PDA

View Full Version : testing finalRender stage-2 for maya


v-reality
06-18-2007, 11:18 AM
hi guys ,i'm a fan of cg rendering , i was workin on MR for maya ,......, but now i'm really tired about this engine and it's hard and complicated workflow , i like hard things but not anymore , cause i need quick results ,especially in setup time , so i think final render is the solution , let's test it

http://img240.imageshack.us/img240/5553/studioshaderstest015525lm0.jpg (http://imageshack.us)

render time : 55:25
setup time 10min

coccosoids
06-18-2007, 01:10 PM
you rendered that in 55 minutes? :) are you still using 486?
It seems waaaay too much for your scene.

v-reality
06-18-2007, 01:38 PM
hum , i'm using a p4 3.0 ghz , i'm trying to aptimize render time

v-reality
06-18-2007, 01:44 PM
hum , i'm using a p4 3.0 ghz , i'm trying to aptimize render time

Bmoner
06-18-2007, 01:54 PM
hum , i'm using a p4 3.0 ghz , i'm trying to aptimize render time

Still what q_vazk is trying to say is that THAT scene should NEVER have taken 55min to render...even with a P4 3.0Ghz. In Mental ray that would've taken all but 5min tops (and that's with FG, GI and super refractive surfaces).

Your best bet to learn Mental ray quickly (but efficiently) is to get Digital-Tutors line of DVDs that focus on Mental ray for Maya. They cover everything for letting you kn ow what all the nodes do, to teaching you how to navigate it's render stat screen (and how to work with it), as well as things to do to drop your rendertime yet still have stunning results.

Mental Ray, in my opinion, will always be a great renderer. Not as long render times as maxwell, not as many issues as finalrender...and costs less than renderman (Because it's free!). It's even been prioduction proven by being used in movies, so check that DVD line out from DTs.

v-reality
06-18-2007, 02:46 PM
here a mental ray test
time : 06:47
http://img74.imageshack.us/img74/6079/mrstudioshaderstest0106jf7.jpg (http://imageshack.us)

coccosoids
06-18-2007, 02:47 PM
Bmoner, hey thanks! That's exactly what I meant, I actually had a p4 3ghz and I never had
to wait more than 10 minutes for a render like that, without much optimizing either.

But, have you actually tried final render for maya? I heard it was faster than mr, even had
a proprietary tesselator for nurbs surfaces.... Can you say something regarding it's GI, FG
implementations and speed? :) Thanks...

vreality, :) I see you've proved yourself that MR is capable of same levels of quality even
in lesser time.

v-reality
06-18-2007, 02:50 PM
here another mental ray test with blurry reflexions ...

time :11:27

http://img101.imageshack.us/img101/3848/mrstudioshaderstest0211gy0.jpg (http://imageshack.us)

Bmoner
06-18-2007, 03:34 PM
Bmoner, hey thanks! That's exactly what I meant, I actually had a p4 3ghz and I never had
to wait more than 10 minutes for a render like that, without much optimizing either.

finalrender seems to have good FG and GI, but also not a lot of control of it (as far as I'm seeing).

sacslacker
06-18-2007, 03:40 PM
It's definitely one of the fastest and nicest GI engines out there. I presonnally love it. However, it doesn't quite reach mental ray. Especially in the stability area. Frankly I'm not too hip on Mental Ray's stability either lately.

v-reality
06-18-2007, 03:54 PM
now have to test the IBL in final render , cause if i remember well i used to got a lot of artefacts in MR , high redertimes and setup times

v-reality
06-18-2007, 04:06 PM
finalrender seems to have good FG and GI, but also not a lot of control of it (as far as I'm seeing).

don't forget the TONE MAPPING wich is very useful for interiors that we fond in finalRender

joie
06-19-2007, 09:30 AM
Things I love in FinalRender...

The hability to render passes from one render and production ready, no setup at all with any kind of shader (MR canīt do this).

Better memmory ussage, I have a scene really big with lots of referenced buildings (something like 10 millions polygons), MR canīt render it..., Final Render renders it without any problems at all.

Onlny with that two things, I think Iīll go for Final Render.

Saturn
06-19-2007, 10:49 AM
Things I love in FinalRender...

The hability to render passes from one render and production ready, no setup at all with any kind of shader (MR canīt do this).


sorry not true. You can output multiple passes in one render. It's called framebuffer.


Better memmory ussage, I have a scene really big with lots of referenced buildings (something like 10 millions polygons), MR canīt render it..., Final Render renders it without any problems at all.


Again not true. Average scenes size is 14 millions poly here ( lot of CAD cars ). We don't even bother to optimise polycount anymore.


Onlny with that two things, I think Iīll go for Final Render.

Maybe you should use MR outside maya ( XSI to name only one ) you may change your opinions.
But If you feel more confortable with FR then fair enough !

coccosoids
06-19-2007, 11:08 AM
Saturn: you mean you can render out passes from mental ray alone?? No custom shaders,
no custom geometry shaders? If that's true could you post an example of some kind?

InTerceptoV
06-19-2007, 03:24 PM
sorry not true. You can output multiple passes in one render. It's called framebuffer.

Yes and it is complicated to setup and as far as I can remember you need to have custom shaders. The way finalRender works is far more user friendly ... everything you need to do is just one click away. Check some other render engines like turtle, vray or renderman for maya and you will see how others do the passes.

Again not true. Average scenes size is 14 millions poly here ( lot of CAD cars ). We don't even bother to optimise polycount anymore.

On what machine ... with 4 gigs of ram? Try to put 14 milion instanced polys in 2gb machine and render. It will crash for sure ... and you can do this in batch rendering. For example you can easily put a hundred of instanced trees in finalrender and render without the problem while on mr it will crash when you get over the limit of about 10 milions. MR in maya just can't handle instances very well, and that is the only truth.

Maybe you should use MR outside maya ( XSI to name only one ) you may change your opinions.
But If you feel more confortable with FR then fair enough !

So what are you suggesting, to leave maya and get to XSI? I know that in XSI mentalray has the best integration and that we will probably wait for years to reach that level in maya, but if you use maya you have alternatives in render engines like fR, Renderman, Vray etc... so pick one that you like!

Michael-McCarthy
06-19-2007, 03:38 PM
Heya Guys,

I saw some talk on render elements and other fR features. Please take a look at these feature videos if you have not yet (flash required):

http://www.cebasusa.com/finalRender%20Stage-2%20for%20Maya/finalRender%20for%20Maya%20Stage-2.html

There is a section on Render Elements that shows the setup and ease of use. Lots of studios are loving the RE 's in fR right now (among many other things) :)

PS: V-reality, you might post that scene up at the cebas fR for Maya forum. It should render in only a few minutes so maybe some users can give you pointers, or maybe a setting or 2 is out of wack. Lots of nice and helpful guys there too.

Thanks
Michael McCarthy
cebas

Bmoner
06-19-2007, 03:40 PM
Well hell, render elements alone (like the ones offered in Max) is enough for me to switch.

A render-layer free scene collection is a dream come true.

lazzhar
06-19-2007, 06:47 PM
Beside what have been mentioned, I find shaodws in FR area lights much more cleaner and faster than in MR. This might turn into a rendering engine war but FR 4 maya made a giant step forward while MR is going slowly cough cough... But the misss_fast thingies in MR are really fast and good to forget about them.

techmage
06-20-2007, 09:41 PM
Just to point out something I saw when evaluating finalRender. I found it be very nice but I don't know if it was because of my inexperience with it. But all the renders seemed to have a very CG-ish type look to them. I can't quite put my finger on what exactly it is. But like when I look at a GI render, I first look at it overall to see if the general color distribution and presence of various things is realistic. Then I look up close at it in a technical sort of way to see how good contact shadows are and how much the radiosity is doing or if theres blurred reflections contributing to it. FinalRender from the technical eye was awesome, it gets GI into places where MR just couldn't and does it more efficiently, smoothly and with perfect contact shadows. But when I sit back and look at it overall finalRender never seemed to get as good of a realistic feeling as mental ray, vray or maxwell. Even though from the technical point of view the GI was perfect, shaders were perfect and I couldn't really tell what was off exactly, it didn't feel perfect overall.

Heres some good example from the cebas image gallery:
http://www.finalrender.com/gallery/v/interior/AlexandreJesus_patio_060607.jpg.html
http://www.finalrender.com/gallery/v/interior/Salone04_800.jpg.html

If you analyze those, technically they are great. The GI calculation gets into every little crack and is perfectly smooth. The shaders are good for their surface. But then look at the overall thing and they are very very obviously CG, even more so than some images that don't even have perfect GI. I don't know what exactly it is but I see alot of stuff coming out of finalRender with this quality on their image forum as well. I want to think it has something to do with the GI calculation which may not be as true to physics as others. But I really have no idea. Or like I said, if it's me just not knowing how to set it up the best because I have seen some excellent architectual stuff out of finalRender.

But I will say, the finalRender workflow is way better and it does render some effects better.

Also, if your looking for mental ray alternatives, don't overlook turtle. Theres two things very notable about it, it's integration into Maya is one of the best, turtle is pretty much designed to match Maya. Also it is notably faster in alot of cases. You may not see awesome interior archviz's coming out it right now. But from the little bit I've played with it so far, it does seem very capable if you took the time to figure out how to set everything up the best.

Michael-McCarthy
06-20-2007, 10:52 PM
Thanks for your insites rygoody

I would have to say this is very much dependant on the Artist. We have many images in our gallery from all different skill levels.

I would suggest you take a look at a few of the images here (more recent R2 stuff) these first 15 or so are quite good (although they scaled for the web):
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Stage1R2_3dsMax/
some of these are larger and clearer here:
http://www.finalrender.com/news/read.php?UD=10-7888-35-788&NID=207

Some great Stage-2 for Maya images (again better at full res, some of these like Toni's images can be found on cgtalk at better res):
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Maya+Images/dawn08_1_.jpg.html
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Maya+Images/1enkelt_0558_1_.jpg.html
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Maya+Images/bioTestSmall0001_1_.jpg.html

There are more Stage-2 for C4D then Maya (as it came out earlier) here:
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/Sebastian_Florand_salon_HD.jpg.html
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/JanneHellmann_Watch2.jpg.html
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/Headphones.bmp.html
http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/moka-reeperbahn.jpg.html

We are certainly proud to make a powerful, fast, and well integrated renderer. I do have to say it is always the fantastic artists that make Amazing artwork! :)

Thanks
Michael McCarthy
cebas

dagon1978
06-20-2007, 11:17 PM
On what machine ... with 4 gigs of ram? Try to put 14 milion instanced polys in 2gb machine and render. It will crash for sure ... and you can do this in batch rendering. For example you can easily put a hundred of instanced trees in finalrender and render without the problem while on mr it will crash when you get over the limit of about 10 milions. MR in maya just can't handle instances very well, and that is the only truth.


absolutely false

here's a test a made in my 2Gb pc some time ago ;)


maya 8.5 mray 3.5

270 ajax x 544566 polys = 147.032.820 polygons
http://img213.imageshack.us/img213/8881/ajax147000000polyskb0.jpg

raytracing
http://img213.imageshack.us/img213/7100/ajax13mlp2.jpg

FG + AO + raytracing
http://img213.imageshack.us/img213/6668/ajaxfg52mci5.jpg

and this is from max 9

13.612 teapots x 25.600 polys each one = 348.467.200 polygons!
http://img213.imageshack.us/img213/2949/teapot1fgmb2m20srt7.jpg

Bmoner
06-21-2007, 03:10 AM
I would love if they had a 64 bit demo version of the software.

yolao
06-21-2007, 04:17 AM
But all the renders seemed to have a very CG-ish type look to them. I can't quite put my finger on what exactly it is. But like when I look at a GI render, I first look at it overall to see if the general color distribution and presence of various things is realistic. Then I look up close at it in a technical sort of way to see how good contact shadows are and how much the radiosity is doing or if theres blurred reflections contributing to it. FinalRender from the technical eye was awesome, it gets GI into places where MR just couldn't and does it more efficiently, smoothly and with perfect contact shadows. But when I sit back and look at it overall finalRender never seemed to get as good of a realistic feeling as mental ray, vray or maxwell. Even though from the technical point of view the GI was perfect, shaders were perfect and I couldn't really tell what was off exactly, it didn't feel perfect overall.

Heres some good example from the cebas image gallery:
http://www.finalrender.com/gallery/v/interior/AlexandreJesus_patio_060607.jpg.html
http://www.finalrender.com/gallery/v/interior/Salone04_800.jpg.html


But I will say, the finalRender workflow is way better and it does render some effects better.



I agree, I haven`t use FR my self, but for some reason the renders from FR does not look as realistic as mental ray or vray....specially regarding to glass, metal reflections and skin SSS...

I would have to say this is very much dependant on the Artist.

Also completely agree!!...

I would suggest you take a look at a few of the images here (more recent R2 stuff) these first 15 or so are quite good (although they scaled for the web):
http://www.finalrender.com/gallery/...tage1R2_3dsMax/ (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Stage1R2_3dsMax/)
some of these are larger and clearer here:
http://www.finalrender.com/news/rea...-35-788&NID=207 (http://www.finalrender.com/news/read.php?UD=10-7888-35-788&NID=207)

Some great Stage-2 for Maya images (again better at full res, some of these like Toni's images can be found on cgtalk at better res):
http://www.finalrender.com/gallery/...n08_1_.jpg.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Maya+Images/dawn08_1_.jpg.html)
http://www.finalrender.com/gallery/...558_1_.jpg.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Maya+Images/1enkelt_0558_1_.jpg.html)
http://www.finalrender.com/gallery/...001_1_.jpg.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/Maya+Images/bioTestSmall0001_1_.jpg.html)

There are more Stage-2 for C4D then Maya (as it came out earlier) here:
http://www.finalrender.com/gallery/...lon_HD.jpg.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/Sebastian_Florand_salon_HD.jpg.html)
http://www.finalrender.com/gallery/...Watch2.jpg.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/JanneHellmann_Watch2.jpg.html)
http://www.finalrender.com/gallery/...phones.bmp.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/Headphones.bmp.html)
http://www.finalrender.com/gallery/...erbahn.jpg.html (http://www.finalrender.com/gallery/v/cebas+Software/finalRender/cinema4d/moka-reeperbahn.jpg.html)

Now this renders looks very very NICE, specially the Sebastian_Florand_salon_HD, but i haven`t seen yet a reallistic render on something organic, specially because of how the SSS looks, i mean, i have seen wonderful and beautiful examples of human character render on FR made by GREAT ARTISTS, but they don`t look very realistic to me....

