Shop Mobile More Submit  Join Login

Featured in Collections


Submitted on
October 29, 2011
Image Size
495 KB


6,923 (6 today)
10 (who?)


Creative Commons License
Some rights reserved. This work is licensed under a
Creative Commons Attribution 3.0 License.
Render Test: Mental Ray VS Cycles VS Luxrender by BrentOGara Render Test: Mental Ray VS Cycles VS Luxrender by BrentOGara
Edit 1/9/13 - Added Luxrender test of the same scene.

Just a test see how well Blender's (in-progress) Cycles render engine stacks up against the (somewhat more expensive) Mental Ray... and how Luxrender beats them both.

MR is standard in Softimage, and a very popular renderer for Max and Maya as well... since Autodesk owns them all, and MR works very well, it's probably one of the most used render engines out there.

Then there's Blender Cycles... still in progress, a GPU accelerated unbiased physically accurate render engine... and it looks just as good.

As an added bonus Blender is free, where the programs that MR comes with cost thousands... and yet, they all do the same things, in much the same way, in the same amount of time.

Luxrender uses some very nicely optimized bi-directional path tracing algorithms with full-spectrum lights (allowing for real dispersion and prismatic effects) to quickly render true-to-life images. You can even choose real world film stock and actual real world lights inside the render engine to exactly match the real world conditions (and that's a lot of real!).

For any 3D work dependent on realism (as so many of them are), having an accurate model of how light actually works is pretty important, and the subtleties of reflected and refracted light is often critical to making a group of digital objects feel like they are real, solid, and sharing the same space.

Here's hoping all of these renders have fooled you into thinking these are all real objects in a real box with a real light on top... and just take a moment to realize one of these renders "cost" thousands of dollars... and the other two were "free".
Add a Comment:
WrathOfCat Featured By Owner Edited Oct 16, 2014  Hobbyist Digital Artist
To get dispersion in cycles you actually need to make that a part of the material nodes. 


Also i'm having a hard time believing that's 8 hours in cycles :v
IamZandar Featured By Owner Oct 17, 2014  Professional Digital Artist
I did recent work in Cycles that ended after 2 days, to complete reduce noise.  So I do believe.  I hard time believe that Luxrender came to no noise in just 8 hrs. :D (Big Grin) 
LuxXeon Featured By Owner Oct 16, 2014  Professional Digital Artist
I agree Luxrender has one of the most sophisticated light transport algorithms of any engine, but having worked extensively with both Luxrender and MentalRay in 3dsmax, I can honestly say I suspect the MR settings were not fully optimized for performance/accuracy here.  Notice the splotchy shadow results, which indicate some low final gather settings.  At 8 hrs, the results of this simple scene should have been much much more refined in MR.  I'd be curious to see an Octane render or Corona render of the same scene for comparison, being they both also use full spectral light with dispersion as well.
BrentOGara Featured By Owner Oct 16, 2014  Hobbyist Digital Artist
I'm sure my MR settings were less than optimized, I've not done a ton with it since I left college. If I had Octane or Corona I'd certainly add them!
LuxXeon Featured By Owner Oct 16, 2014  Professional Digital Artist
I see.  Otherwise, I think it's a fair test, and set up properly.  I could help you with MR settings, if you ever decide to revisit this test.  I think speed vs. quality, MR should produce a much more viable result, and certainly much faster than the competitors.  It's been a go-to realistic engine for some very large and reputable studios for over a decade, and there's a valid reason behind it.  I've personally found it to be very reliable and predictable, once the settings are mastered, but the problem is that settings need to be adjusted on a scene to scene basis, which isn't the case in path tracing render engines.  I prefer path tracing to other algorithms, for accuracy and ease of use.  The fastest path tracer I've ever used is Corona.  The results are amazing, but the algorithm doesn't seem to be quite as accurate as Luxrender, in my opinion.  It's not bi-directional, but does have a built-in secondary algorithm which can be made to do bi-directional path tracing.  The results of that are, however, far slower than the standard progressive algorithm.  Regarding many scenes, the difference between single and bidirectional path tracing is irrelevant.  You wouldn't notice the subtle difference.  However, scenes with a lot of refractive materials benefit greatly from it.
BrentOGara Featured By Owner Oct 16, 2014  Hobbyist Digital Artist
I loved MR when I was using it in college. We used Maya and XSI, and some students chose to use Max. By the time I left school Autodesk had bought them all, and the MR was identical across all 3 systems. I never had much luck optimizing MR's Photon Mapping or Final Gather. I read a lot of information on it, followed every tutorial, but I never could get the balance of speed/accuracy where I wanted it. I recall that when I first ran the tests, I was surprised that it took 8 hours to complete the MR render, as I had expected closer to 2-3 from my college experiences with it.

I'll have to look at Corona, I do love a fast & accurate physical render engine. I simply adore LuxRender's bi-directional path tracing, it produces caustics to die for. I just with Cycles would get bi-directional tracing so I can stop crying over my muddy, scattered, indistinct glass/water renders.

Path tracing is just so easy to use... it makes all the fiddly little settings of biased engines seem so baroque.  
stephane-richard Featured By Owner Oct 6, 2014  Professional Digital Artist
I agree with the last comment. And to add more, Mental ray in Maya let's say, is all about mia_material, linear workflow and physical lighting. I would say you would have to use all three renderer with the best of their ability in some kind of standardized protocol (linear workflow and correct gamma correction at least) to make a descent comparison.
BrentOGara Featured By Owner Oct 7, 2014  Hobbyist Digital Artist
That's the problem with comparing different render engines, you have to keep the parameters very simple (like the Cornell Box) to make the comparison meaningful. With an old image like this one, it's even harder, because all the render systems involved have moved on quite a bit by now.

In the end, it's like comparing a Fuji apple vs a Pink Lady apple vs a Pineapple.

The fact that the 3 engines can produce extremely similar results in the same time frame cannot be denied, but some will quibble over specific details of implementation (or flat out call you a liar) until the end of time.
DynamicEvolution Featured By Owner Aug 30, 2014  Hobbyist Artist
Problem with this test is there are no screenshots of the settings for each renderer.  So we don't know how many samples and proof of the samples used. So you can lower the sample rate on Mental ray to improve your claim.   You state 1024 @ 32 but we don't see proof of it.  I'm sorry Maybe i should get all 3 and fake a test just as you guys =3  
BrentOGara Featured By Owner Aug 31, 2014  Hobbyist Digital Artist
Problem with this comment is there are no screenshots of the poster actually typing it. So we don't know if it's a sentient life form or a spam bot. So if, for some reason, some idiot did post a false comparison image, what would stop anyone else from repeating the experiment and discovering the truth?

Nothing. Nothing is stopping anyone with a decent computer and internet access from doing this same test themselves. That's how science works. You see an interesting result, and you try it yourself to see if it's repeatable.

You could, in fact, get all 3 and fake a test... but what would you gain from that? What would it profit you? What would you have learned? What good would that bring into the world? How would that help anyone?

It wouldn't even help the program you were lying in favor of, because anyone testing it would quickly discover the lie by direct, personal experience.

You'd have to be an imbecile to try and fake something like this.

You'd have to be lacking in critical-thinking ability to even believe such a transparent deception might work, or be a viable means of changing minds in the real world.

Don't be an imbicile, do a little thinking, try it for yourself. All you have to lose is your ignorance. 
Add a Comment: