Mental Ray vs other renderers?

[wurp]

How does MR compare to a different GI renderer such as Brazil or Vray in terms of speed on GI and normal rendering? Have anyone had experience with both? I realize that one renderer isnt faster than the other in all cases, but how do they compare in general. I know brazil and vray are really fast renderers, and yet they are developed by small companies, they are nothing like Mental Images and Avid. I think Mental Images should be way ahead of these renderers because they got so much more resources, but that doesnt seem to be the case.

[sonne]

Well, according to Henrik Wann Jensen ( http://graphics.stanford.edu/~henrik/ ) - the inventor of the Photonmapping technique - Mental Ray is a rather primitive renderer.
F.ex. you can't use a 3d object as photon emitter.
This means that if, f.ex., you want to use HDRI in your scenes, you will only get a "half" result.
Normally you would map your HDR image on a sphere surrounding your scene, and use that to light your scene. That is possible with MR, but the result will be somewhat limited, because you will have to turn on Final Gathering, and with FG the rays only have 2 bounces, whereas they with Global Illumination (GI) can have as many as you like (turning on both at the same time won't help, because GI needs lights that emit photons, and with Mental Ray objects can't do that)
In other words, you can't do this http://www.finalrender.com/finalrender/ima...\_room\_web\_o.jpg . At least not if you want to use the wall as a light source.
You can't do this either http://graphics.stanford.edu/~henrik/images/imgs/cboxray.jpg, and keep in mind that this is only raytracing. If you want to do this in Mental Ray, you have to turn on FG.
I'm not an expert, but you're right, it does seems strange that Mental Ray isn't far ahead, now that is has been out for such a long time.

/Peter

[deaniopo]

You must keep in mind that Mental Ray is a very big and complex renderer. Integrating GI that works with all the other features of MR (volume shaders, motion blur, lens shaders, etc) takes time. Mental Images just can't integrate all the new GI techniques as they appear.

And besides, global illumination is very cool for rendering models to post on CGChannel, but in production, most of the time we still fake the effect with a few lights... and some times with a dome-light. Having photons go with more than two bounces is useless.

We've been having GI for almost three years now in Mental Ray. And I can count the times we've used it in production on the fingers of one hand.

flord.

[sonne]

Well, aparently it's not a problem for the people behind Brazil, Fastrender and Arnold.
Those renderers are all newcomers, but are already used in production - and that includes Global Illumination.
Take a look at these places: http://www.3dluvr.com/brucester/
http://www.rithuset.se/ (Tobias Dahl

[devdoka]

Shaders are allowed to cast light from any place in 3D space. A light shader does not have to emit light from the light source. The shader is allowed to pick arbitrary geometry and cast the ray from the object's surface if desired. This abstraction is what allows area lights to work. Photons have the same flexibility.

The next release of mental ray will have the capability to emit light from arbitrary 3D objects as a built-in feature rather than relying on shaders to be developed. So there, problem solved.

Matt

---

Matt Lind
Animator / Technical Director
Softimage certified instructor:
Softimage|3D
Softimage|XSI
speye_21@hotmail.com

[pnoizat2]

Im with matt,
as far as i can tell mental images is ramping up feature integration, have a look at http://www.mental.com/p207.html

plenty of wholesome features just around the corner.

[geoffcrowther]

yes, mental is good wholesome goodness.

[srvargas]

Oh Goodie...
question - when will we see MR 3.1 integrated in XSI ?
Sounds like it can finally really compete with RenderMan - with the addition of multisegment motion-blur and sub-pixel displacement (the things that only PRMan could do...)

[atyss]

Yes but first, Renderman is much more than multi-segment motion blur and micro-polygon displacement. It's a complete different architecture and langage.

Second, I'm very curious to see how those features will affect render times, because in terms of speed Renderman is way ahead of mental ray.

Salutations - Cheers
Bernard Lebel

[srvargas]

Indeed PRMan is totally different in terms of tech and speed (especially speed)...
But MR has the advantage of great integration in XSI in particular - the render region and especially the render tree for creation of complex shader graphs - I mean it is more usable by people who are not programmers

[atyss]

I agree with you. But the thing is that large studios have those programmers and expect short render times with top quality. I think mental ray, although already used in this industry, has still a long way to go before being a serious threat to Renderman.

Oh by the way the next version of Renderman will feature raytracing, and Global Illumination, among other things. If those feature render much faster than mental ray, I think mental ray will have a real hard time!

Bernard

[devdoka]

I think you'll see Renderman slow down when used as a raytracer because raytracing involves significantly more overhead. As far as raytracers go, mental ray is very fast, but when compared to a scanline renderer, of course it will seem slow.

Matt

---

Matt Lind
Animator / Technical Director
Softimage certified instructor:
Softimage|3D
Softimage|XSI
speye_21@hotmail.com

[netiglobal]

I love both Renderman and MR but MR's integration into XSI and its overall ease of use is what makes it stand out. I can't wait to see MR 3.1's support of accurate light models like those used in Ligthscape, etc. Also, I thought I heard about a RIB exporter for XSI......could be wrong, maybe it already exists??? It would be great to be able to render XSI scenes with RenderMan. I'm sure this would allow big production houses that are already built around RenderMan to integrate XSI into the pipeline faster.