Is FR more for achitecture and that kind of things than for organic work?.....

Thanks, and again some of the renders that you linked looks WONDERFUL.

Cheers

techmage
06-21-2007, 04:34 AM
@McCarthy, those are very nice. I'm thinking now what I talked about earlier could be due to lack of blurry reflections on things. But the main concern those pictures leave me with is, where is fr-fillet and fr-archi shader in maya?! :shrug:


and dagon, how did you get mental ray to render that much?! Are those instances? How did you get instances out of mayatomr? I have never managed to get mental ray to render more than like 2 million in any sort of reliable or respectable amount of time.

coccosoids
06-21-2007, 05:38 AM
Any demo for osX ?

InTerceptoV
06-21-2007, 06:36 AM
Again these funny simple tests that doesn't show any real situations when you create true scenes. Come one dagon, make some real scenes and try to get a real expirience. We done the forest with instanced polygons with mentalray sometime ago and it crashed. The turtle was able to render about 2 times the polygons mentaray has and finalRender doesn't have a true limit...

absolutely false

here's a test a made in my 2Gb pc some time ago ;)


maya 8.5 mray 3.5

270 ajax x 544566 polys = 147.032.820 polygons

InTerceptoV
06-21-2007, 09:04 AM
Hey Dagon is it possible to know rendering time for these?

I want to test exactly the same scene in both renderers so we can compare them.
I suppose you used large BSP in combination with low memory settings (500mb?) ... but I am wondering did you used placeholders for the objects or not?

Michael-McCarthy
06-21-2007, 11:32 AM
Now this renders looks very very NICE, specially the Sebastian_Florand_salon_HD, but i haven`t seen yet a reallistic render on something organic, specially because of how the SSS looks, i mean, i have seen wonderful and beautiful examples of human character render on FR made by GREAT ARTISTS, but they don`t look very realistic to me....

Is FR more for achitecture and that kind of things than for organic work?.....

Thanks, and again some of the renders that you linked looks WONDERFUL.

Cheers


Hello yolao,

Thanks for the incite. I agree Sebastian is a very talented artist and is continual producing great works.

Here are two great examples of organic characters in finalRender:
http://www.finalrender.com/products/feature.php?UD=10-7888-35-788&PID=36&FID=578
http://www.finalrender.com/products/feature.php?UD=10-7888-35-788&PID=45&FID=585

Loocas and Steve are both very talented artists.

Loocas also did a great free Video training segment on realistic skin creation using finalRender here:
http://forums.cgsociety.org/showthread.php?f=96&t=492825&highlight=finalrender+skin

finalRender is surely a great Architectural tool but it does amazing work with CG characters. One reason for this is its fast and flexible 3s methods. You can see the monster rendered from behind on the fast skin page. This is more of a technical demo then a beauty image. The character is high detail MTD, with 3S, effected by GI, that you can animate flicker free with the GI and 3S. This is quite a challenge and not possible at all, in some other software. I think just 3S with MTD is a challenge or not doable for some packages, let alone flicker free and effected by GI. Oh and it renders in 2 minutes 30 seconds a frame on a single 3400 + :D...

Also fR is very strong in FX animations. It has recently been used with great success for shots in Spiderman 3, Superman Returns, and a host of other great FX animation. This is largely do to its fast flicker free GI and flexible Render elements (studios love the many and versatile Render Ellements).

Hope that's not more then you needed to know :)

Thanks
Michael McCarthy

Saturn
06-21-2007, 11:33 AM
Yes and it is complicated to setup and as far as I can remember you need to have custom shaders. The way finalRender works is far more user friendly ... everything you need to do is just one click away. Check some other render engines like turtle, vray or renderman for maya and you will see how others do the passes.



On what machine ... with 4 gigs of ram? Try to put 14 milion instanced polys in 2gb machine and render. It will crash for sure ... and you can do this in batch rendering. For example you can easily put a hundred of instanced trees in finalrender and render without the problem while on mr it will crash when you get over the limit of about 10 milions. MR in maya just can't handle instances very well, and that is the only truth.



So what are you suggesting, to leave maya and get to XSI? I know that in XSI mentalray has the best integration and that we will probably wait for years to reach that level in maya, but if you use maya you have alternatives in render engines like fR, Renderman, Vray etc... so pick one that you like!

I don't have more than 2 gig. And actually I could render over 900 millions poly with 1 GB of ram. Ok it took 24 hours and was under XSI but still it s MR render.

Same Framebuffer are done in one click in XSI. So don't need to get a custom shader.

What I am saying is not to switch to XSI, or maybe yes, But it 's not a MR problem it 's a maya problem.

yolao
06-21-2007, 01:31 PM
Hello yolao,

Thanks for the incite. I agree Sebastian is a very talented artist and is continual producing great works.

Here are two great examples of organic characters in finalRender:
http://www.finalrender.com/products/feature.php?UD=10-7888-35-788&PID=36&FID=578
http://www.finalrender.com/products/feature.php?UD=10-7888-35-788&PID=45&FID=585

Loocas and Steve are both very talented artists.



Those renders looks very nice indeed, agree Loocas and Steve are wonderful artists.

-I was checking the way FR deals with render passes (elements) and is very impressive, but i haven`t seen yet a video showing the making of custom passes, which i know that FR of course can do, which will be the workflow to make a custom pass?

-Also wich will be the workflow to make a light pass for individual lights?

-Can you speak a bit of FR for Character animation...advantages over MR for maya?...

-And what about hair?...any plan to support shave and haircut...or maya hair?

FR really looks nice, congratulations on what you have done so far!!

Cheers

joie
06-21-2007, 01:53 PM
My problem with MR in this job (the one Iīm currently working on), is with very high texture resolutions, I have many 2048 pix size textures, and more of 4096 pix size textures, MR canīt handle those and fails to render with MAYA opened, I can only render them via batch render..., so I canīt make test renders in order to see whatīs going on...

So my problem is not about the millions of polys but with millions of TEXTURED polygons, and textured with very high resolution textures.

Finalrender can handle them perfectly without the need of being a super guru...

Michael-McCarthy
06-21-2007, 02:12 PM
Those renders looks very nice indeed, agree Loocas and Steve are wonderful artists.

-I was checking the way FR deals with render passes (elements) and is very impressive, but i haven`t seen yet a video showing the making of custom passes, which i know that FR of course can do, which will be the workflow to make a custom pass?

-Also wich will be the workflow to make a light pass for individual lights?

-Can you speak a bit of FR for Character animation...advantages over MR for maya?...

-And what about hair?...any plan to support shave and haircut...or maya hair?

FR really looks nice, congratulations on what you have done so far!!

Cheers

fR for Maya has the ability to do custom render passes. These custom passes are pretty simple to set up, Just right click and add them and then choose what you want in them. They have inclusion and exclusion lists so can include and exclude what you like. Also we have added even more very cool control to the next service pack (SP3) which is in RC now and should be out soon!

As for character animations there is many advantages of fR. I would say the largest advantage would be flicker free GI and flicker free Skin shading, as well as the interaction between the 2. Also the fact that these things look very nice, are easy to setup, and render very fast.

We do plan and are currently working on Hair and fluids in Maya, so you will likely see this in the future.

Thanks
Michael McCarthy

yolao
06-21-2007, 02:43 PM
fR for Maya has the ability to do custom render passes. These custom passes are pretty simple to set up, Just right click and add them and then choose what you want in them. They have inclusion and exclusion lists so can include and exclude what you like. Also we have added even more very cool control to the next service pack (SP3) which is in RC now and should be out soon!

As for character animations there is many advantages of fR. I would say the largest advantage would be flicker free GI and flicker free Skin shading, as well as the interaction between the 2. Also the fact that these things look very nice, are easy to setup, and render very fast.

We do plan and are currently working on Hair and fluids in Maya, so you will likely see this in the future.

Thanks
Michael McCarthy

thanks Michael.

By seeing the videos FR seems to be far more intuitive and straight forward than other renders in the market, and that`s fantastic.

Looking forward the next FR for maya, and i`m glad to see that you guys are working in support hair and fluids in maya, that`s very very important.

I hope to see some videos showing the new features when the new version of FR for maya is realesed.

thanks again Michael, and thanks Joie for the info on textures.

Visor66
06-21-2007, 03:25 PM
Is there a demo-version for Maya 8.5? Couldnt find anything so far!

Thanks

dagon1978
06-21-2007, 03:33 PM
Again these funny simple tests that doesn't show any real situations when you create true scenes. Come one dagon, make some real scenes and try to get a real expirience. We done the forest with instanced polygons with mentalray sometime ago and it crashed. The turtle was able to render about 2 times the polygons mentaray has and finalRender doesn't have a true limit...

interceptov you are very funny for me ;)
you are talking about 14m polygons inteances and i showing you 300milion polygons renders!!

Try to put 14 milion instanced polys in 2gb machine and render. It will crash for sure ...
:rolleyes:

so... what are you triyng to say??


Hey Dagon is it possible to know rendering time for these?

I want to test exactly the same scene in both renderers so we can compare them.
I suppose you used large BSP in combination with low memory settings (500mb?) ... but I am wondering did you used placeholders for the objects or not?

this is a funny test, why do you wanna test it with fR? :rolleyes:
admit your sentence is completely false and maybe i can share this or another funny test with you




and dagon, how did you get mental ray to render that much?! Are those instances? How did you get instances out of mayatomr? I have never managed to get mental ray to render more than like 2 million in any sort of reliable or respectable amount of time.

yeah, absolutely instances... as interceptov say ;)
mray use maya instances, so i dont know what kind of problem you have...

Bmoner
06-21-2007, 03:36 PM
Is there a demo-version for Maya 8.5? Couldnt find anything so far!

Thanks

I think people are ignoring us on the demo subject lol.

Michael-McCarthy
06-21-2007, 03:44 PM
finalRender does not currently have a demo version. We are looking at options for this in the future. You can however feel free to contact us via the cebas shop or info@cebas.com (http://forums.cgsociety.org/info@cebas.com) for a full 15 day license. We do offer this but you may need to be patient, with all our new releases, siggraph etc it's a busy time :)

Thanks
Michael McCarthy
cebas

coccosoids
06-21-2007, 04:35 PM
Any versions for osX in the "making" ? :)

InTerceptoV
06-21-2007, 05:41 PM
interceptov you are very funny for me ;)
you are talking about 14m polygons inteances and i showing you 300milion polygons renders!!
:rolleyes:
so... what are you triyng to say??


That those tests with a lot of objects are a simple tests that aren't very usefull for some real situations. It is just different when you try to create something more complicated in 3D and many times mentalray showed his real face when scenes with many polygons, mostly instanced ones, came for render. Yes, many polys can be rendered with MR, but if takes 20 times more than in finalRender (for example), and it is so usefull that that it basically means MR can't do the job as it should, or can't do at all.


this is a funny test, why do you wanna test it with fR? :rolleyes:
admit your sentence is completely false and maybe i can share this or another funny test with you

Yes my sentence is wrong when you just want to ask if MR can render this. But I wanted to do a simple test to show you the time differences between them, just so you can see that MR doesn't know how to handle instances very well and is flushing so often that it slow down so much it became usefull.

The simple test is coming... and after that bye bye since I am going to vacations for 3 weeks... :-)

dagon1978
06-21-2007, 05:54 PM
That those tests with a lot of objects are a simple tests that aren't very usefull for some real situations. It is just different when you try to create something more complicated in 3D and many times mentalray showed his real face when scenes with many polygons, mostly instanced ones, came for render. Yes, many polys can be rendered with MR, but if takes 20 times more than in finalRender (for example), and it is so usefull that that it basically means MR can't do the job as it should, or can't do at all.


...or (most probably) you dont know HOW to do it right :rolleyes:


Yes my sentence is wrong when you just want to ask if MR can render this. But I wanted to do a simple test to show you the time differences between them, just so you can see that MR doesn't know how to handle instances very well and is flushing so often that it slow down so much it became usefull.

The simple test is coming... and after that bye bye since I am going to vacations for 3 weeks... :-)
mray can render this very well and very fast, the max test with more then 300M polys was 2m 20s, beating vray without problems, try this "funny" test (http://rapidshare.com/files/29241798/teierine_pesantine.rar.html) with your fR and let me know your rendertimes :rolleyes:

InTerceptoV
06-21-2007, 07:54 PM
Here is one test ... maya, finalrender.
These are instanced particles...

I just clicked on render and FR rendered the picture very fast...

http://interstation3d.com/cgtalk/fr_tests/fr_instances.jpg

I don't know why you send the scene in 3DSmax when we are talking about maya MR implementation?

With the options you have in maya, without addons to mentalray ... which means that you can use memory limit, bsp setup and possibly a placeholders, mr can't render this amount of geometry. Or, maybe I am wrong, so if you can give us the test scene with teapots instanced on particles that render in that amount of time so we can see it?

BTW see you in three weeks since I am at vacations from tomorow ... I am very interested to see what is the solution in mr...

Visor66
06-21-2007, 08:00 PM
Sounds as its getting more rough in here! ;)

InTerceptoV: Nice test! How about giving us the Maya-scene so we all can have a try! I would also like to test my system! That would be great!

Dagon: Are you running Maya on a 64Bit-System?


