PDA

View Full Version : Mental Ray 3.5 Features


bjoern
06-12-2006, 06:10 PM
http://www.vfxtalk.com/forum/showthread.php?t=5816

Rolf Herken "CEO Mental Images": Well let’s see...in the immediate future mental ray 3.5 will be released.
The development of version 3.5 is focused on improvements of performance and usability, while ensuring a high level of compatibility with earlier versions. The following areas are at the focus of the improvements:
- global illumination and final gathering ease of use and performance
- ease of shader creation and shading performance
- rasterizer performance
- ray tracing robustness
- ease of parameter tuning for options and scene data
- backward compatibility
- physically based rendering

For the future we are working towards the ultimate goal of being able to render incredibly complex scenes (with billions of triangles) using raytracing based global illumination algorithms with complex shading including BRDFs, subsurface scattering, and frequency dependent phenomena
*in real-time*.

Our next generation software rendering technology that is built into RealityServer, achieves rendering times of less than 40 milliseconds per frame (i.e., realtime) with full raytracing (and very soon full global illumination) on small 64-bit commodity multiprocessor machines or blades on scenes with tens of millions of polygons.

Parts of this entirely new, revolutionary rendering technology such as our dynamic BSP acceleration methods are being integrated into mental ray, beginning with version 3.5. In mental ray 3.6 we will have support for assemblies, allowing rendering of arbitrarily large scenes, e.g., crowd scenes with a million creatures, even with our raytracing and global illumination algorithms.

mech7
06-12-2006, 06:51 PM
Bring it on :buttrock:

*in real-time*.

mdee
06-12-2006, 06:58 PM
That's very exciting news. MR is already top renderer and the way they improve it (adding and expanding very advanced features while making them faster and easier to use at the same time) is great. Seeing it used heavily in feature movies recently makes me really happy.

I have very big respect for people like Rolf Herken who are the inovators in CG world. Without them, there would be no CG as we know it.

mustique
06-12-2006, 07:01 PM
Good news. Whish they will add faster motionblur and DOF too.

Clanger
06-12-2006, 07:19 PM
Will this mean I can finally render high res stills >3500 pixels without memory errors?

Bonedaddy
06-12-2006, 07:30 PM
All this seems nice, but, unless I'm missing something, they're just saying, "We do everything we used to do, but more betterly!"

Could someone up on the MR tip decode the PR-speak a little?

gent_k
06-12-2006, 07:45 PM
All this seems nice, but, unless I'm missing something, they're just saying, "We do everything we used to do, but more betterly!"

Could someone up on the MR tip decode the PR-speak a little?

Well that can't be bad when it comes to MR, as there isn't much it doesn't do already, so speed and performace improvements (and ease of use for the less technical people) are the top on the list.

The MetaSL thing sounds interesting and is something 'new', even though I don't exactly understand what it can do.

ca11j31
06-12-2006, 07:58 PM
Hello there everybody,

I might be missing something here but I have always believed(trough years of experience) that the current version of Mental Ray for 3ds max was already capable of most things mentioned above...I could be wrong tough...can anyone enlighten me please?!

Kind Regards

Jamie

Strang
06-12-2006, 08:01 PM
The MetaSL thing sounds interesting and is something 'new', even though I don't exactly understand what it can do.

hopefully cross platform shaders with ease and maybe an interface for non-programmer types

P_T
06-12-2006, 08:28 PM
I like the way that "EASE OF USE" sounds. :D

It'll also be great when renderers start taking advantage of any graphic cards in their engine. Gelato does this but it only works with NVidia cards right?

CupOWonton
06-12-2006, 08:56 PM
"Reality Server"? Is this part of their deal with the makers of the Render Drive?

malducin
06-12-2006, 09:23 PM
All this seems nice, but, unless I'm missing something, they're just saying, "We do everything we used to do, but more betterly!" Could someone up on the MR tip decode the PR-speak a little? ... I might be missing something here but I have always believed(trough years of experience) that the current version of Mental Ray for 3ds max was already capable of most things mentioned above...I could be wrong tough...can anyone enlighten me please?!


That's because it's a minor update, from 3.4 to 3.5. You'll have to wait for 4.0 for any major changes. So it's mainly bug fixes and performance improvements, it's the same features so it should be the same in 3DMax, Maya, XSI, etc.

"Reality Server"? Is this part of their deal with the makers of the Render Drive?

Reality Server predates that. It's a server based solution for internet and intranet applications.

stew
06-12-2006, 09:50 PM
All this seems nice, but, unless I'm missing something, they're just saying, "We do everything we used to do, but more betterly!"
Which in the case of MR is a lot, IMHO.

oddforce
06-12-2006, 10:18 PM
...using raytracing based global illumination algorithms with complex shading including BRDFs, subsurface scattering, and frequency dependent phenomena
*in real-time*.

BTW I've always been wondering whether this gesture existed in english speaking countries where you pull down one lower eyelid with your index finger :) Would apply so nicely here but I guess this might be a germany only thingie.

sdserbos
06-13-2006, 01:22 AM
Oh and please make sure you tell the developers of the host applications how to implement all these wonderful features that you are adding. And some useful documentation would be ... useful too.
And try not to break anything that ain't already broken....

shingo
06-13-2006, 01:30 AM
Sure, that's precisely the point. If Mental Ray did all of these with a minimum of fuss and overhead, it would be a near perfect renderer, which is why Renderman is still the standard. Renderman, for all it's genius, is still based on a lot of clever cheating.

MR has alwasy sufferred from it's high accuracy, which has meant long render times and large memory footprints.


All this seems nice, but, unless I'm missing something, they're just saying, "We do everything we used to do, but more betterly!"

Could someone up on the MR tip decode the PR-speak a little?

Tripdragon
06-13-2006, 01:42 AM
holy cow the second interview was sooo full of ... errr &^* Not to say Mental ray is no great it's just like he sat there and bad mouth'd ayone that stood up to try ...

Oh well.. Can't wait till realtime raytraceing comes from everywhere

NUKE-CG
06-13-2006, 02:34 AM
Thanks for the link. The Poseidon images alone are worth a look. ILM always bring their A-Game to the screen, that's for sure.

mental Ray's accuracy is always in the back of my mind, because I think it could be doing things faster if it faked or loosened up on trying to be so precise. Rendering in general is not accurate, or we would be rendering with huge AA levels, with lights casting billions of photons etc.. so I think mental Images should drop its stance on physically correct and focus on speed. With this real-time comment, I guess they have.

Szos
06-13-2006, 02:40 AM
Where is the support for graphics card-assisted rendering?!?

(yes, I know I ask this very question at every new release of gfx card hardware and software, but enough is enough with this - soooo much wasted processing power. It's pissing me off)

Lorecanth
06-13-2006, 02:42 AM
didn't really see too much BS there...a lot of these features have been shown at various mental ray functions over the last couple of years. Just because they arn't built in with a particular button into your choice of 3d application doesn't mean the features arn't there. As far as the real time rendering its rather obvious that its artvps's rendering chips. Whats unique though is that the rendering engine and user writable shader system is going to be platform agnostic.

mustique
06-13-2006, 08:54 AM
Rolf Herken is a great man, but I don't agree on some things he mentioned in the interview.

Games and films are unlikely to converge as far as rendering goes. You can use models
and similar assets made for games and use them for film work too, (vice versa)
but the rendering proccess will remain "realtime vs software" rendering.

MR "is dipping it's toes into all markets" and the inevitable conclusion is it will be
a good all around solution, but not the best.

Its going to be hard for MR to replace Pixars Renderman when it comes to films.
(Especially with the strong Prman 13) The visualisation market on the other hand,
is not without competition too.

Easy shader/GI creation along with raytracing acceleration via GPUs sounds interesting.
But I hope Mentalimages won't rely on a niche hardware company like ArtVps for breakthrough speed improvements.

-Vormav-
06-13-2006, 12:22 PM
Heh. MetaSL, eh?
There goes all that knowledge of writing shaders the traditional C way...
But I guess it was bound to happen eventually.

wizzackr
06-13-2006, 03:53 PM
raytracing acceleration via GPUs sounds interesting.
But I hope Mentalimages won't rely on a niche hardware company like ArtVps for breakthrough speed improvements.

I personally would love to see a close collaboration between mental images and ArtVps. Imagine the potential of an ArtVps chip to fit AMDs recently announced HTX slots - the potential gains from a specialised processor with direct CPU interfacing could surpass the gains from GPU acceleration by far, niche market or not.

mustique
06-13-2006, 04:19 PM
I personally would love to see a close collaboration between mental images and ArtVps. Imagine the potential of an ArtVps chip to fit AMDs recently announced HTX slots - the potential gains from a specialised processor with direct CPU interfacing could surpass the gains from GPU acceleration by far, niche market or not.

Physics, raytrace and other co-processors will fight for one HTX slot then... :)
I'm OK with it after all, as long as these co-processors don't cost thousands of dollars.

dagon1978
06-13-2006, 05:37 PM
Rolf Herken is a great man, but I don't agree on some things he mentioned in the interview.

The main reason why MR is widely used is because it comes free with the main host apps.


uh?
are you trying to say that ILM use mental ray because it's free?? :argh:
please man!

albedo4800hp
06-13-2006, 06:39 PM
Easy shader/GI creation along with raytracing acceleration via GPUs sounds interesting.
But I hope Mentalimages won't rely on a niche hardware company like ArtVps for breakthrough speed improvements.

As far as I understand GPU acceleration means basically lifting scanline jobs to the GPU. Lets face GPUs are extremely bad for accelerating raytracing unless very specific parts like maybe the ray shading there are far too many bottlenecks and current GPU architectures stands crosswards to the raytracing algorithm. Most of the graphics research done in this area clearly show that. PPU instead might be a far more interesting technology to try out their design is much closer to the raytracing algorithm especially when you considering collision handlings and probably feedback to main CPU

mustique
06-13-2006, 07:24 PM
uh?
are you trying to say that ILM use mental ray because it's free?? :argh:
please man!

I removed that statement cause it sounded like I was bashing MR which was not the case.

From the article:
... "I don't think mental ray "is dipping its toes into all of these markets". mental ray rather has a very large, if not the largest footprint in each of these markets."...

I was refering to this statement. Indeed MR is used almost everywhere by even the smallest studios, but that's just because everyone gets it free with Maya/Max/XSI. The userbase of the standalone version of MR is not as populated as perceived. Of course big studios like ILM use it cause they dig a bit deeper than most small studios/individuals.

Strang
06-13-2006, 07:46 PM
As far as I understand GPU acceleration means basically lifting scanline jobs to the GPU.

i think these guys are doing more than just putting scanline jobs on the GPU

http://homepages.paradise.net.nz/nickamy/realtimerad/realtimerad.html
http://www.cs.unc.edu/~coombe/research/radiosity/
http://www.openrt.de/
http://www.bee-www.com/parthenon/index.htm

but i dont know

stew
06-13-2006, 08:10 PM
i think these guys are doing more than just putting scanline jobs on the GPU

http://homepages.paradise.net.nz/nickamy/realtimerad/realtimerad.html
Actually, this method (IGI) is using the GPU only for scanline jobs. It's calculating hundreds of regular bounce lights that are then rendered on the GPU with OpenGL.
http://www.openrt.de/
Are these guys using the GPU? I know they're making heavy use of SSE and clusters, but I wasn't aware of any GPU acceleration.
http://www.bee-www.com/parthenon/index.htm
This one is using the GPU for final gathering. In some ways, even his method also boils down to scanline rasterization, if I undersood it correctly.

There are research renderers that implement general raytracing on the GPU, but I haven't seen any implementation that would be anywhere near production ready. The most usable GPU assisted renderer these days is probably Gelato.

Strang
06-13-2006, 08:25 PM
Are these guys using the GPU? I know they're making heavy use of SSE and clusters, but I wasn't aware of any GPU acceleration.

as i said.. "i dont know". I was hoping these links could further the GPU acceleration discussion...

soulburn3d
06-13-2006, 09:48 PM
Just because they aren't built in with a particular button into your choice of 3d application doesn't mean the features arn't there.

But if there's no way to access it beyond programming or using the standalone version, might it just as well not be there since 95% of its users can't get to it?

- Neil

azshall
06-14-2006, 02:14 AM
Good news. Whish they will add faster motionblur and DOF too.

amen to that!

P_T
06-14-2006, 02:32 AM
http://www.bee-www.com/parthenon/index.htm There are research renderers that implement general raytracing on the GPU, but I haven't seen any implementation that would be anywhere near production ready. The most usable GPU assisted renderer these days is probably Gelato.

Even if it only saves minutes from rendertime it'll still be useful for people who don't have access to renderfarm no? The examples given in that site looks pretty good too, good enough for a smaller scale archvis project at least.
http://www.bee-www.com/parthenon/stairs.png
Now, if only the big boys implement something like this in the big name renderers, it'll be great. Maxwell in particular would benefit a lot from it.

Hoppergrass
06-14-2006, 03:48 AM
Even if it only saves minutes from rendertime it'll still be useful for people who don't have access to renderfarm no? The examples given in that site looks pretty good too, good enough for a smaller scale archvis project at least.
http://www.bee-www.com/parthenon/stairs.png
Now, if only the big boys implement something like this in the big name renderers, it'll be great. Maxwell in particular would benefit a lot from it.

