums.odforce.net/topic/17960-pbr-render-test/ [forums.odforce.net]
check the mantra settings
i think in my experience for full HD frame and decent sampling it is very hard to get bellow 30min per frame
on my dual xenon 72gb ram machine

hope this helps

Korny Klown2
Ok, all the time I was thinking, I was doing something wrong but now that I've watched the tutorials you have linked here, some of them I already knew, I can now officially say that Mantra is a pretty bad renderer. Very sad.

This is the 6th post of the thread and until this statement there were nothing but honest and helpful replies. Maybe instead of calling Mantra a ‘pretty bad renderer’ you should invest a little more time getting to know it.

Admitting to the fact that you still have a hard time getting acceptable results, leads to more help from forum members. On the other hand oversimplified statements about mantra raises questions about your understanding of render engines in general.

Back on topic:

Maybe it's a good idea to make your Maya scene available to have a look at how you set up things. Linear workflow, correct falloff on the area light, Min/Max settings on antialiasing, FG settings.

Thanks for your time,

Andy

Andy58
Maybe instead of calling Mantra a ‘pretty bad renderer’ you should invest a little more time getting to know it.

Maybe I just should have appended “for my needs” after “pretty bad renderer”.

Andy58
On the other hand oversimplified statements about mantra raises questions about your understanding of render engines in general.

As I said above, as a beginner in regards of rendering Mantra I was looking for as much information as possible. I was watching some tuts before this thread and they confused me because I didn't see someone dailing parameters for an optimal renderesult. The next thing I did was hovering over each parameter in Mantra node to see if I can get any more information from there and I also tried reading the documentation, still not finding any more info. Then I started this thread thinking: Before researching lost and aimless on my own, the guys at the forum can surely direct me to the necessary infos a lot quicker. The feedback then was basically all the stuff I already looked up and just a few links I haven't looked at so far. After that journey searching for infos and still not finding any more then “one should usualy use PBR” and “there are only two controls for quality” only then I was simplifying because I thought: "If even these guys don't have more information, there's probably not much power to it.

Andy58
Maybe it's a good idea to make your Maya scene available to have a look at how you set up things. Linear workflow, correct falloff on the area light, Min/Max settings on antialiasing, FG settings.

Linear workflow: yes
Falloff: quadratic
Anti-Aliasing: min 0, max 2
FG: default but Point Interpolation 20, Secondary diffuse bounces 2

But the Maya scene is not my problem. I would like to know techniques to optimize a Mantra rendering.

Thanks for your objectivity

klown matey
this attitude wont get you far
hundreds of people with experience far greater than mine or yours used Mantra on very demandin productions
and calling it bad just wont cut it
stick with it, learn it well and reap the benefits
its not a matter to be solved in a week

arzo
hi
here is a render of mine
https://vimeo.com/69748935 [vimeo.com]
and here is a scene
http://forums.odforce.net/topic/17960-pbr-render-test/ [forums.odforce.net]
check the mantra settings
i think in my experience for full HD frame and decent sampling it is very hard to get bellow 30min per frame
on my dual xenon 72gb ram machine

hope this helps

Hi,
I have taken a look at the scene.

1) Write all your geometry to disk, right now your scene cooks all your poly modelling. Thats one big no no .

2) Explore instancing and packed primitives

3) There are some old shaders being used there. not sure why ?

4) Is your ray tracing bias set correctly for your scene size ? Put a bound over the scene ie 70000 unit scene size x 0.01 .

5 Why have you set 2 diffuse bounces on the mantra ROP and set min ray samples to 5 ? with pixel samples 7 x 7 with PBR. What was your thinking ?

6) Any chance you can post up a full scene in a folder structure with maps etc. It would make a good test scene for beginners to use as there's no highly reflective surfaces and its set indoors. Should be pretty straight forward.

7) Render times … 30mins for your scene seems excessive, mind you on Enders game I was rendering 60million instanced ships with absolutely everything turned on and that was using near to 80 gigs of ram and took 3hrs a frame .
:?

Rob

@KK2 - Optimizing Mantra… It's not like Mental Ray of yesteryear so you kind of have to forget that. Or if you're familiar with the new Mental Ray Unified Sampling method of rendering it's similar to that.

Here's a modified scene. 35 seconds to render on my machine. Mimics your non-reflective Maya render as was discussed earlier but seems to be ignored?

I can't tell what you really want to know. You seem to know what you need to know but aren't wanting to accept it? It is actually that simplified when it's a noise problem. I can't think of a modern rendering engine that is not using this approach. And I'm thankful as I don't have to spend hours dialing in numbers and spend more time on the art.

@circusmonkey
6) I love that suggestion circusmonkey! I was working on the latest CGTalk lighting challenge [forums.cgsociety.org] to learn this very topic of optimizing Mantra, but got majorly sidelined by work. I've got the scene but no shaders set up.

If that scene isn't fully shared, maybe I can set up this older CG Lighting Challenge [3drender.com] that can be used.

7) Was that 3 hours at 2k? Arzo's dual xeon with 72gb of ram.. I can only imagine dual 3.0ghz 8 cores at a total of 32 threads. And yeah 30 minutes a frame on that machine seems like a lot.