Keep on with the competition guys...there is always stuff to learn!

Greetz

Pascal

dagon1978
06-21-2007, 08:21 PM
Here is one test ... maya, finalrender.
These are instanced particles...

I just clicked on render and FR rendered the picture very fast...

http://interstation3d.com/cgtalk/fr_tests/fr_instances.jpg

I don't know why you send the scene in 3DSmax when we are talking about maya MR implementation?

cause i dont have the ajax scene here and why dont you send your scene? ;)


With the options you have in maya, without addons to mentalray ... which means that you can use memory limit, bsp setup and possibly a placeholders, mr can't render this amount of geometry. Or, maybe I am wrong, so if you can give us the test scene with teapots instanced on particles that render in that amount of time so we can see it?

BTW see you in three weeks since I am at vacations from tomorow ... I am very interested to see what is the solution in mr...
ahaha and why i can't use addons? i never know that was forbidden :)
but ok, i can do it without addons if you like it :)

dagon1978
06-21-2007, 08:23 PM
Dagon: Are you running Maya on a 64Bit-System?


not in these scenes ;)
now i'm using a 64bit system, but probably interceptov dont like i use a 64bit software :p so, i will use a 32bit one in my tests

Visor66
06-21-2007, 08:29 PM
not in these scenes ;)
now i'm using a 64bit system, but probably interceptov dont like i use a 64bit software :p so, i will use a 32bit one in my tests

hehe! Thats good, so I can probably test your scene with my 32bit system as well. Im looking forward to your tests! Your posts made me learn Mental Ray (at least, the first steps :thumbsup: ). Thanks for sharing

Pascal

InTerceptoV
06-21-2007, 09:29 PM
Ok still have some time till tomorow. :)

Here is the new test. Now all materials have the reflections...

http://interstation3d.com/cgtalk/fr_tests/fr_instances1.jpg

The scene is here. You can play it till end to get the image above or you can try with lower frames to get a little less...

http://interstation3d.com/cgtalk/fr_tests/instances01.rar

joie
06-21-2007, 10:16 PM
5,376,349,442 polygons?????, thatīs 5,376 millions polygons????, MAYA can handle that huge amount of geometry in viewport?, what the hell..., I canīt believe it 8O

BillSpradlin
06-22-2007, 01:28 AM
My problem with MR in this job (the one Iīm currently working on), is with very high texture resolutions, I have many 2048 pix size textures, and more of 4096 pix size textures, MR canīt handle those and fails to render with MAYA opened, I can only render them via batch render..., so I canīt make test renders in order to see whatīs going on...

So my problem is not about the millions of polys but with millions of TEXTURED polygons, and textured with very high resolution textures.

Finalrender can handle them perfectly without the need of being a super guru...

If you aren't converting your texture maps to .map format then yea, you are going to experience crashing. Mental ray can handle thousands of 4k+ maps if you set it up properly.

Myliobatidae
06-22-2007, 02:06 AM
...or (most probably) you dont know HOW to do it right :rolleyes:


mray can render this very well and very fast, the max test with more then 300M polys was 2m 20s, beating vray without problems, try this "funny" test (http://rapidshare.com/files/29241798/teierine_pesantine.rar.html) with your fR and let me know your rendertimes :rolleyes:

Ummm...that test scene that you rendered in 2m 20s, has been rendering on my computer for the last 6 hours with Mray and only 2 buckets are complete, not to mention its eating up 6 gigs of ram to do it...also as far as I can tell there are no lights in the scene casting shadows, or GI of any kind...is that correct ?

It crashed when I tried to render it on my 32bit machine, so I tried it on the 64bit machine, and of coarse realized why it crashed, its rediculus to be using that much ram just for this...

I'm going to render it with Vray, with lights and shadows and even GI, it will probably take 5 min and won't even use 1 gig of ram...300 million polys will make Vray Yawn, I've done well over 5 billion with no trouble at all...

Rerender that scene with lights and shadows and GI and let us know the results...

dagon1978
06-22-2007, 05:33 AM
Ummm...that test scene that you rendered in 2m 20s, has been rendering on my computer for the last 6 hours with Mray and only 2 buckets are complete, not to mention its eating up 6 gigs of ram to do it...also as far as I can tell there are no lights in the scene casting shadows, or GI of any kind...is that correct ?

It crashed when I tried to render it on my 32bit machine, so I tried it on the 64bit machine, and of coarse realized why it crashed, its rediculus to be using that much ram just for this...

I'm going to render it with Vray, with lights and shadows and even GI, it will probably take 5 min and won't even use 1 gig of ram...300 million polys will make Vray Yawn, I've done well over 5 billion with no trouble at all...

Rerender that scene with lights and shadows and GI and let us know the results...


that's because this is not my scene :D is the standard scene to test :D
here's mine (http://rapidshare.com/files/38641604/teierine_pesantine_mat.rar) ;)

you have skylight+FG+AO
without FG/AO the scene is rendered in 30s

InTerceptoV
06-22-2007, 05:54 AM
that's because this is not my scene :D is the standard scene to test :D
here's mine (http://rapidshare.com/files/38641604/teierine_pesantine_mat.rar) ;)

you have skylight+FG+AO
without FG/AO the scene is rendered in 30s

This is what I was talking about. The grid subdivision works great when everything is inline and homogenuous since it doesn't do adaptive voxel creation. But, it sucked when I tried to render my temple (check the portfolio) in decent amount of time. The fR, on other hand done this easy, but unfortunately I didn't have the ability to render that image in fR so i chosed to do it in 4 layers in mr...

I would like that we can do some more tests .... but I have to go now, really!. C U in 3 weeks... stay cool...

dagon1978
06-22-2007, 06:00 AM
oh, interceptov, nice to see you here again :D

so, your test...

i'm at a friend home to test with his fR4maya and mray... to campare rendertimes... so...



frame 20 obj 3168 polys 1.293.564.096

mr 47s
http://img98.imageshack.us/img98/8452/mrteapotf2047snt8.jpg

fR 1m49s
http://img98.imageshack.us/img98/1818/frteapotf201m49smy7.jpg



frame 30 obj 4835 polys 1.974.232.035

mr 1m16s
http://img98.imageshack.us/img98/4434/mrteapotf301m16syn4.jpg

fR 2m16s
http://img98.imageshack.us/img98/1563/frteapotf302m15sby8.jpg



frame 35 obj 5168 polys 2.110.202.928

mr 1m25s
http://img98.imageshack.us/img98/9818/mrteapotf351m25sxq7.jpg

fR 2m31s
http://img98.imageshack.us/img98/4192/frteapotf352m31szf6.jpg

dagon1978
06-22-2007, 06:00 AM
frame 40: obj 5668 polys 2.314.363.428

mr 1m33s
http://img98.imageshack.us/img98/4837/mrteapotf401m33sfz7.jpg

fR 2m50s
http://img207.imageshack.us/img207/3986/frteapotf402m57sxp9.jpg



frame 45 obj 7335 polys 2.995.034.535

mr 1m46s
http://img207.imageshack.us/img207/2568/mrteapotf451m46shp8.jpg

fr 3m47s
http://img207.imageshack.us/img207/9087/frteapotf453m47sgq1.jpg



frame 80 obj 12335 polys 5.036.639.535

fR 5m50s
http://img207.imageshack.us/img207/9625/frteapotf805m56seg0.jpg


mray cannot go over frame 45 in my system (ok, it could, but with very slow rendertimes)


conclusions:

mray is faster under 3 bilions polys, and i think 3.000M polys are a bit more then 14M polys, dont you think interceptov?
so, your sentece IS completely false


over 3bilions polys fR is actually faster, and (surprisingly) dont have really a poly limit for instances

so, if you want to render more then 3.000M polys (!!) you can buy fR... OR you can buy 2Gb more ram and get it in mray :p

dagon1978
06-22-2007, 06:08 AM
This is what I was talking about. The grid subdivision works great when everything is inline and homogenuous since it doesn't do adaptive voxel creation. But, it sucked when I tried to render my temple (check the portfolio) in decent amount of time. The fR, on other hand done this easy, but unfortunately I didn't have the ability to render that image in fR so i chosed to do it in 4 layers in mr...

I would like that we can do some more tests .... but I have to go now, really!. C U in 3 weeks... stay cool...

there are different solutions for different problems interceptov :scream:
see ya and good vacation ;)

InTerceptoV
06-22-2007, 06:08 AM
Nice tests. But now that I know you are using the GRID voxel subidivions of the scene I would like to make one more test. Unfortunately I don't have time any more so I'll let you know when I get back an we can continue...:thumbsup:

dagon1978
06-22-2007, 06:10 AM
Nice tests. But now that I know you are using the GRID voxel subidivions of the scene I would like to make one more test. Unfortunately I don't have time any more so I'll let you know when I get back an we can continue...:thumbsup:

ok :thumbsup: i'll wait

Myliobatidae
06-22-2007, 07:07 AM
Well I'm impressed, the scene you created worked great, but I don't understand what you did to make it work...

Visor66
06-22-2007, 09:32 AM
hey dagon,

nice tests! Could you give us some insight on how you made it work in Maya and probably the scene file so we all can have a look too?

That would be great! Thanks


Pascal

joie
06-22-2007, 10:04 AM
@ Bill Spradling: Yes, all my huge textures are already in .MAP format and it crashes everytime..., I can render in batch mode render but I cannot render it in the render view...

lazzhar
06-22-2007, 11:35 AM
The thing is that in FR it's in most case a matter of click render and see results while in MR you have always to go tweaking something here or there. But This would push us to ask why a a renderer that has been implemented since Maya 4.5 still causes a headache for simple users like us especially for starters while the recent one is much more easier to use and gives better results in certain areas or in general faster ?! I know MR comes for free with Maya and this is its big asset for now.

slipknot66
06-22-2007, 12:37 PM
Well I'm impressed, the scene you created worked great, but I don't understand what you did to make it work...

To render things like this with mental ray, its a good idea to use the grid acceleration instead of bsp or large bsp.

floze
06-22-2007, 12:40 PM
To render things like this with mental ray, its a good idea to use the grid acceleration instead of bsp or large bsp.
..though it does not support motion blur.

slipknot66
06-22-2007, 12:58 PM
Here is a test i did with an old project, 22 million polys, no instances , rendered in 1hour45min

http://img267.imageshack.us/img267/3117/santcugatbs9.th.jpg (http://img267.imageshack.us/my.php?image=santcugatbs9.jpg)

Visor66
06-22-2007, 01:02 PM
Here is a test i did with an old project, 22 million polys, no instances , rendered in 1hour45min


and you rendered that with Grid-Acceleration?

slipknot66
06-22-2007, 01:09 PM
and you rendered that with Grid-Acceleration?

yes, rendered using grid accelaration.

dagon1978
06-22-2007, 01:53 PM
Well I'm impressed, the scene you created worked great, but I don't understand what you did to make it work...

eheh there are 3/4 things to setup up in this kind of scenes:
- raytracing acceleration (grid works fine with huge amount of geometries)
- placeholders
- memory limits
- scanline/raytracing (when you have many geometries in scene you have to disable the scanline alghorithm for the first rays)


hey dagon,

nice tests! Could you give us some insight on how you made it work in Maya and probably the scene file so we all can have a look too?

That would be great! Thanks


Pascal

why not? ;)
here's the interceptov scene "reloaded" (http://rapidshare.com/files/38703872/teapot_fR.rar)


The thing is that in FR it's in most case a matter of click render and see results while in MR you have always to go tweaking something here or there. But This would push us to ask why a a renderer that has been implemented since Maya 4.5 still causes a headache for simple users like us especially for starters while the recent one is much more easier to use and gives better results in certain areas or in general faster ?! I know MR comes for free with Maya and this is its big asset for now.

mray it's evolving "slowly", especially the OEM interfece part... i use max and i know there are parts of mray4maya very annoying, and this is true for some part of mray4max too... and i know there are many other parts working much better in xsi... then, probably we have to ask to Autodesk a bit more attention of what they're doing with mray

Visor66
06-22-2007, 03:36 PM
@dagon: Thanks for the scene! Although it crashes on my machine when I hit render! Is there anything I still have to set up?

Also I would like to know more about the placeholders you were talking about! What is it and where can I learn something about it!? :thumbsup:

Visor66
06-22-2007, 03:41 PM
oh, now it renders fine! Awesome! Still I would like to know more about the placeholders...

Thanks

dagon1978
06-22-2007, 04:16 PM
..though it does not support motion blur.

that's true floze, but... comparing apples-to-apples, if you try to eneble motionblur in the fR scene you have the same problems...
you get:
translation, then RAM>RAM>RAM, then infinite rendertimes, then something like this:

"Warning: NMSP : not enough memory for compiling faces for Teapot01Shape
Warning: NMSP : not enough memory for compiling faces for Teapot01Shape
Warning: NMSP : switch to conserve memory mode for Teapot01Shape
Warning: NMSP : switch to conserve memory mode for Teapot01Shape
Warning: NMSP : not enough memory to build full MSP for instance Teapot01Shape
[cut]
Error: NMSP : not enough memory to build MSP for instance Teapot01Shape
Warning: NMSP : not enough memory to build full MSP for Teapot01Shape
Warning: NMSP : this may result in low performance
Warning: NMSPot01Shape
Error: NMSP : not enough memory to performance
Warning:Warning: MAIN : system idle
Time for frame render (hh:mm:ss.mil): 00:14'46"594
Total frame computation time (hh:mm:ss.mil): 00:17'28"688"

and a render with just 10/15 teapot instead of 5.000 :p
so, the fantastic fR instances mechanism dont work very well with motion blur... and this is true for mray too... in this type of renders renderman is the only choice

oh, now it renders fine! Awesome! Still I would like to know more about the placeholders...