Sorry but I am not impressed by examples such as these, simple scenes like these should render in minutes with ease. To impress me show me a fully detailed room with obects of both large and small scale, detailed shaders and tricky nooks and crannies to get light into on top of that render it to a decent resolution (nothing less that A3 @ 300dpi). Render that in minutes and I will get my wallet out.

I am looking forward to the release

JHV

CupOWonton
06-14-2006, 03:51 AM
Sorry but I am not impressed by examples such as these, simple scenes like these should render in minutes with ease. To impress me show me a fully detailed room with obects of both large and small scale, detailed shaders and tricky nooks and crannies to get light into on top of that render it to a decent resolution (nothing less that A3 @ 300dpi). Render that in minutes and I will get my wallet out.

I am looking forward to the release

JHV
Yeah, but it seems that most of these images were set to render for 1 minute.
I think he should get ahold of some of the full scenes I believe on Evermotion and place some nice materials in them and render that for a good demo.

P_T
06-14-2006, 04:29 AM
Sorry but I am not impressed by examples such as these, simple scenes like these should render in minutes with ease. To impress me show me a fully detailed room with obects of both large and small scale, detailed shaders and tricky nooks and crannies to get light into on top of that render it to a decent resolution (nothing less that A3 @ 300dpi). Render that in minutes and I will get my wallet out.

I am looking forward to the release

JHV
Well to be fair, it's been developed since 2002, who knows how old those examples are and he did said they were rendered within few minutes, wish he would specify the rendertime though. That fat Buddha is a high poly 3D scan IIRC, hundreds of thousands of poly rendered with GI and blurred reflection in a few minutes without the typical Monte Carlo noise is great a few years ago no?

You don't need to get your wallet out, it's a free renderer that you can dload from here (http://www.bee-www.com/parthenon/Parthenon.zip). Why don't you give it a try?

Now like I said, if Mental Image or Pixar implement something like this in their renderers it'll be great.

gent_k
06-14-2006, 08:44 AM
Quote:
Originally Posted by mustique
Good news. Whish they will add faster motionblur and DOF too.



amen to that!

-"rasterizer performance" is listed there, hope the improvement will be substantial.

albedo4800hp
06-14-2006, 09:05 AM
Now, if only the big boys implement something like this in the big name renderers, it'll be great. Maxwell in particular would benefit a lot from it.
Nope that will never happen because Maxwell uses a completely different approach it does not use final gathering at all which they perseve as a biased method which is true AFAIK.

Jadetiger
06-14-2006, 09:23 AM
Incoming newb question: If I bought Maya 7.0 recently, how hard/expensive would it be to upgrade to the latest mental ray? Is it even worth it? I'm a student so I doubt its very necessary, I'm just curious thouhg.

Mic_Ma
06-14-2006, 11:13 AM
Originally Posted by mustique
Good news. Whish they will add faster motionblur and DOF too.


But if:

Our next generation software rendering technology that is built into RealityServer, achieves rendering times of less than 40 milliseconds per frame (i.e., realtime) with full raytracing (and very soon full global illumination) on small 64-bit commodity multiprocessor machines or blades on scenes with tens of millions of polygons.


Then what is the point of faster motion blur?

P_T
06-14-2006, 11:23 AM
Nope that will never happen because Maxwell uses a completely different approach it does not use final gathering at all which they perseve as a biased method which is true AFAIK.

I didn't mean the FG, I meant the GPU assisted rendering.

mustique
06-14-2006, 12:12 PM
But if:

...

Then what is the point of faster motion blur?

It's not just about motionblur, it's about motionblur + DOF.
Eventually MB + DOF + Displacements + Dmap shadows.

If MR 3.5 will accellerate those too, I'm a happy camper! :)

pixelmonk
06-14-2006, 03:35 PM
Incoming newb question: If I bought Maya 7.0 recently, how hard/expensive would it be to upgrade to the latest mental ray? Is it even worth it? I'm a student so I doubt its very necessary, I'm just curious thouhg.

you have the latest mental ray that works with maya.

albedo4800hp
06-14-2006, 06:28 PM
I didn't mean the FG, I meant the GPU assisted rendering.
I repeat myself again there is not much it can really help in assisting in that process unless you go for hybrid rendering (mixing scanline with raytracing) or ray shading. Everything else is at the moment not just worth approaching

lazzhar
06-14-2006, 08:27 PM
I repeat myself again there is not much it can really help in assisting in that process unless you go for hybrid rendering (mixing scanline with raytracing) or ray shading. Everything else is at the moment not just worth approaching

I'm no coder but I'm thinking that distributing jobs over multiple processors would only slow down renderings. If they to optimize the already algorithms to use the power of single units that would be more interesting IMO , especially the fact chipsets are getting faster and stronger.

