crazy homeless guy

A couple of months ago, Mental Ray announced that they were rolling out a proxy system. Well, with the release of Mental Ray 3.6, the proxy system was rolled out. So, I am doing a head to head comparison between the two. I am a Vray user who is in the process of adopting Mental Ray. I feel off hand, at first stab, Vray is going to come out on top. I can say this because I don’t know how to speed tweak Mental Ray like I can Vray. Hopefully I will learn some tweaks during this process, making the final result a valid comparison. All times posted include the time needed for the lighting calculations. Basically, they are from the time I hit render, until the time the rendering finishes and save to the hard drive.

Below is a link to the screen shot of the basic setup I am using for the comparison. It consists of a plane, 2 cameras, and a daylight system. I will be setting the daylight system, materials, and cameras up for both render engines so that they can have full access to what makes them unique. The ground plane will be a neutral gray of 128,128,128. I will be working with a gamma of 1.0. The scene will not have reflection on any objects to try and keep the playing field level on the proxy issue, and not bring in too many other factors.

http://www.phase22.com/misc/cgarchit...es/initial.jpg

I was starting off setting this scene up using Onyx trees, but I think I am going to switch to Dryad Tree Generator. I would like to post the scene for others to tweak and try. If I use Onyx trees, I will not be able to do that legally.

My proxy object has roughly 444,000 polygons, and I am using 300 instances of it. So basically it is a scene with more or less 133 millions polygons. Though I don’t know if you can really say that you are rendering that many poly’s since we are using round about methods to do it. It does thought give you an idea of the complexity that can be achieved with proxies. The size of the tree after converted to a Mental Ray proxy is 67.7 megs. The size of the tree after converted to a Vray proxy is 63.2 megs.

I should also mention my computer specs and software used. Dell Precision 490 - Windows 64bit, Quad Core, 8 gigs of ram. All Vray tests will be done using Max2008 with Vray 1.5sp1. All Mental Ray test are being done using Max2009 with Mental Ray 3.6.

Here are the links to the scenes...

VRAY : http://www.phase22.com/misc/cgarchit...oxies/vray.zip
MENTAL RAY : http://www.phase22.com/misc/cgarchit.../mentalray.zip

These test are also going to create a comparison of some of the nuances that the differ in. How the color toning differs, how the sky system differs, how the camera settings differ, etc..

_________________________________________


The results from the Vray test.

The aerial shot rendered in 13 minutes and 33 seconds. The ground level shot rendered in 10 minutes and 11 seconds. I am trying to keep the playing field as level as possible, so I will also make note of AA settings, as well as comment on the GI methods used. The GI methods used cannot be equal because they are the primary thing that makes each engine unique to itself.

The Vray scene is setup using Irradiance for the first pass, and Light Cache for the second pass. I won’t go into the details of the individual settings for Vray, if you are curious, download the file and look. The image sampling is set to Adaptive DMC at 1,4. I am using no AA filter. I do most of my production work at 2,6, but I have dropped it to 1,4 during production when I simply need to get something out the door.

With the scene loaded in Max2008, the RAM usage sits at 228megs. When I hit render, the usage climbs to 603 megs during the lighting calculations, but back off to 527 megs while rendering.

Vray Aerial > http://www.phase22.com/misc/cgarchit...ray_aerial.jpg
Vray Ground > http://www.phase22.com/misc/cgarchit...ray_ground.jpg


________________________________


The results from the Mental Ray test.

The aerial shot rendered in 11 minutes and 45 seconds. The ground level shot rendered in 13 minutes and 8 seconds.

With the Mental Ray scene loaded in Max2009, the RAM usage sits at 414 megs. Keep in mind that these are different versions of Max, so that could count for the extra RAM usage. While processing the Final Gather and rendering the image, the RAM usage peeks somewhere slightly over a gig. That is slightly more than Vray, but nothing to get upset about. It may be due to my lack of knowledge on how to setup Mental Ray for maximum efficiency.

I am using one of the presets included in the Mental Ray dialog window, mental.ray.daylight. The only setting change I made was switching to BPS2 instead of BPS. This preset seems to produce fast results, but will definitely need to be tweaked to produce a low end production quality image like the one created using Vray. The obvious blemish is with the shadows kind of disappearing around the base of some of the trees. I am using fairly low sampling and AA settings. The sampling is set to 1/4 : 4 with a Mitchell filter set to 2 : 2. I need to learn a little more about the AA filters in Mental Ray. As said above, I like to disable them altogether in Vray to increase my speed. I am not sure if Mental Ray will respond the same way, or if there is even a possibility to disable them altogether.