Thanks

;) it crashes sometimes to me too, over frame 35

oh, remember the 3gb swith for win xp

p.s.
smoething about placeholders (http://www.lamrug.org/resources/placeholdertips.html)
this is a standalone page, in maya you have just a "threshold" (numbers of polys over mray use the placeholders mechanism) and nothing more ;)

Michael-McCarthy
06-22-2007, 09:09 PM
These are all very interesting tests. But as with all "benchmarks" I feel they can be "and are" skewed one way or another. This topic was set to discus some finalRender for Maya tests. While some comparisons can and should be made I think we reach a point of diminishing returns (as fare as useful information) when we really try to pit 2 different scenes or features against each other.

for example, this instance rendering. fR does it very well and its easy to set up. MR can also do some instance rendering but it takes a bit more tuning and has some drawbacks. Once you start tuning and setting different MSPS, grid methods and talking about "well these teapots render Ok but not a real scene", you are really starting to compare apples and oranges.

It looks like fR does good instance rendering that just works and is easy to setup (I know this to be true). It has been show that MR can also do instance rendering (I also know this to be true) but may take some more tweeking or tricks to get it going.

Also this looks like some tests are happening were an MR expert (or at least knolageable person) is setting up MR scenes and tweaking them and then (with little or no fR experiance) rendeing in fR. This again happens all the time (even on the fR side too) and never produces good results.

An example of this is the instances with Motion blur setup. fR has no problem doing this but you are running out of RAM. There are many ways to for this to work but not being familiar with fR you may not use these methods. fR has twean, as well and camera and 3d mblur each that might get you the result you want as well as 64 bit rendering which can allow for huge ram renders. The dynamic MSP can also offer huge ram savings so you can render these types of scenes. None of these are really tricky to set up or hard to use but a complete none fR users will just not do them.

Anyhow. Id like to see many more fR tests and opinions, but lets try to keep it as positive as possible... App wars and negative spirals rarely yeild good or helpful information :) I hope I'm not to optimistic :)

Thanks a bunch
Michael McCarthy

coccosoids
06-22-2007, 09:13 PM
I'd like to see the classroom scene rendered at 2k resolution! ;)

yolao
06-22-2007, 10:56 PM
hey Michael

-do you have an aprox. realese date for FR sp3 for maya?

-Can you talk about what new features will have?...i know that you already said about more control to the render passes...which i`m sure will be nice:), but any other new features that you can talk about?

Thanks

Michael-McCarthy
06-23-2007, 01:31 AM
Well we are RC now so it should be soon. Our RCs don't tent to last more then a few weeks (but you never know what might pop up)


Cant talk to much about features but there will be some powerfull new shaders, Impressive new RE stuff, and actually some new tools that can make the stuff rendered here look silly :). These are some highlights but there is actually much more. Of course there are many fixes too(the real point of an SP). I think all our users will be super impressed for a free SP :D

Thanks
Michael McCarthy

yolao
06-23-2007, 02:34 AM
thanks Michael.

Bmoner
06-23-2007, 02:40 AM
Where can I get scenes like the class room scene, or that buddah statue or dragon statue?

coccosoids
06-23-2007, 07:51 AM
You can find a link to the classroom scene in the Vray like renders with MR thread... just
search for it.

Michael: Can you say anything about an osX version of finalRender?

coccosoids
06-23-2007, 12:05 PM
:)
Well, I have to say that I am a little impressed by the GI implementation of finalRender.
It makes a hell of a first impression. No blotches, no loose photons, quite fast...

Although I think it needs a megaShader, like MIA in maya... Area lights I didn't quite get.
I overdone the ambiantocclusion, sorry. :)
And I don't really get AA either, it seems a little slow on that side.
Glossy reflections, reflections+refractions I didn't even touch so I can't compare.
Anyone care(dare) to redo this scene in mental ray to compare? (dagon ? ;) )
Also, if anyone thinks there's a lot of improvement possible with this scene I could
upload it somewhere.

Render time: 1h (intelCore2duo 2.1)

http://img519.imageshack.us/img519/9973/zzznz0.th.jpg (http://img519.imageshack.us/my.php?image=zzznz0.jpg)

-forgive the jpeg quality.

dagon1978
06-23-2007, 03:15 PM
:)
Anyone care(dare) to redo this scene in mental ray to compare? (dagon ? ;) )


where i can download it? ;)

for the glossy, i know fR stage-1 for max has many powerful shaders, the architectural one is identical to the mia_material (and for identical i mean really identical)

coccosoids
06-23-2007, 03:38 PM
Check your PM please! :)

You got it dagon?

Here's my best with mental ray... not the same scene but the blotchiness steals the
show either way...

http://img252.imageshack.us/img252/1755/testphotonroomenclosedghi2.th.png (http://img252.imageshack.us/my.php?image=testphotonroomenclosedghi2.png)

You've got mail!! :)

techmage
06-23-2007, 11:40 PM
yeah, absolutely instances... as interceptov say ;)
mray use maya instances, so i dont know what kind of problem you have...

I see, for some reason I was under the impression that the mayatomr instancing wasn't up to par. But looking at the exported .mi files they technically are instanced. I think the thing that was confusing me is memory usage seemed to increase proportional to the vertice count, even if everything was just instanced.

But now I'm thinking the increased memory usage probably isn't due to storing more vertices from instances. Do you know what mental ray is doing internally to increase memory usage proportionally with instance count? Is it the BSP tree that has to take up more memory cause of more instances?


I've been fiddling around with this a bit. I can only get high levels of instanced faces (I got 10 million faces in the current scene) to render with grid acceleration. Do you know if largeBSP is at all capable of handling this much geometry in anyway? Using largeBSP would crash my scene.

coccosoids
06-24-2007, 07:33 AM
dagon : I only get 3 fR shaders in maya (skin, glass, metal)... Am I missing something here?

floze
06-24-2007, 02:15 PM
I totally agree with Michael, this shouldnt turn into an app-war. And I do agree that we indeed are comparing apples to oranges, specially because noone seems to post their system specs. So just to keep it continuous I wont do that either.

Anyway, for the fun of it I picked up where dagon left, continuing a rather meaningless test and polygon war - so please dont take it too serious. And for the even extra more fun of it I added glossy reflections, ambient occlusion and soft shadows... and a little custom built global tonemapping operator that's not worth mentioning.


13.166 instances, 408.320 faces each, makes total a of 5.375.941.120 tris. Of course there's a little kung-fu action going on, but hey.. take it with a little grain of salt.

http://individual.floze.de/fileadmin/files/cgtalk/mr_instances2.jpg
-
http://individual.floze.de/fileadmin/files/cgtalk/mr_instances1.jpg


I'd really love to see some more final render tests coming along, the sky portal seems to be very interesting for example.

jude3d
06-24-2007, 03:15 PM
your teapot test is all about instance, try to render millions of polycount with animation motion and textures without any instancing you should crash the render quickly.
instance cost no memory so it's easy for any render to keep billions of faces alive, this is not really usefull in animation. renderman is the only renderer could handle millions of polycount without crashing.
so cool test but not really impressive...:shrug:

dagon1978
06-24-2007, 03:18 PM
dagon : I only get 3 fR shaders in maya (skin, glass, metal)... Am I missing something here?

that's because fR stage-1 for max is much more complete then fR stage-2 for maya... and i can't understand why... stage-2 is not newer then stage-1? doh



I see, for some reason I was under the impression that the mayatomr instancing wasn't up to par. But looking at the exported .mi files they technically are instanced. I think the thing that was confusing me is memory usage seemed to increase proportional to the vertice count, even if everything was just instanced.

But now I'm thinking the increased memory usage probably isn't due to storing more vertices from instances. Do you know what mental ray is doing internally to increase memory usage proportionally with instance count? Is it the BSP tree that has to take up more memory cause of more instances?


I've been fiddling around with this a bit. I can only get high levels of instanced faces (I got 10 million faces in the current scene) to render with grid acceleration. Do you know if largeBSP is at all capable of handling this much geometry in anyway? Using largeBSP would crash my scene.

mmm i dont really know the technical part of the work, maybe someone like paolo berto or gga or someone of the power users can give us a clue

for the largeBSP i think you can get more then 10M instances... but i dont know how many more... in a 32bit system... ;)



I totally agree with Michael, this shouldnt turn into an app-war. And I do agree that we indeed are comparing apples to oranges, specially because noone seems to post their system specs. So just to keep it continuous I wont do that either.

Anyway, for the fun of it I picked up where dagon left, continuing a rather meaningless test and polygon war - so please dont take it too serious. And for the even extra more fun of it I added glossy reflections, ambient occlusion and soft shadows... and a little custom built global tonemapping operator that's not worth mentioning.


13.166 instances, 408.320 faces each, makes total a of 5.375.941.120 tris. Of course there's a little kung-fu action going on, but hey.. take it with a little grain of salt.



I'd really love to see some more final render tests coming along, the sky portal seems to be very interesting for example.

my god! how many teapots! :D
great works floze... your system specs?

dagon1978
06-24-2007, 03:21 PM
your teapot test is all about instance, try to render millions of polycount with animation motion and textures without any instancing you should crash the render quickly.
instance cost no memory so it's easy for any render to keep billions of faces alive, this is not really usefull in animation. renderman is the only renderer could handle millions of polycount without crashing.
so cool test but not really impressive...:shrug:

i would like to see your test with renderman and 5bilions of teapots with glossy and AO :)

oh, and raytracing obviously... :rolleyes:

floze
06-24-2007, 03:31 PM
my god! how many teapots! :D
great works floze... your system specs?
Uhmmm... Intel Xeon dual quad-core @1.86GHz and 4GB of RAM. :cool:

jude3d: With mental ray (x64) I rendered CAD car models that had several millions of polygons, at the least usually around 6.000.000, without major problems.

coccosoids
06-24-2007, 03:37 PM
your teapot test is all about instance, try to render millions of polycount with animation motion and textures without any instancing you should crash the render quickly. instance cost no memory so it's easy for any render to keep billions of faces alive, this is not really usefull in animation. renderman is the only renderer could handle millions of polycount without crashing. so cool test but not really impressive...

As jude3d points out these tests are somewhat irrelevant bearing no meaning of real
life situations... Unless dreamworks is working on the next big hit: "10 buzzillion teapots"

I want to see something like ArchViz or (as they advertise) "flicker free GI"...
That's why I put together that little test...
And, since we're dealing with the subject, adding glossy and refractions seems to alter
greatly rendering times - and for the worse, ;) ... and it all seems to be related to the
GI (FG? ;) ) rays shot during final rendering. Any tips there my comrades?
I will probably have to render a little project (interior) sometime next and I would like
to employ the power of fR.

dagon: any chance of seeing something from you that can 'compete' with fR ? :)
I know you have a few tricks up your sleeve!

EDIT: woops.. Hey, just read your PMs. I'm waiting anxiously :)

floze
06-24-2007, 03:53 PM
As jude3d points out these tests are somewhat irrelevant bearing no meaning of real
life situations... Unless dreamworks is working on the next big hit: "10 buzzillion teapots"

I want to see something like ArchViz or (as they advertise) "flicker free GI"...
That's why I put together that little test...
And, since we're dealing with the subject, adding glossy and refractions seems to alter
greatly rendering times
Thanks for pointing this out q_vazk, this is what I actually wanted to express with the little 'test'. Unless there is no truly standardized benchmark scene we're always comparing apples to oranges. So it's all a pretty much useless though sometimes even funny pissing contest (sorry).

Maybe let's even keep this contests completely out of this still promising thread? And see the strongs and weaknesses of stage-2 come along by serious final render exercises, instead of bashing each other? That's my 2 cents. ;)

Hasnz
06-24-2007, 03:58 PM
Nice Tests !!! but they wont impress me as im stuck with a more practical problem :

I have a class room scene with around 30 characters . averaging around 10,000 polys each with mr Subd Approx applied to literally everything , even the duster and chalks have Subd applied . All textures and tga from 512 to 2048 res . ( not converted to map format yet )
I have physical sun sun applied with two area lights and one extra directional light .

This scene simply wont render as a whole . It doesnt crash Maya , but is giving fatal errors after which i need to get Maya restarted . Im afraid ill have to render different groups of Characters and comp then in the end .
What what setting will atleast let me see the test render even at half res .
I have a core 2 duo with 2 GB RAM , which i think is more then enough to handle this scale of scenes .

Saturn
06-24-2007, 04:03 PM
your teapot test is all about instance, try to render millions of polycount with animation motion and textures without any instancing you should crash the render quickly.
instance cost no memory so it's easy for any render to keep billions of faces alive, this is not really usefull in animation. renderman is the only renderer could handle millions of polycount without crashing.
so cool test but not really impressive...:shrug:

http://www.glassworks.co.uk/search_archive/jobs/vauxhall_astrabatics/index.shtml

It was done last year and we got only 3 semaines to finish this commercial .

Here we got :
7 millions poly per cars ( no instances )
Quite heavy use of raytrace.
HD resolution.
Motion-Blur
Flicker free GI
Caustics.

In others word today in 2007 Renderman don't get anymore the monopole of rendering a lot of geometry. I will push this a bit more in 2007 any rendered can perform this.
So please stop living in the past.

At the end of day we are all making images, no matter which tools we are using. It's a matter of personal preferences.

I have jumped on this thread only because untrue statement was done.

dagon1978
06-24-2007, 04:21 PM
Nice Tests !!! but they wont impress me as im stuck with a more practical problem :

I have a class room scene with around 30 characters . averaging around 10,000 polys each with mr Subd Approx applied to literally everything , even the duster and chalks have Subd applied . All textures and tga from 512 to 2048 res . ( not converted to map format yet )
I have physical sun sun applied with two area lights and one extra directional light .