P_T
06-14-2006, 08:40 PM
I repeat myself again there is not much it can really help in assisting in that process unless you go for hybrid rendering (mixing scanline with raytracing) or ray shading. Everything else is at the moment not just worth approaching

Mate, I only posted that in reponse to your previous post which sounded like you were implying that I suggested Maxwell to use FG in their renderer which I didn't.

I don't know anything about the technology behind it nor do I give a flyin rat's ass what method they use if MR or PRman or whatever company decides to use GPU together with CPU in their renderers, as long as the result is good quality with faster rendertime I'm a happy camper.

I'm just saying it'll be great if they can utilise GPUs for rendering as well as viewport display and I think those renderers, parthenon, gelato etc. is a step in the right direction.

CupOWonton
06-14-2006, 09:20 PM
I'm just saying it'll be great if they can utilise GPUs for rendering as well as viewport display and I think those renderers, parthenon, gelato etc. is a step in the right direction.
Ah, now see, thats creative thinking.
Maybe these renderers dont have to use it for a finalised render. Maybe they can just incorperate a GPU preview/viewport renderer that's set up to emulate the settings of the final renderer. Renderers as incredibly slow as say... Maxwell, would greatly improve through the use of a GPU preview. Though, still, the final product is VERY clean for 1 minute renders on parthenon alone, I need a new video card so I can test it out on a large scene.

Jadetiger
06-14-2006, 09:20 PM
you have the latest mental ray that works with maya.

oh, hehe, my bad ^^; my newbie underbelly is showing

Strang
07-04-2006, 12:02 AM
here is a pdf about mental mill, goes into pretty good detail about mental mill's capability

http://www.mentalimages.com/2_4_mentalmill/mental_mill_functional_overview-Januar4-2006.pdf

enygma
07-04-2006, 06:48 AM
i think these guys are doing more than just putting scanline jobs on the GPU

http://www.openrt.de/

but i dont know
I think the furthest they've gone to using special hardware for OpenRT is by programming an FPGA to do its ray processing. I haven't really heard of any instance of them using a GPU to do any real time ray tracing. The FPGA based ray-tracer was demoed at SIGGRAPH 2005 during a paper session. Mercury Computers also had a Cell based system running OpenRT on there as well at pretty steady speeds.

Lorecanth
07-04-2006, 08:48 AM
But if there's no way to access it beyond programming or using the standalone version, might it just as well not be there since 95% of its users can't get to it?

- Neil

Well that depends on how good and dedicated your user base is. Honestly I havn't seen the commitment that francesca luce, puppet, pixaro, or any of the other wonderful shader writers have put in, in ANY other render community. Not only have they extended the functionality of mental ray they have really pushed those wonderful german bastards hard into getting the functionality thats wanted.

The reality is that this is very very hard stuff. My honest opinion is that 3d artists should have a fundemental understanding of our tools. After all what happens when a director says I want this reflection to be placed over here where its not physically accurate, and by god you can't do that in maxwell or vray. One button solutions are great for some problems but it doesn't beat a thorough understanding of the algorithms and processes involved in generating imagery.

Davinci probably complained about having to study the interaction with various types of oil paints on canvas and the years of practice it took to understand the concepts and techniques. That doesn't mean he didn't always strive to learn.

playmesumch00ns
07-04-2006, 09:58 AM
Oh deary me. Rolf Herken is nothing if not a great self-promoter.

Notice how he says "soon" and "working towards" a lot. i.e. there's nothing revolutionary at the moment at all.

So they can raytrace scenes of millions of polygons in realtime on a 64-bit (read: 16GB of RAM) multiprocessor (read quad- or oct-core) "commodity" machine. Big woop. :-s. Do it on a dual-core laptop with 2GB of RAM and I'd be impressed.

mental ray tries to be a jack-of-all-trades and doesn't really succeed. It's not a production renderer. Vray is a faster approximating GI engine. Maxwell will soon have the market sewn up for those who care about physical accuracy.

I'm looking forward to seeing what they can do with mental ray 4.0. Hopefully they'll take the opportunity to completely trash all their legacy code and interface and just do something new. They've got some incredibly clever people working there, it's a shame they're tied down working on this ugly beast that is mr 3.x

rendermaniac
07-04-2006, 10:14 AM
here is a pdf about mental mill, goes into pretty good detail about mental mill's capability

http://www.mentalimages.com/2_4_mentalmill/mental_mill_functional_overview-Januar4-2006.pdf

Very interesting. This is definitely the direction shaders should be heading. I guess they have no incentive to target RenderMan, which is a shame, but I guess it would at least be possible.