The color toning in the Mental Ray image is on the dark side right now. This is something I should be able to fix in the Mental Ray Photography exposure dialog, and should not effect the overall rendering time of the image. I will probably re-render it later today with a few exposure adjustments, and just edit this post to reflect that.

Mental Ray Aerial > http://www.phase22.com/misc/cgarchit.../mr_aerial.jpg
Mental Ray Ground> http://www.phase22.com/misc/cgarchit.../mr_ground.jpg

________________________________


It is obvious that the Mental Ray proxies are going to be capable of the complexity that can be created with Vray, which is what this test was about. I won't have a solid hold on how they compare speed wise until I am a more in depth Mental Ray user, and can really tweak the engine with settings that I know will make it fly.

I am going to continue further testing with these 2 scenes, slowly adding in a glass box, ies lighting, and maybe a few other things to get a further feel for how it is going to respond. As of now, neither of the scenes have reflection in them.

Feel free to download the scenes, and play with them. Tweak the settings, run tests, or whatever you please. When you find things worth sharing, post them so we can all learn.

I must say, that I am fairly happy with the initial results.

__________________
travis schmiesing

manta

And now you can compare them even more evenly since vray SP2 is out and also has animated proxies...

__________________
DESIGN FORWARD
http://www.mantastudios.com

manta

Speaking of animated proxy objects...anyone know if the file size will be huge for say a tree blowing in the wind...its going to have to somehow retain all that info in the file right ?

__________________
DESIGN FORWARD
http://www.mantastudios.com

crazy homeless guy

Not sure. A proxy is still vector data, so applying animation settings to it would require a dataset that is overly complex. ..but the only way to find out is to run some tests on it. Maybe take an object that is not animated, and make a proxy of it. Then take the same object, animate it, and make a proxy of it.

__________________
travis schmiesing

Justin Hunt

Is it possible to use Vray proxies with particle Flow?

jhv

crazy homeless guy

I don't know if it is possible or not. I would imagine so. Bobo's script fixes instancing for pFlow, I don't think it is just for Mental Ray proxies. I think he says it works with any object.

My goal of this thread was not to create one where it was Mental Ray is better than Vray, or vice versa. Which I am afraid is the direction it is starting to go. It is my feeling that blanket statements that try to generalize things as being better or worse tend to happen in those discussion without giving a solid look at each product.

I use Vray Proxies heavily in my daily work flow. I am in the process of implementing Mental Ray into my work flow. I wanted to make sure that Mental Ray's proxy abilities were capable of producing what Vray's were. I wanted to share my results with the community so we could all learn from them.

__________________
travis schmiesing

Justin Hunt

Your right, I just wanted to know if it was possible nothing more

Been following this thread, nice testing thanks

jhv

MasterZap

The Bobo script I post about in my latest blog post is renderer agnostic (actually object agnostic - it simply makes "real instances" of whatever you give it for the PFlow object), so in theory it would even work with vRay.

/Z

__________________
mentalraytips.blogspot.com - mr Zap Andersson - a completely mental shader guy

crazy homeless guy

A couple of new tests.

A new scene this time. The scene has 1.44 billion polygons via 2 proxy objects. The two images are the same, just different sampling and AA filters. The first image is a box filter set to 1 : 16, the image with these settings took just shy of 5 hours to complete. The second is the Lanczos filter set to 4 : 4. With these setting the image took 2 hours and 46 minutes to complete. The box filter looks better, but at nearly double the time. Though I would rather use a sharpening filter or no filter for a still. Box is a blurrying, though it is still usable.

I think I might set up some extensive sampling and AA tests this weekend.

The times are for lighting calculation and rendering at 3000 pixels wide with a MR sun and sky setup. The lighting calculation finished under 2 minutes, so the vast majority of the time is sent sampling and filtering.

Still no reflections in the scene.

Lanczos 4 : 4 – 2 hours and 46 minutes > http://www.phase22.com/misc/cgarchit...st_lan_4-4.jpg

Box 1 : 16 – 4 hours 55 minutes > http://www.phase22.com/misc/cgarchit...t_box_1-16.jpg


.

__________________
travis schmiesing

demo38

Thanks for the testing and the thread on this. The Box filter is definitely a nicer image quality. The trees look great.