This scene simply wont render as a whole . It doesnt crash Maya , but is giving fatal errors after which i need to get Maya restarted . Im afraid ill have to render different groups of Characters and comp then in the end .
What what setting will atleast let me see the test render even at half res .
I have a core 2 duo with 2 GB RAM , which i think is more then enough to handle this scale of scenes .

mmm maybe we are a little bit OT here... but, check this (http://forums.cgsociety.org/showpost.php?p=4467203&postcount=67)

lazzhar
06-24-2007, 05:48 PM
The thing I like the must is that it doesnt need experience to get decent GI.
http://img260.imageshack.us/img260/9160/frteapot01sz1.jpg

On a dual core 3.2 GHZ

I just cranked up GI setting and went hanging out. I'm rendering now with better anti aliasing.

coccosoids
06-24-2007, 06:01 PM
lazzhar: are you using a simple maya area light? :)
And, may I say? The render times are quite high!!... are you rendering at 2k resolution?
Check back a few posts, to see my test, took only 1h. (? hope I remember correctly)

I would like to propose that if we are to compare -still- "apples to oranges" we should do it
at something like 2k res. Why? Because every renderer can get away with decent GI (and
decent start-to-finish rendering lap times :) ) at lower resolutions. Compare my two test
renders: MR -> blotchy, fR -> smooth as a baby's skin :)

dagon1978
06-24-2007, 06:04 PM
with an athlon xp 3200+@2400

http://img370.imageshack.us/img370/5125/teapotgi2m33sov3.jpg

;)

coccosoids
06-24-2007, 06:23 PM
:) Dagon: are you using AO on that one?
Grr... it has a little eerie and unnatural feeling to it (shadow in the foreground too strong
maybe?) plus it is tainted by the bloootchynesssss
:)
Did you try a 2k render? One overnight maybe?

I would suggest (to whom is more liberal in spirit) to share some, or maybe all of the
scenes that are circulating around here - thus it would be much easier for everyone to
optimize and modify settings and actually render, rather than wasting precious time
recreating specific scenes or situations. Just a thought! :)

Michael-McCarthy
06-24-2007, 06:40 PM
that's because fR stage-1 for max is much more complete then fR stage-2 for maya... and i can't understand why... stage-2 is not newer then stage-1? doh


This is incorrect and missleading. fR Stage-2 for Maya (and C4D) are fully complete and hugely well integrated into the application themselfs. In most cases we are much better integrated then even the default renderers (This is certainly true for Maya).

Stage-1 and Stage-2 are different in many ways (not more complete or less). First Stage-2 is a standalone renderer and Stage-1 is a plugin renderer. I will not go into all the differences there but the basic idea is a Stand alone has some more stability, speed and flexibility. While a Plugin rendererr is more easily able to support other plugins and extension features etc.

fR Stage-2 for Maya has most of the same shaders that stage-1 for 3dsmax does and some that it does not. The case here is that in Maya we support 99% of the shader tree, which is very powerful and with nodes like sampler info etc you can create a huge amount of shaders (carpaint in a snap). 3dsmax suffers from the lack of a powerfull shader tree so we have things like finalShaders etc to augment that lack. The other thing to take into consideration is release cycle. Stage-2 for Maya was released a while ago and the new Stage-1 for 3dsmax just released 2 months ago. So if you asked 2 months ago you would have gotten people on the 3dsmax side saying "why don't I have Dirt, and occlusion and the Maya guys do". That being said you will see some cool things in SP3 for Maya as well as C4D :D

Thanks
Michael McCarthy

coccosoids
06-24-2007, 06:53 PM
I would like it very much if you would release an official statement regarding the
possibility of an osX version of finalRender. This is the last time I'm asking this as it is
third strike in a row.

Michael-McCarthy
06-24-2007, 08:00 PM
Sure... must have mist it.

OSX is in development for Stage-2 Maya and C4D. We do not have a release date for the products. The port to OSX has taken longer then we would have liked for a number of reasons but we are still committed to releasing a great product.

Thanks
Michael McCarthy

lazzhar
06-24-2007, 08:05 PM
lazzhar: are you using a simple maya area light? :)
And, may I say? The render times are quite high!!... are you rendering at 2k resolution?
Check back a few posts, to see my test, took only 1h. (? hope I remember correctly)
:)

It's Maya Area Lights yeah and the rendering time is high because i have no experience in FR but you don't need to get mental like with the math of mental ray. It's like one button sollution. And it's grain free as well guys !! ;)

lazzhar
06-24-2007, 08:15 PM
Forgot to add the last rendering with better antialiasing where some details are more sharp. Sorry I forgot tosave rendering time because it was almost the same.
I'm not benchmarking here to be precise. The idea is that when it comes to GI I think that Final Render is much more user friendly and this what most people are asking for.
http://img526.imageshack.us/img526/6903/frteapot02dr1.jpg

coccosoids
06-24-2007, 08:49 PM
Michael McCarthy: Thank you. I'm glad to hear that a commitment has been made to osX,
in the meantime let's hope it won't be that long untill the next version of fR for windows.
:) I'm really looking forward to these new, surprise features you hinted at.

Can somebody share the scene file I'm seeing lately? I'd like to have a go at it with both
MR and fR. Thanks :)

dagon1978
06-24-2007, 09:52 PM
This is incorrect and missleading. fR Stage-2 for Maya (and C4D) are fully complete and hugely well integrated into the application themselfs. In most cases we are much better integrated then even the default renderers (This is certainly true for Maya).


Stage-1 and Stage-2 are different in many ways (not more complete or less). First Stage-2 is a standalone renderer and Stage-1 is a plugin renderer. I will not go into all the differences there but the basic idea is a Stand alone has some more stability, speed and flexibility. While a Plugin rendererr is more easily able to support other plugins and extension features etc.

fR Stage-2 for Maya has most of the same shaders that stage-1 for 3dsmax does and some that it does not. The case here is that in Maya we support 99% of the shader tree, which is very powerful and with nodes like sampler info etc you can create a huge amount of shaders (carpaint in a snap). 3dsmax suffers from the lack of a powerfull shader tree so we have things like finalShaders etc to augment that lack. The other thing to take into consideration is release cycle. Stage-2 for Maya was released a while ago and the new Stage-1 for 3dsmax just released 2 months ago. So if you asked 2 months ago you would have gotten people on the 3dsmax side saying "why don't I have Dirt, and occlusion and the Maya guys do". That being said you will see some cool things in SP3 for Maya as well as C4D :D

Thanks
Michael McCarthy

i'm talking about stage-1 max NOW and stage-2 maya NOW! not 2 months ago or 3 years ago
NOW in max (stage-1) you have MANY more shader than maya, and this is correct, or correct me if you can (NOW)

dagon1978
06-24-2007, 10:03 PM
:) Dagon: are you using AO on that one?
Grr... it has a little eerie and unnatural feeling to it (shadow in the foreground too strong
maybe?) plus it is tainted by the bloootchynesssss
:)
Did you try a 2k render? One overnight maybe?

I would suggest (to whom is more liberal in spirit) to share some, or maybe all of the
scenes that are circulating around here - thus it would be much easier for everyone to
optimize and modify settings and actually render, rather than wasting precious time
recreating specific scenes or situations. Just a thought! :)

no AO, this is just a 2minutes rendering with an old pc, with 2 hours in a core2 (o quad) pc you can get something better and easier with maxwell (or fry render or fprime), but this is not what i want...

2k render (http://forums.cgsociety.org/showthread.php?p=4471102#post4471102)

jude3d
06-25-2007, 12:24 AM
to saturne: I know final render could handle millions of polygons, , I like stuff you rendered with but I spoke about really hard to render in cgi with character animation a lot of displacement or things like that, those things crank up the rendering and need really good engine.

I don't think lot of studio use final render, Vray for plenty of reason , look at the last Ilm working on poc 3 the amount of polygons with GI was done in only 4 hours a frame. rendering a car is not the same as rendering a complex character with SSS, displacement, GI flicker free etc etc... for exemple rendering an occlusion pass with complex geometry and displacement need 1 to 3 hour in final render and could be rendered in 5 minutes in renderman 13. of course rendering plane surface of smooth texturing surface is fast for all renderer this is really different. with final render Vray etc... you could crank up the render time from minutes to hours in few cliks to improve the quality. this is the raytraced way. so it's not the past man , try renderman 13.5 and you would see the future of an 23 years old engine. But all those engine are really great for most of the project , short, commercial, archi and still image...but not the best for what I need.:) this is just my own point of view and experience in 12 years of cgi.

final render VRay and mental ray are really cool and fast for architectural, cars, commercials, but it's not the best for character animation. most of our work is about character animation. sometimes it works good sometime not it depend to the project man. they are all great render so you have to choose the right could handle your project. that's all

we add a lot of problem with high polygon count with vray mr or fr, some time the render crashed down and it's not really stable, this is one of the main reason why we changed our entire pipeline from mental ray to renderman few month ago. it was a choice to solve many of our big account problems and full of 8 to 16 k texture.

Michael-McCarthy
06-25-2007, 12:41 AM
i'm talking about stage-1 max NOW and stage-2 maya NOW! not 2 months ago or 3 years ago
NOW in max (stage-1) you have MANY more shader than maya, and this is correct, or correct me if you can (NOW)

Again this is not really correct for the reasons I already explained. Right now they bassicaly have the same shaders. As I said in Maya there is not a need for Carpaint, or many others because of our full support of the shader tree. In 3dsmax these shaders are needed but in Maya they would be nothing more then presets. Right now 3dsmax does have the Archi shader which Is cool but Maya has other shaders that 3dsmax does not. So there are small differences for sure but Maya or C4d are surely not missing a lot of shaders or "not complete" in any way.

Thanks
Michael McCarthy

dagon1978
06-25-2007, 03:35 AM
Again this is not really correct for the reasons I already explained. Right now they bassicaly have the same shaders. As I said in Maya there is not a need for Carpaint, or many others because of our full support of the shader tree. In 3dsmax these shaders are needed but in Maya they would be nothing more then presets. Right now 3dsmax does have the Archi shader which Is cool but Maya has other shaders that 3dsmax does not. So there are small differences for sure but Maya or C4d are surely not missing a lot of shaders or "not complete" in any way.

Thanks
Michael McCarthy

presets? really?

this is what "your" help file says:

"The fR-Advanced material on the other hand, is an incredibly powerful and complex material. If you need to build a really sophisticated material, fR-Advanced is the one to go with. However, be warned that this material type can take some time to learn. By taking the time to master this material, you can produce exceptionally realistic and robust surfaces for your scenes."

but, now i know this is just a preset :rolleyes:

p.s.

and stage-1 supports also all the max internal material/textures:

"finalRender Stage-1 comes with a whole load of materials, settings, textures and shaders. Besides the highly optimized native materials and texture maps, finalRender also supports nearly every standard 3ds Max material or texture map"

but hey, if you think the maya shaders are good enough dont worry, is not my problem ;)

Saturn
06-25-2007, 08:37 AM
to saturne: I know final render could handle millions of polygons, , I like stuff you rendered with but I spoke about really hard to render in cgi with character animation a lot of displacement or things like that, those things crank up the rendering and need really good engine.

I don't think lot of studio use final render, Vray for plenty of reason , look at the last Ilm working on poc 3 the amount of polygons with GI was done in only 4 hours a frame. rendering a car is not the same as rendering a complex character with SSS, displacement, GI flicker free etc etc... for exemple rendering an occlusion pass with complex geometry and displacement need 1 to 3 hour in final render and could be rendered in 5 minutes in renderman 13. of course rendering plane surface of smooth texturing surface is fast for all renderer this is really different. with final render Vray etc... you could crank up the render time from minutes to hours in few cliks to improve the quality. this is the raytraced way. so it's not the past man , try renderman 13.5 and you would see the future of an 23 years old engine. But all those engine are really great for most of the project , short, commercial, archi and still image...but not the best for what I need.:) this is just my own point of view and experience in 12 years of cgi.

final render VRay and mental ray are really cool and fast for architectural, cars, commercials, but it's not the best for character animation. most of our work is about character animation. sometimes it works good sometime not it depend to the project man. they are all great render so you have to choose the right could handle your project. that's all

we add a lot of problem with high polygon count with vray mr or fr, some time the render crashed down and it's not really stable, this is one of the main reason why we changed our entire pipeline from mental ray to renderman few month ago. it was a choice to solve many of our big account problems and full of 8 to 16 k texture.

Actually rendering car envolve a lot of raytracing. Glossy reflection and caustics because if you need a good looking shader you need to add theses effects.

But don't worry I got also plenty exemple of complex render as you say. Character with SSS displacement, FG, high res texture and even with Fur.

http://www.glassworks.co.uk/search_archive/jobs/ford_flexifuel_next_generation/index.shtml

and here :

http://perso.orange.fr/harry.bardak/video/PSP_comp__h264.mov

or even better the super punch scene in Matrix 3.


I am working also with Rman. So I know why I will use it and why I will not use it.
And before we go on point based ambient occlusion, you should try FG based occlusion which give you similar performance on same complex object.


As I said before If rman because it answers to your need. Then fine excellent for you.
I am just here to correct UNTRUE statements. Please don't assume you are higher because you use Rman on complex scenes in 2K. The industry got a lot of kind of guys we don't need go that low. At the end we produce all images and honestly I don't care what renderer you use.


The problem is that you probably use MR only through Maya. And honestly I understand all your statement. I am frustrated too when I using it under Maya.

Saturn
06-25-2007, 08:41 AM
Again this is not really correct for the reasons I already explained. Right now they bassicaly have the same shaders. As I said in Maya there is not a need for Carpaint, or many others because of our full support of the shader tree. In 3dsmax these shaders are needed but in Maya they would be nothing more then presets. Right now 3dsmax does have the Archi shader which Is cool but Maya has other shaders that 3dsmax does not. So there are small differences for sure but Maya or C4d are surely not missing a lot of shaders or "not complete" in any way.

Thanks
Michael McCarthy

Does it mean that convert also MR shaders ( like Car paint ) to FR ? If it s true that really an amazing features or you got equivalent shader that just take it place when you use car paint node ?

lazzhar
06-25-2007, 09:10 AM
...

http://www.glassworks.co.uk/search_archive/jobs/ford_flexifuel_next_generation/index.shtml

...