Simon

earlyworm
07-04-2006, 11:15 AM
mental ray tries to be a jack-of-all-trades and doesn't really succeed. It's not a production renderer.

Ofcourse what wasn't clearly stated in the interview is that mental ray isn't actually used to render images at ILM, it's used as a coffee machine to keep TDs "rendering" away.

Nothing like millions of litres of coffee produced in realtime. Good on ya' mental images.

And with the new mental mill everyone will have easy access to all of mental ray's coffee making features.

So they can raytrace scenes of millions of polygons in realtime on a 64-bit (read: 16GB of RAM) multiprocessor (read quad- or oct-core) "commodity" machine. Big woop

Agreed. I won't be impressed until I can render ONE BILLION (capitalized for effect) polygons in realtime on my PDA. Muhahaha.

playmesumch00ns
07-04-2006, 11:29 AM
Very interesting. This is definitely the direction shaders should be heading. I guess they have no incentive to target RenderMan, which is a shame, but I guess it would at least be possible.

Simon

It seems like it will be perfectly possible to target RSL for oneself.

I have serious reservations about the idea of having one uber-language that would simultaneously compile to phen, Cg and RSL. However, mental mill itself does look like it could be a great shader designer, especially with the debugger.

Kabab
07-04-2006, 11:33 AM
mental ray tries to be a jack-of-all-trades and doesn't really succeed. It's not a production renderer. Vray is a faster approximating GI
To bad mental ray comes free with Max, Maya and XSI hrmm whats that about 90% of the market ?

With that amount of market penetration/revenue i would think Menta Ray is in a far better postion to win this rendering war then any other product.

DotPainter
07-04-2006, 12:25 PM
The document is interesting, but it doesnt address the implications at the architectural level, in terms of hardware support for multiple image buffers and swapping data back and forth from the gpu to the cpu in order to create layers and passes that get composed into a final image. This is a RAW set of core features that show what mental ray will have under the hood, but anyone that has followed the development of mental ray will know that the integration of mental ray into 3rd party applications is a mixed bag. All applications dont support all the features of mental ray, even in its current generation. So it remains to be seen how much of an impact this will have in terms of the 3d apps that use mental ray. How many of them will port all of their shaders to the new MR shader language? How many will allow for mixed gpu/cpu rendering? All of this would require much more in terms of coding on the part of 3d apps in integrating these features fully into the application. Of course, it may be that none of these advanced features are integrated into most 3d apps using MR, with the main benefit being that you have a nice tool that can write MR shaders that can be used in almost any 3d package running MR, which is not necessarily the case today.

playmesumch00ns
07-04-2006, 01:36 PM
To bad mental ray comes free with Max, Maya and XSI hrmm whats that about 90% of the market ?

With that amount of market penetration/revenue i would think Menta Ray is in a far better postion to win this rendering war then any other product.

I wasn't talking about market saturation, I was talking about whether or not it's actually any good.

CaptainObvious
07-04-2006, 02:47 PM
mental ray tries to be a jack-of-all-trades and doesn't really succeed. It's not a production renderer. Vray is a faster approximating GI engine. Maxwell will soon have the market sewn up for those who care about physical accuracy.
It sure sees an awful lot of use in production for something that isn't a production renderer...

CupOWonton
07-04-2006, 06:16 PM
Originally Posted by playmesumch00ns
mental ray tries to be a jack-of-all-trades and doesn't really succeed. It's not a production renderer. Vray is a faster approximating GI engine. Maxwell will soon have the market sewn up for those who care about physical accuracy.

It sure sees an awful lot of use in production for something that isn't a production renderer...
I like Vray a lot, lately Ive seen that MentalRay realy does have some speed to it when used properly, but Maxwell is still a freakin joke. Photons are technicaly acurate, maxwell just wastes unnessisary time resampling pixels to get its image, rather than a lot of shortcuts that come out with nearly the exact same image minus unnessisary caustics etc. Whats the point of 90% physicaly acurate rendering with Maxwell if no one's going to notice when compared to Vray or Mental Ray? It isnt saving overall time, thats for sure.

I still wish I got better feedback from the people at MentalImages about some of the faulty problems with MR and Photometric area lights in max, it leaves dots. They need to create a light like Vray's which can intersect geometry and still cast light properly, and have a viewable light plane. The photometric area light in max isnt visible, I end up wiring a plane to it.