Holly sh!t !Any hope to see a making of or a behind the scenes somehow? Amazing stuff! super bien réalisé ! :thumbsup:

Michael-McCarthy
06-25-2007, 09:37 AM
presets? really?

this is what "your" help file says:

"The fR-Advanced material on the other hand, is an incredibly powerful and complex material. If you need to build a really sophisticated material, fR-Advanced is the one to go with. However, be warned that this material type can take some time to learn. By taking the time to master this material, you can produce exceptionally realistic and robust surfaces for your scenes."

but, now i know this is just a preset :rolleyes:

p.s.

and stage-1 supports also all the max internal material/textures:

"finalRender Stage-1 comes with a whole load of materials, settings, textures and shaders. Besides the highly optimized native materials and texture maps, finalRender also supports nearly every standard 3ds Max material or texture map"

but hey, if you think the maya shaders are good enough dont worry, is not my problem ;)

Yes but again 3dsmax does not have a shader tree. So yes many of the things you can do with fR advanced would just be presets with Maya's powerful shader tree. These are not possible in 3dsmax without a specific shader. That being said; it is nice to have caned shaders like Carpaint even if you can easily do them with a moderate amount of shader tree experience, so we are looking into making finalShaders for Maya or C4d.

I am happy to clarify but I believe all this was clear in the first 1 or 2 posts I made here regarding this subject. dagon, your posts are argumenitive and misleading in places. At the onset I was surppised with this adversarial tone you took with Toni as he is an Amazing artist that creates incredible work. Now I am seeing this in response to my posts even though I think I have posted clear and understandable explanations.

I am more then happy to answer anyone's questions here but I don't want to be goaded by any flamey posts. So don't be surprised if posts like this go unaswerd. Also I would take a look at this "Style" of presenting yourself, I feel it is detrimental to credibility.

Thanks
Michael McCarthy

yann22
06-25-2007, 09:54 AM
Hi Michael,

I would be really interested in an answer to saturn's question regarding the "full support of the Maya shader tree", e.g. carpaint, as well. Does FR convert/translate all MR shaders?

cheers,
Yann

Michael-McCarthy
06-25-2007, 10:57 AM
No fR definitely does not convert or translate MR shaders. Lots of Carpaint shaders should be easy enough to make with Mayas basic shaders with fR nodes. Again we are thinking of having more canned "finalShaders" for Maya as some people seem to like this idea that heve less experience with the shader tree.

Always good to have feedback on what users want :D

Thanks
Michael McCarthy

coccosoids
06-25-2007, 11:28 AM
Well, I would say that "canned" shaders - as you call them - are essential for any rendering
engine for maya. I am surprised you haven't seen/been aware about the welcome
shaders like MIA, TS2, CTRL_shading had for mental ray. Everyone is true to the concept
"all in one" these days.

Michael-McCarthy
06-25-2007, 11:34 AM
Ahh we are aware :) You will see some cool stuff in SP3 and future releases :D. Our goal it to make rendering advanced effects as easy as possible and this can help there.

Thanks
Michael McCarthy

yann22
06-25-2007, 11:55 AM
No fR definitely does not convert or translate MR shaders.

That's what I thought :D, it just sounded too weird, but thanks for the confirmation.

And I agree with q_vazk in that, while connecting large networks of nodes in the Hypershade is providing me with endless hours of joy, the trend is definitely going towards 'canned shaders' like mia and car-paint :).

Saturn
06-25-2007, 12:23 PM
Ahh we are aware :) You will see some cool stuff in SP3 and future releases :D. Our goal it to make rendering advanced effects as easy as possible and this can help there.

Thanks
Michael McCarthy

That the right way to go !

dagon1978
06-25-2007, 02:36 PM
Yes but again 3dsmax does not have a shader tree. So yes many of the things you can do with fR advanced would just be presets with Maya's powerful shader tree. These are not possible in 3dsmax without a specific shader. That being said; it is nice to have caned shaders like Carpaint even if you can easily do them with a moderate amount of shader tree experience, so we are looking into making finalShaders for Maya or C4d.

I am happy to clarify but I believe all this was clear in the first 1 or 2 posts I made here regarding this subject. dagon, your posts are argumenitive and misleading in places. At the onset I was surppised with this adversarial tone you took with Toni as he is an Amazing artist that creates incredible work. Now I am seeing this in response to my posts even though I think I have posted clear and understandable explanations.

I am more then happy to answer anyone's questions here but I don't want to be goaded by any flamey posts. So don't be surprised if posts like this go unaswerd. Also I would take a look at this "Style" of presenting yourself, I feel it is detrimental to credibility.

Thanks
Michael McCarthy

feel free to unanswer my posts, my "style" is not a problem for me
it's funny you see my words "detrimental to credibility" but you considers Toni an "Amazin artist" even with really misleading words like this: "Try to put 14 milion instanced polys in 2gb machine and render. It will crash for sure ..."

maybe it's because i'm talking against fR and he talks against mray?

eheh

bye Michael

techmage
06-26-2007, 07:18 AM
I was preparing an archviz comparison. But I ran into a little issue on mental ray's side using the grid acceleration. The grid seems to cutting my scene in half. I'm just using the default settings for the grid acceleration. I tried fiddling around with them a little bit, which did cause it to cut the scene differently, but it never got rid of the cut completely. Anyone know how to fix this?

Libor
06-26-2007, 01:07 PM
Well its very interesting thread and it seems there are many specialists on rendering using MR and FR! One thing which Im really curious about is framebuffer size (read output resolution)

We needed in our company to render 2 mil polygon mesh (exported from realflow) in huge resolution for big print ad in 17k x 10k pixels but there werent any chances even on a 3GB of ram (Win32bit). And the scene was really simple in terms of rendering (1 light, 1 reflection and some ramp controled by sampler info for fresnel like reflection falloff)

So does anybody know if it is a MR problem or in FR too??

Maybe some benchmarks on that?

Keep posting people its really interesting to read all of the experiences and opinions!

Libor

Michael-McCarthy
06-26-2007, 01:31 PM
Hello Libor,

This is no problem in fR. You can use the Dynamic Bitmap Pager for both Textures as well as the virtual frame buffer. It works well and is easy to setup.

Actually Digital Tutors has one of their free finalRender videos on High res textures in fR here:

http://www.digitaltutors.com/digital_tutors/tutorials.php?cat=finalrender

Thanks
Michael McCarthy

marckury
06-26-2007, 02:26 PM
i dont understand this... i have interior scene with area light in normal mode and all pretty nice , but if i change mode to sky or light portal all turn to black, I forgot do some connection? and another question , u can say to a mirror emit light to make shadows? I mean if light bounce on a mirror , mirror acts as a light generating more shadows, it can be done in final render or any engine?

yolao
06-26-2007, 02:58 PM
Hey Michael

i like mix20layer or mix8layer with MR, is there something similar in FR for maya?

Thanks

Michael-McCarthy
06-26-2007, 06:26 PM
Hello marckury,

The portal modes of the lights don't provide any light themselves - they act as 'holes' through which other lights can pass. So the behaviour you describe is as expected if you have no other light source in your scene. (I think you already got this answer today on our forum)

As for bounced light and shadows. This would be a caustic effect and fR has some great casutics. We support volume caustics also.

yolao,

We support Mayas LayerShader and Layer Texture nodes which do basically the same thing (I beliieve). I actually think the Maya nodes are more advanced and offer more options.

Thanks
Michael McCarthy

BillSpradlin
06-27-2007, 07:01 AM
Well its very interesting thread and it seems there are many specialists on rendering using MR and FR! One thing which Im really curious about is framebuffer size (read output resolution)

We needed in our company to render 2 mil polygon mesh (exported from realflow) in huge resolution for big print ad in 17k x 10k pixels but there werent any chances even on a 3GB of ram (Win32bit). And the scene was really simple in terms of rendering (1 light, 1 reflection and some ramp controled by sampler info for fresnel like reflection falloff)

So does anybody know if it is a MR problem or in FR too??

Maybe some benchmarks on that?

Keep posting people its really interesting to read all of the experiences and opinions!

Libor

The problem when rendering resolutions of that size isn't typically the fault of the renderer, but more so when the file gets written to disk. A file of that size is going to be gigabites in size and loading that (in Windows) for output usually causes a segment fault. You can get around this by rendering regions with mental ray and batching that out with a simple mel script and then using a compositing package that has scripting (Nuke or Shake) to bring in the regioned images and rebuilt the image.

BillSpradlin
06-27-2007, 07:04 AM
Hello marckury,
yolao,

We support Mayas LayerShader and Layer Texture nodes which do basically the same thing (I beliieve). I actually think the Maya nodes are more advanced and offer more options.



The reason the mix* shaders were initialy created was because mental ray for maya didn't fully support the maya layered shader/texture nodes (some final gathering/gi support issues) so if fR supports Maya's base shader package then you should be fine.

Libor
06-27-2007, 07:57 AM
Michael> that sounds great! I saw the video on DigitalTutors and definetely looks promising...
So if I get it right, there is actually no limit of output res (theoretically) ??

Bill> well thats exactly the way we tried to solve it but for some reason the sampler info which was sending facing ratio of camera to ramp texture changed even when we just offset the camera by script (without persp changes). Is there any chance to get my hands on some MEL script which solves this (web)?

In the end our tiles didnt match because of diferent reflection amount on each :scream:

Regarding the rendering failure it seems that MR even couldnt allocate memory/disk space for creating the framebuffer! Is it the same what you said? I guess you thought writting the final img after render so to me its not WinXP problem but the Maya/MR.

Any thought on this?

And thanks for your replies guys!

alexandre jesus
06-30-2007, 03:10 PM
Heres some good example from the cebas image gallery:
http://www.finalrender.com/gallery/v/interior/AlexandreJesus_patio_060607.jpg.html


Hi I rendered this image, like Michael said wisely is the artists that make artwork, the renderer is just a tool. Never used others renderers so i can`t compare.
I have final render because i have 3ds max and to make exterior shots was ok but long rendering times with default 3ds max render but for interiors sucks or take too long if was used with radiosity(other 3ds max "advanced lighting" renderer)
I like final render because is straight forward and now with final render stage 1 r2.0 handles nicely my scenes
http://www.idea-3.com/fr/85-final.jpg


This scene have 18 million polys and rendered everything in one pass.


btw, notice that the maya users talk wonderfull of the mia and ts2 shader, what have so great about?

techmage
06-30-2007, 10:27 PM
Alright so, I will donate one of my house models to this comparison (this is imported from sketchup). I've set up this quick little scene in maya mental ray with all of the latest bells and whistles in mental ray.

It currently contains:
-about 34 million triangles, alot of instances, trees and grass is all 3D (frame 48 is what I rendered at for the grass particle emmiter)
-physical sun/sky, sun is area light with an angle of 1
-mia_material on everything
-The grass and leaves are translucent through 1 face.
-Only one reflection level on everything (need refraction of 4 to get through windows)
-Only one FG bouce, no GI. The FG is calculated with rather low settings, maybe a -4 -2 sampling rate in fR would be comparable.
-I put round corners on everything but with a radius of 1 (except grass dirt leaves)
-the mia_material 'details' ambient occlusion checked on everything with a radius of 600. (so use the 'details detection' in fR)
-sampling level of 0, 2. Except the grass is overrided to -1, 1 and the trees are overidden to -2,0.
-The camera position is keyed at frame 1.

Now the shaders:
miaDirt, miaFloor, miaGrass, miaStone, miaTrunk - mostly lambert but does have 'highlight only' checked, I don't think fR has a feature comparable to this so I guess just use lambert, the MR docs say 'highlight only' adds very little rendertime so it's prolly negligible.

miaFrontSupports - this is brushed metal, .8 reflection blur (0 is highest amount of blur, 1 is least in mia)

miaGlass - physical glass preset, no blur at all

miaWhiteWall, miaWindowFrames, miaLeaf - These have interpolated blurry reflections. I don't think fR has a featue comparable to this so just use normal ones. The reflection blur is .3

I rendered at 800x600 using large BSP, took 1 hour 56 minutes I have a dual opteron 246 with 2gb of ram. I still couldn't figure out how to get the hierchal grid acceleration to work on this scene.

Alright I think thats all we need to make a good comparison.

Now one thing to note about this scene is there are multiple materials on the same object. So you either need to hook up new shaders to the existing shadingGroup nodes (what I do). Or in the hypershade right click on the mia_materials and 'select objects with material', then right click on the new shaders and 'assign material to objects'.

Here is the scene http://rapidshare.com/files/40305356/fR_MR_Test.zip.html

MaxL
07-01-2007, 04:18 AM
I'm Max Liani the developer of finalRender for Maya translator.
I'm sorry I didn't see the discussion earlier, but now I'm here to clarify some points and even gather suggestions about how to improve finalRender.

First of all I want to say something about the number of shader provided by fR for maya.
As Michael sayd 99% of maya shader are supported by fR (and some of them also provides some cool improvements) there are a number of effects you can achieve by using those shader. I understand that for non expert people is not easy to put togheter a good shading network and obtain reliasitc results. For example you can esily simulate a dielettric material without a dielettric shader. But if you don't know what is the physic behind that effect you don't know how to achieve that result.

It's a common mistake to think finalRender stage-2 is just an upgrade of stage-1. It's brand new engine, 90% rewritten from scratch as a generation step up.
As each brand new product you need to proced by steps:
1) add base functionalities and support for the most important platform (maya in this case) features.
2) reach a good balance of stability and performance to make it usefull in production.
3) add secondary features like new powerfull shaders, take advantage of a proficuos comunity that write and publish new features (like shaders and scripts) on the internet.

Untill now we took care of important features, core components, shader SDK, fast an reliable translation and rendering. They are far beyond more important than some few uber shaders. More shader will come, trust me... a lot more... and much much better designed that the one coming with MR for maya ;)

An other important thing to say is that we are improving faster than what MR for maya is doing. Somebody just sayd that there are mix8 and mix20 shaders downloadable because MR never supported properly maya layered shader/texture... This should make you think about it: if you have a problem (bug) with MR, nobody is listening to solve it on the next release. In the other side we are listening to all bug enquiry, thrying to solve them as quickly as possible. (This is one of the most important feature of a software.. but hardly mentioned in any comparison!)


I mainly use prman for my daily work, but I know mentalray as well.
One of the reason I started writing the translator of fR was because I was sick of the bad integration of mentalray in maya, where you need many nodes to setup a bump... or you need to download a custom shader from the internet to simplify a such common operation.
I asked myself, let's see if we can do better than that....

... and for people killing eachother about 20 seconds less per frame, or 1 gazzillion polygons more... do yourself a favor: try to "benchmark" how much time it could take to learn how to use a new render engine, or how much time you spend strugling each day on a bad designed workflow... 20 seconds more in rendertime, but zero seconds to prepare all your commonly used render passess, priceless.

Max

slipknot66
07-01-2007, 06:04 AM
Well, i think the question here is not about learning a new render engine, but the quality of the render.
From everything i saw at cebas website, mental ray have a superior quality, compared to Fr and thats very impotant.
Among other things that i think mental ray is far superior, but oh well.. thats just me :)

coccosoids
07-01-2007, 07:57 AM
Yeah - based on their gallery one might say that mental ray IS superior... But it's not fair
competition. I mean, where do you find a mental ray gallery? In the movies maybe...
Because if you compare Vray's gallery then you very well might say that Vray IS superior.
So that is what this thread is all about - a direct comparison of the two: speed, QUALITY,
versatility etc...
I'm waiting for a FR militant to post his version of the house and then thing will really start
to get interesting... ;)

MaxL
07-01-2007, 09:06 AM
Well, i think the question here is not about learning a new render engine, but the quality of the render.

Ok, sorry if I was OT.
But quality is and hard thing to measure. I could split "quality" in 4 components:
1) Core features quality, like is MotionBlur good or not.. this doesn't depend on users skills, but just on the render engine algorithms, and if they are not good, it's hard for the user to find a workaround.

2) Shaders quality, this is what usually people refer when they talk about quality. The big advantage of MR in this matter is that you can find many free shaders on the internet, but implementing the same shader in different engines should bring exactly the same quality, only the speed would change.

3) Setup complexity, how easily you achieve a specific effect (not render time, but user time spent playing with the GUI and with the provided tools). The faster the better because you will have more time to spend in achieving a better quality.

4) User skills

This last one is by far the most important and can be further subdivided in two parts.
1) Artistic skills, the talent to create a good image, sapient usage of lights and colors to trigger emotions. (absolutely non sense to bench)

2) Technical skills, the ability to use a particular tool to achieve the desired look/effect. This involves lot of knowledge in many subjects, physics, optics, and specific understanding of the tool features and settings.


From everything i saw at cebas website, mental ray have a superior quality, compared to Fr and thats very impotant.
Among other things that i think mental ray is far superior, but oh well.. thats just me

What do you mean in this case? What is superior, because if we know where fR "fails" in terms of quality, we can think about it and come back with better features in the next releases.
I don't think people here is interested in reading "Among other things that i think mental ray is far superior, but oh well.. thats just me". Please talk about these superior features, say what is better and why. Otherwise you are not contributing at the comparison. And people won't have an idea where and why an engine is better than an other for a their specific task.

I know that image galleries are made in purpose to show the quality of a tool, but in my opinion it's a very superficial mistake to use image galleries as a score.

Just as stupid example. If you see in a picture poor textures, visible tiling, fake reflective surfaces... you can't say the problem is the render engine, because we are talking about commercial products, and a product with that problems won't be bought by anyone.
If you see that the image is graysh and have a classic poor CG look, it's hard to say if it's because of the render engine or the user who didn't know how to set the shaders...
But if the render engine is so hard to set that only experts can achieve the desired quality, then the lack of quality is due to the engine too complex to use.
To explain with a case mentioned in this thread. If a lot of people says it's impossible to render more than 14 millions of polygons in MR, this is not the "absolute true", but if you need to be a super expert to set the right combination of options in order to get better results, then the tool is just too hard to use and for normal people the limit is formally 14 millions.
You can show images with lot more complexity and show how cool YOU are to achieve that, rather than how cool the engine is.

More interesting is q_vazk try. Provide a common scene and see what people will come up with. Even better would be if people that have access to both fR and MR would publish what best they would do using both software, and the feeling they had using them. But in this last case people need to play fare.

Max

slipknot66
07-01-2007, 04:55 PM
Well, i meant the GI quality, light quality etc..
And when i said "well its just me", its because thats the way i see it.
Most of the images i saw at the cebas website, have that cg look, and thats a render engine problem, not an artist problem.All the images are very good in terms of art, (ideas..concepts ..etc).

marckury
07-01-2007, 06:25 PM
well in the official mental ray web i havent seen any special cool interior renders,actually its strange because they are quite bad anyways i use and have been using mental ray long time ago ,and i know mental ray is one of the best renderers but these days im trying final render and im getting the same or better quality easier and faster,im not talking about mental ray is worse because i have seen some renders of other people i dont like very much, im talking since my experience working with mental ray and now testing final render, so the best thing to do would be share the same scene and render it with diferent engines so we can see which engine is better in this case , anyways with that judgement way only seeing renders , I think vray for architecture is better than mental ray

slipknot66
07-01-2007, 06:51 PM
Well, you are right there, it would be a good idea if cebas could release a demo version of final render, so we could test the engine and compare.
The only thing i didnt liked in this thread, was when someone mentioned that was not possible to render millions of polys in mental ray for maya, well that was not correct as was proved.
And about the quality of the render, as i said, thats the way i see it, it does not mean others will see the same way.Again thats why i said "maybe its just me"... was a personal opinion.:)

dagon1978
07-01-2007, 07:13 PM
More shader will come, trust me... a lot more... and much much better designed that the one coming with MR for maya ;)


really?
we'll see something better then arch&design and not just a copy as in fR4max? :)

MaxL
07-01-2007, 11:24 PM
Most of the images i saw at the cebas website, have that cg look, and thats a render engine problem, not an artist problem.All the images are very good in terms of art, (ideas..concepts ..etc).

In almost 3 years of development of finalRender I saw lot of cool images done by our beta testers but the most of the time we coudn't publish those images because we didn't have the copyrights to do that. Many studio doen't allow software houses to publish their stuff.
In the art gallery you find what people decide to share.

I can't say much about the shaders that will come. But just think that no one of the shaders coming with mentalray are able to extract proper render passes (Render Elements in fR terms). Yes there are some you can find in the web, I saw custom solution, one big shader that does everything, and you need also its custom light in order to make it work, but that's not the way to provide a solution to a such common task, the solution it's way too close and limited to a single plug-in package. This is an example of what I sayd as design.
All shaders in finalRender for Maya, including maya standard shaders and custom shaders can export specular, diffuse, reflections and a lot more stuff (quite a long list actually).
So if you connect some materials to layered shader, you will get always proper render passes to comp together. You don't even need to modify your maya shaders from your ald scenes.
This is a good shaders design.

Name an other render engine that can do this in maya now. Try to use the ctrl_buffer to get the same easy to use solution, amd you will understand what I'm talking about.

Max

Michael-McCarthy
07-02-2007, 12:07 AM
really?
we'll see something better then arch&design and not just a copy as in fR4max? :)

fR for 3dsmax always had the capability to do what was introdused in MR Arch material and just as easily as MR Arch (with fR advanced or other shaders). Because so many MR guys liked the MR arch workflow and layout, we emulated its UI so MR guys would feel comfortable in fR. So it looks and feels mostly the same for MR guys but uses 100% fR power and functionaluty meaning it is usually (in our and our beta testers works) faster and better quality.

So this is really not "just a copy" but it is a nice UI for MR guys that want to use fR, fast and easy :) with minimal learning curve!

Thanks
Michael McCarthy

slipknot66
07-02-2007, 01:00 AM
Well, i think ctrl_buffer its easy to use, i see nothing wrong with that,and you also have a nice tonemmaper when using ctrl_buffer.And theres also the mia_wrapper that is available for free.Another way its using the maya layer system, and maya shaders as they are supported with mr base dll.And again, as i said i think the GI quality in final render its not as good as in mr, and thats my opinion and my experience, as ive seen ppl using final render in Max.So to me its not a problem if i need to connect some extra nodes to have a bump or a pass that i need, actually i prefer that way, as i can have much more control.And when i said that mental ray was far superior, i was not talking about mental ray for maya only, i was talking about the fact that you can write your own shaders with it, you can create your own interface for the shaders if you want , using phenomena, again among other things. To be really honest the only thing i think mental ray in genaral needs to gets better, its with displace maps.But its always good to have choices, so theres where final render, maxwell v-ray comes in. Nothing wrong with that.

MaxL
07-02-2007, 01:06 AM
really?
we'll see something better then arch&design and not just a copy as in fR4max? :)


It's always the same old story. You do something new, and people says: "I don't like it, it's not like the one I was using before...". You do something similar to make people comfortable and people says: "Hey there is nothing new in this stuff, there is not worth for it.". http://www.cebas.de/forums/smf/Smileys/default/smiley.gif Ah the irony...

But waiting for people to post their versions of the shared scenes, let's keep up the discussion.

What is a good shader for you?
You seems to be quite a technical person using MR. Do you like uber shaders with lot of parameters, like the architectural shader? Or do you like smaller more generic shaders that once connected together you can really really do whatever you want?
If you have and you are able to use these last ones, you don't need to wait for kind people to share their specialized custom shaders on the net, like mia_*, ctrl_*. You could already do 99.9% of that stuff with the things you already got.

Uber shaders are totally worthless in the film industry (VFX, animated features), but they are appreciated by architects who are not experts to create specialized shading networks.
Each of us is different, but I wouldn't expect from a rendering technically skilled person to mention the arch&design shader as a cool thing, neither the default MR custom shaders, because they are even far too low level to be practical, even for skilled people that knows exactly how to use them. Also because they don't provide a simple unified way to combine them together, they don't provide a way to easily manage render passes (absolute common practice in the industry since several years).

I would like to heard you sincere opinion.
Once again pure short critics are worthless to me.

Max

MaxL
07-02-2007, 03:03 AM
Well, i think ctrl_buffer its easy to use, i see nothing wrong with that,and you also have a nice tonemmaper when using ctrl_buffer.

I found people struggling aroung crtl_buffer, and I had to redesign that node GUI to make it a bit easier to use for normal people. But as I sayd in the previus message, using a free node as the very base of your rendering pipeline (it have to be in all shaders in the scene in order to get the proper render passes) it's a big risk if you are designing your workflow around it. This is unless you have also the source code and you can (or pay someone) to recompile it if the original developer don't want to maintain it anymore.

Another way its using the maya layer system, and maya shaders as they are supported with mr base dll

This can be a nightmare solution. Using maya render layers and reference for example is a pure nightmare where you loose overrides and your scene get corrupted quite frequently. No one of the big studios I know is using Maya render layers. ANd I know a lot od people working in small studios complaining a lot about this.

And again, as i said i think the GI quality in final render its not as good as in mr, and thats my opinion and my experience, as ive seen ppl using final render in Max.

Cool with that. This is your opinion and it counts in this comparison, as well as other people saying it's far more easier and faster in fR to setup a good GI.

And when i said that mental ray was far superior, i was not talking about mental ray for maya only.

We are talking about rendering in maya. If MR is super cool in XSI, this is OT. If Mentalray for maya have a cheap translator you must count this in your equation, just because you have to deal with it.

I was talking about the fact that you can write your own shaders with it, you can create your own interface for the shaders if you want, using phenomena, again among other things.

We have shading SDK too, and the way you can integrate shaders inside maya is even more easier end powerfull. The shader SDK is not released in the main installer because at this stage things are evolving very fast, and at each new service release, custom shaders should be recompiled, making a problem to share shaders on the internet. But this will change very soon. Very soon. And trust me, there is no shader you can implement in MR you can't implement easily in fR.

But you are right, if you need to implement your custom shaders NOW, you need prman (or RI compatible engines like Air) or MR. Very soon there will be also fR in this short list.

Max

dagon1978
07-02-2007, 04:57 AM
It's always the same old story. You do something new, and people says: "I don't like it, it's not like the one I was using before...". You do something similar to make people comfortable and people says: "Hey there is nothing new in this stuff, there is not worth for it.". http://www.cebas.de/forums/smf/Smileys/default/smiley.gif Ah the irony...

But waiting for people to post their versions of the shared scenes, let's keep up the discussion.

What is a good shader for you?
You seems to be quite a technical person using MR. Do you like uber shaders with lot of parameters, like the architectural shader? Or do you like smaller more generic shaders that once connected together you can really really do whatever you want?
If you have and you are able to use these last ones, you don't need to wait for kind people to share their specialized custom shaders on the net, like mia_*, ctrl_*. You could already do 99.9% of that stuff with the things you already got.

Uber shaders are totally worthless in the film industry (VFX, animated features), but they are appreciated by architects who are not experts to create specialized shading networks.
Each of us is different, but I wouldn't expect from a rendering technically skilled person to mention the arch&design shader as a cool thing, neither the default MR custom shaders, because they are even far too low level to be practical, even for skilled people that knows exactly how to use them. Also because they don't provide a simple unified way to combine them together, they don't provide a way to easily manage render passes (absolute common practice in the industry since several years).

I would like to heard you sincere opinion.
Once again pure short critics are worthless to me.