Ryan-B
07-04-2006, 08:27 PM
I still wish I got better feedback from the people at MentalImages about some of the faulty problems with MR and Photometric area lights in max, it leaves dots. They need to create a light like Vray's which can intersect geometry and still cast light properly, and have a viewable light plane. The photometric area light in max isnt visible, I end up wiring a plane to it.

You can increase the area sampling within the parameters of each area light. This may help with leaving "dots". Also, you may need to move the light further away from the surfaces being lit if the light is too close and the pattern of the sampling is visible.

The shape of the photometric area light has always been visible to me and visible on every copy of max I've ever used.

BillSpradlin
07-04-2006, 10:54 PM
You should get with the people at Autodesk if you have problems with mental ray for Max, or join the mental ray maililng list on highend3d and ask your question there. Many mental images people read that list and some extremely experienced mental ray users post there.

BillSpradlin
07-04-2006, 11:20 PM
Oh deary me. Rolf Herken is nothing if not a great self-promoter.

Notice how he says "soon" and "working towards" a lot. i.e. there's nothing revolutionary at the moment at all.

So they can raytrace scenes of millions of polygons in realtime on a 64-bit (read: 16GB of RAM) multiprocessor (read quad- or oct-core) "commodity" machine. Big woop. :-s. Do it on a dual-core laptop with 2GB of RAM and I'd be impressed.

mental ray tries to be a jack-of-all-trades and doesn't really succeed. It's not a production renderer. Vray is a faster approximating GI engine. Maxwell will soon have the market sewn up for those who care about physical accuracy.

I'm looking forward to seeing what they can do with mental ray 4.0. Hopefully they'll take the opportunity to completely trash all their legacy code and interface and just do something new. They've got some incredibly clever people working there, it's a shame they're tied down working on this ugly beast that is mr 3.x

I wasn't talking about market saturation, I was talking about whether or not it's actually any good.

You're right, mental ray isn't a production renderer, it hasn't been used on a PLETHORA of feature films, from the Matrix movies, Shark Tale, Day After Tomorrow (which it was used on for the longest photo realistic full CG scene in a live action film to date: opening sequence), Spiderman 1, 2, Hulk and the list goes on to a TON of features in production right now. You obviously have drifted off into total fanboyism of your favorite renderer Renderman and appear to be threatened by more advacement in mental ray. It's rather amusing to see such statements made by someone who's been in the industry as long as yourself and you of all people should know, that you use the tool that works best for the situation. I think it's great to see that they are "working toward" advancements and can't wait to see what they come up with.

Studios choose the tool that best fits the need of the work which is why a lot of studios do not rely soley on one renderering system to get jobs done. If I can do something better in Renderman then I'll do it, if I can do it better with mental ray, then I'll use mental ray. Simple as that.

Edit: And I'm not even going to get into Maxwell....they have a LOOOOONG way to go before I'll EVER use it in production, one button solutions are not viable for the kind of work I do.

CupOWonton
07-05-2006, 12:34 AM
You can increase the area sampling within the parameters of each area light. This may help with leaving "dots". Also, you may need to move the light further away from the surfaces being lit if the light is too close and the pattern of the sampling is visible.

The shape of the photometric area light has always been visible to me and visible on every copy of max I've ever used.
Increasing the sampling increases the render time quite a bit. Ive tried suggesting they take a look at their competition and how they construct lights, *cough Vray cough* and maybe get someone over at autodesk to make one or something. Vrays area lights are incredibly smooth, I hope they add that into 3.5 and autodesk actualy takes TIME TO APPLY EVERYTHING AVAILABLE IN MENTAL RAY INTO MAX INSTEAD OF HIDING THEIR SHORTCOMMINGS with "hidden" tags :banghead:

Lorecanth
07-05-2006, 12:45 AM
I'm with Bill...all i know is with francesca's buffer system, Garramuño's mrLiquid, puppets megatk I get great results. Even in a small production enviroment.

As far as production tested ? I personally finished nearly 60 MINUTES of technical animation that would have been impossible on vray, brazil, maxwell or any of the other small team renderers. The render times for the desired images were far too high. I effectivly rendered it in mental ray in 3 months on 20 procs. They just don't have the community, that puts out wonderful things like ctrl openexr buffer output system that are really the tools that let a pipeline be built for consistent imagery.

I'm looking forward to renderers like maxwell, but it cracks me up that people are going to throw away the tools and knowledge that will let them move that reflection a quarter of an inch over when the client MUST have it that way, even if its not physically accurate.

CupOWonton
07-05-2006, 01:04 AM
So are we in agreement that MentalImages should put out easy to understand instructions on how everything in MR works? because I think it lies to me in the max version when it tells me photons will be 1/100th of the scene size by default... well if Im rendering at 1 inch i get huge photons, but if I render at 1 foot, i get tiny ones. 1/100th apparently is a lie... You have to restrict the size in max to get them to shrink down for rendering an exact size gold ring. All the demo-renders of gold rings and glass caustics are built at like.. 10Feet, theyre huge.

ajsfuxor
07-05-2006, 01:09 AM
hehehe...couldnt have said it better myself Bill.

And Maxwell, well sure its nice for a still here and there. But for animation, no chance. Its a loooooooong way off.

Mental Ray has such a huge community as well, so any problem you have can be solved in a matter of minutes. People often complain about it being so slow to render and complex. Well if people did some homework, and render in layers, do DOF and Motion Blur in post, Mental Ray is very hard to beat. This isnt to say that other renderers are bad, i dont want to start a flame war. Theres a good reason that Renderman, Brazil and Vray etc has been used on countless shots. The point of what i think Bill is saying that all renderers have their good points. Having such a closed mindset about renderers can only lead to problems in production. The best renderer for the job I say.

playmesumch00ns
07-05-2006, 10:14 AM
You're right, mental ray isn't a production renderer, it hasn't been used on a PLETHORA of feature films, from the Matrix movies, Shark Tale, Day After Tomorrow (which it was used on for the longest photo realistic full CG scene in a live action film to date: opening sequence), Spiderman 1, 2, Hulk and the list goes on to a TON of features in production right now. You obviously have drifted off into total fanboyism of your favorite renderer Renderman

Hehehe. You're right, I was trolling a bit. Sorry :) Seriously tho, have you spoken to anyone who used mr at ESC? Not fun at all.

I don't think mental ray's a bad renderer and I'm not denying that it has been used to render whole shots -- i.e. not just being used to do the odd occlusion or final gather pass like in some of those examples you cite. My point was that in all the fields where it tries to compete, there's a better solution available.


EDIT: sorry I'm verging on flame-baiting again. I'll shut up now.

gent_k
07-05-2006, 12:04 PM
Ya, and we all know how fun Renderman is to use. :rolleyes:

rendermaniac
07-05-2006, 01:10 PM
Hehehe. You're right, I was trolling a bit. Sorry :) Seriously tho, have you spoken to anyone who used mr at ESC? Not fun at all.

I don't think mental ray's a bad renderer and I'm not denying that it has been used to render whole shots -- i.e. not just being used to do the odd occlusion or final gather pass like in some of those examples you cite. My point was that in all the fields where it tries to compete, there's a better solution available.


EDIT: sorry I'm verging on flame-baiting again. I'll shut up now.

I suspect the Mental Ray hostility comes from trying to use it for production shots with a lot of water in them ;)

ESC were a bit masochistic - they had a Windows renderfarm for a start!

There are two main reasons Mental Ray is the most popular renderer - firstly the price, and secondly it is integrated in most software packages. This makes it easier to get up and running quicker and more people are likely to have used it.

Ironically there is probably more useful information on the internet than in the manuals! (renderman is the opposite).

Mental Ray excels at the fiddly bits - raytracing, baking, subsurface scatter etc, but fall down on the basics - most notably motion blur. I can see motion blur being a real issue with a big transparent/refractive particles/fluid/volumetric systems - very difficult to do in 2d.

Simon

dagon1978
07-05-2006, 04:54 PM
Increasing the sampling increases the render time quite a bit. Ive tried suggesting they take a look at their competition and how they construct lights, *cough Vray cough* and maybe get someone over at autodesk to make one or something. Vrays area lights are incredibly smooth, I hope they add that into 3.5 and autodesk actualy takes TIME TO APPLY EVERYTHING AVAILABLE IN MENTAL RAY INTO MAX INSTEAD OF HIDING THEIR SHORTCOMMINGS with "hidden" tags :banghead:

actually mray area lights work very well and fast in many case, probably faster then the vraylights, if you dont have geometries too close, in this case the vraylights are very powerful

you need to try a good integration of mental ray if you want to compare it to vray (wich is a great renderer), in maya the mray are lights are way better implemented then in max, and you can also use the phisical mray light (very similar to the vraylight) and the great ctrl.geolights.

and yeah, i hope we can have a good mray for max too, it's the time to wake up autodesk!

CGTalk Moderation
07-05-2006, 04:54 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.