Max

Max, seems just ironic you was talking about a shader "much much better designed that the one coming with MR for maya" but you (i dont know if really you, but fR) are copying exactly this shader in one of his releases, i'm not saying that i dont like it, i think it's very good for the fR4max users (in facts i was referring on this and some other shaders, some posts ago)

but, answering to your question, i'm not a technical person, i dont work in VFX, i dont work in animated features, i'm a graphic designer and i work for design and architecture... and for my hobby too ;)
so, arch&design is very good for me
but, i would like to know why do you think it's not pratical in production... i mean the "practical" reasons ;)


Mat

MaxL
07-02-2007, 05:55 AM
Max, seems just ironic you was talking about a shader "much much better designed that the one coming with MR for maya" but you (i dont know if really you, but fR) are copying exactly this shader in one of his releases, i'm not saying that i dont like it, i think it's very good for the fR4max users (in facts i was referring on this and some other shaders, some posts ago)

Different actions are taken for different reasons.
The Architectural shader for fR has been added to 3dsMax to make 3dsMax people using mentalray confortable with fR (since that shader seems to be so popular). This specific shader has been added "almost as-is" to Maya too, to maintain a sort of similar feeling on both platforms.
Other shaders are on development right now. There will be super complex shaders (like architectural) and other smaller shaders. Smaller but powerful and I can't tell you much on this.

but, answering to your question, i'm not a technical person, i dont work in VFX, i dont work in animated features, i'm a graphic designer and i work for design and architecture... and for my hobby too

Pleased to meet you ;)
I work on the film industry in a large studio, I'm mainly a lighter/compositor. (and yes, I develop finalRender translator too... life some time is hard)

arch&design is very good for me
but, i would like to know why do you think it's not pratical in production... i mean the "practical" reasons

Sure! For VFX or animated features we never deliver an image out of the 3d render. The 60% of the image quality is boosted in comp, not through painting or cheating, but by interactively fine tuning all light component aspects.
For me and my colleagues it is a must be able to obtain from a single render pass all those information. We can't prepare different set of shaders to have speculars, then diffuse, diffuse without shadows, reflection and so on.. with the complexity we have to manage this would literally kill us.
A shader that doesn't take care of this doesn't provide me the required set of features.
You could say: "I could use several archi-shaders, and turn on and off the single aspects I want, pass by pass", but in my workflow (not mine, but VFX/animation typical workflow) this is wrong for three reasons:

1) You need to break down a big shader in components, ending up having many big shaders, all the same but with different flags turned on/off. So you loose the "easy-to-use" part of that shader. You need to understand how to split the complex shader, same complexity as connecting smaller specialized shaders, but the problem is seen in the opposite direction, how to split instead of how to sum.
If you want to tweak your material you need to change the same properties in many different nodes. In the pressure of the deadline, this workflow is easy to break.

2) You have an all in one shader and you can't replace its parts (single features) with better in house solutions. Like, the uber shader provides a cool SSS part, but you have a custom ambient occlusion shader that renders much quickly. You can't plug your AmbOcc shader into the archi&design, because it's a black box, so you are stuck.

3) You can't reuse computationally expensive parts. You can't bake the occlusion to textures, because the shader doesn't allow you to do this. You need to backe the whole shader, but then you can't change any aspect. (this is connected to point 2.. if you could replace that part with your shader who implements reading/writing of baked occlusion, you would be able to solve the problem). But this not just about of occlusion, but everything. I'll keep mentioning the occlusion just for simplicity.
If I have a small network of smaller shaders, I can reuse the occlusion shader in many parts of the shading network by just replugging the same occlusion node, like on specular, diffuse, scatter, output occlusion in its own... The render engine will compute that node only once per sample, sharing the cost off the node across many effect the node is used for. But fractioning the shader in parts to extract render elements, occlusion could be computed many times instead. This would have a drastic negative impact on rendering performance.

There are more reasons, but in my opinion those three are the most important.

Max

coccosoids
07-02-2007, 06:23 AM
The main reason I'm not eager yet to try out a side by side comparison of MR and FR is
because of the lack of such a shader (in maya) as Arch&Design... I will try though... Just
to get a feel of the speed and quality the current implementation has in maya and hopefully
get a feeling of what is to come. ;)

lazzhar
07-02-2007, 07:52 AM
I'm really surprised that mates here are saying that some renderers are not able to render the same quality as other renderers when they have both very similar tools.
I'd venture saying that in most cases, when you are working with Maya Shaders and rendering in MR, it's only a matter of a switch to FR and you should get very close results. And again in term of "smooth" and "grain free" GI, I didn't see anything coming from MR that could compete with what Vray or FR are able torender out... or it's just me guys?!

coccosoids
07-02-2007, 08:15 AM
I wouldn't say that... Instead: "MR it IS capable of delivering smooth and grain free GI (in
stills at least, as I have no experience in animation) but at what cost?!"
Speed is an issue... So in these times you can't afford to render a typical interior for
4 to 5 hours (when you require like 5, 6 shots) and even more so just to find out when the
render finishes that you've got aliasing artefacts, some hotspots, fairy dust in your glossy
and so on... ;)

Michael-McCarthy
07-02-2007, 03:11 PM
If you guys didn't see this we also released another 30 mins of fRee training on fR for Maya last week. This looks at MTD, some motion blur and GI working together here:

http://forums.cgsociety.org/showthread.php?f=59&t=511807&highlight=finalRender

We are keeping them coming and to date have over 8 hours of fRee finalRender video training up on the web for our users (c4d, 3dsmax, Maya)


Thanks a bunch
Michael McCarthy

dagon1978
07-02-2007, 07:42 PM
Different actions are taken for different reasons.
The Architectural shader for fR has been added to 3dsMax to make 3dsMax people using mentalray confortable with fR (since that shader seems to be so popular). This specific shader has been added "almost as-is" to Maya too, to maintain a sort of similar feeling on both platforms.
Other shaders are on development right now. There will be super complex shaders (like architectural) and other smaller shaders. Smaller but powerful and I can't tell you much on this.


It's clear for me, but i simply dont want to copy something that i dont like, instead of do it better...
or at least i wouldn't talk as a bad solution
in italian i call it "sputare sul piatto in cui si mangia" ;)


Pleased to meet you ;)
I work on the film industry in a large studio, I'm mainly a lighter/compositor. (and yes, I develop finalRender translator too... life some time is hard)

Pleased to meet you, i know who you are
i followed your fR for Maya progress from the beginning steps in Renderglobal.org ;)
and i know your name also because you are a great example of an italian doing great things in the VFX




Sure! For VFX or animated features we never deliver an image out of the 3d render. The 60% of the image quality is boosted in comp, not through painting or cheating, but by interactively fine tuning all light component aspects.
For me and my colleagues it is a must be able to obtain from a single render pass all those information. We can't prepare different set of shaders to have speculars, then diffuse, diffuse without shadows, reflection and so on.. with the complexity we have to manage this would literally kill us.
A shader that doesn't take care of this doesn't provide me the required set of features.
You could say: "I could use several archi-shaders, and turn on and off the single aspects I want, pass by pass", but in my workflow (not mine, but VFX/animation typical workflow) this is wrong for three reasons:

1) You need to break down a big shader in components, ending up having many big shaders, all the same but with different flags turned on/off. So you loose the "easy-to-use" part of that shader. You need to understand how to split the complex shader, same complexity as connecting smaller specialized shaders, but the problem is seen in the opposite direction, how to split instead of how to sum.
If you want to tweak your material you need to change the same properties in many different nodes. In the pressure of the deadline, this workflow is easy to break.


mmm this is not the way to go...
i'm not as skilled as you but, believe me, i'm using mia_material with pass :)
is not that difficult to make it work in a wrapper...
ok, there's something you have to deal with, but ehy, i think a technical supervisor in a VXF/ani workflow wouldn't have any problem to make it work...
again, i'm not in this field, so, i could be wrong (i'm just talking about my non-technical experience)



2) You have an all in one shader and you can't replace its parts (single features) with better in house solutions. Like, the uber shader provides a cool SSS part, but you have a custom ambient occlusion shader that renders much quickly. You can't plug your AmbOcc shader into the archi&design, because it's a black box, so you are stuck.


3) You can't reuse computationally expensive parts. You can't bake the occlusion to textures, because the shader doesn't allow you to do this. You need to backe the whole shader, but then you can't change any aspect. (this is connected to point 2.. if you could replace that part with your shader who implements reading/writing of baked occlusion, you would be able to solve the problem). But this not just about of occlusion, but everything. I'll keep mentioning the occlusion just for simplicity.
If I have a small network of smaller shaders, I can reuse the occlusion shader in many parts of the shading network by just replugging the same occlusion node, like on specular, diffuse, scatter, output occlusion in its own... The render engine will compute that node only once per sample, sharing the cost off the node across many effect the node is used for. But fractioning the shader in parts to extract render elements, occlusion could be computed many times instead. This would have a drastic negative impact on rendering performance.

There are more reasons, but in my opinion those three are the most important.

Max


mmm i can't understand the AO problem
we are talking about a B/W AO pass to comp in a multiply layer?
why dont do it as a separated pass?
i can't understand your need to have an AO interacting with specular, diffuse, scatter...
but i think there is not a great problem to make it work in this way with a layer shader ( mix8layer split all the layers for the buffer_api without problem...)




I'm really surprised that mates here are saying that some renderers are not able to render the same quality as other renderers when they have both very similar tools.
I'd venture saying that in most cases, when you are working with Maya Shaders and rendering in MR, it's only a matter of a switch to FR and you should get very close results. And again in term of "smooth" and "grain free" GI, I didn't see anything coming from MR that could compete with what Vray or FR are able torender out... or it's just me guys?!

it's just you ;)
but i'm with you for the quality question, it's just a matter of how the user works... both renderers are very powerful... the problem in fR is not really the GI quality

MaxL
07-02-2007, 11:17 PM
Thanks for the kind words about my reputation :)

Appreciated.




It's clear for me, but i simply dont want to copy something that i dont like, instead of do it better...
or at least i wouldn't talk as a bad solution
in italian i call it "sputare sul piatto in cui si mangia" ;)

Actually I said it can be good for architects, but it's not that good for VFX/animation.
A long time ago there wasn't shading networks, and people were forced to use super complex shaders that did a bit of everything.. well there wasn't any SSS, any occlusion, dispersion ,etc... 3dsMax is still a bit like that, you don't have true shading networks so you need dedicated shader for specific tasks.
Then many software started to offer shading networks to split the complexity of a single shader in smaller building blocks, offering more power and control, easier to learn, easier to share and recombine.
Now we are seeing a reversed trend, where people suddenly discovered the old fashion super complex shader+presets, and many (I'm not saying you reading) just don't care what the shader does.

This way to go prevents people to learn not only how to prepare a shader, but how to fix a problem. And you say people complaining: “this is slow, or this render so bad” in both fR, MR and many others. Just because they are used to find ready solutions, and when they have to play a bit, they don't know what to do. Or they starts looking for a tutorial.
Now I'm a bit drastic, But I'm amplifying it in purpose to make it clearer.
If I would have to interview a guy for an opened surfacing position, and he would show me an image telling me, "there is the mia shader in all parts"... I would say “ok see you in a few years, you aren't ready to work here”. But as I said, visualization and film have completely different needs!

I'm not criticizing, I'm describing different worlds.


I'm using mia_material with pass :)
is not that difficult to make it work in a wrapper...
ok, there's something you have to deal with, but ehy, i think a technical supervisor in a VXF/ani workflow wouldn't have any problem to make it work...

This is why almost all VFX houses write theyr own shaders, because commercial shaders/free shaders usually don't meet requirements for this job. And mia shaders are that kind of shaders.
They are cool to use "as-is", you forget about the complexity, you don't even need to know how they works. But if you need wrapping them, you have to understand and spend time to implement the wrap.. probably, at that, point is better to implement your solution. If you frequently want to use the scissor from a swiss knife, you don't create more confortable handles to attach to the scissor, you just pick up a true scissor.

mmm i can't understand the AO problem
we are talking about a B/W AO pass to comp in a multiply layer?
why dont do it as a separated pass?

i can't understand your need to have an AO interacting with specular, diffuse, scatter...
but i think there is not a great problem to make it work in this way with a layer shader ( mix8layer split all the layers for the buffer_api without problem...)


So why are there occlusion control in the mia shader? It may have sense for centain people that doesn't know exactely how to use occlusion, or that knows, but they don't have time to deal with.
If I need to apply an external occlusion the controls for occlusion in mia shaders loose sense to exists.

One of the first principle of software development is that in your software there should be one entity and only one (shader in this case) responsible for a specific task (occlusion). Well you can have many if they do different thing in different ways to obtain different results. But it sounds like different entity then ;). Sincne occlusion in mia shader is just a generic occlusion, it shoudn't be there at all, instead it should offer a slot to connect my favorite occlusion shader, and this only.
But it's my personal opinion on how shaders should be designed.

I'm writing shaders now for fR. If you love mia shaders, you may hate mine... or you may appreciate a rather different approach. I'm trying to push fR as harded as I can into the VFX/animation industry, and in this thread I can feel more "archi & desing" artists.. so maybe I'M in the wrong spot here.

Max

MaxL
07-02-2007, 11:19 PM
ouch looking at the last posts we did... this is getting longer and longer.... people will never read all this stuff.

Jozvex
07-02-2007, 11:49 PM
Some people will! ;)

It's a good thread.

:thumbsup:

dagon1978
07-03-2007, 01:39 AM
Thanks for the kind words about my reputation :)

Appreciated.

Actually I said it can be good for architects, but it's not that good for VFX/animation.
A long time ago there wasn't shading networks, and people were forced to use super complex shaders that did a bit of everything.. well there wasn't any SSS, any occlusion, dispersion ,etc... 3dsMax is still a bit like that, you don't have true shading networks so you need dedicated shader for specific tasks.
Then many software started to offer shading networks to split the complexity of a single shader in smaller building blocks, offering more power and control, easier to learn, easier to share and recombine.
Now we are seeing a reversed trend, where people suddenly discovered the old fashion super complex shader+presets, and many (I'm not saying you reading) just don't care what the shader does.

This way to go prevents people to learn not only how to prepare a shader, but how to fix a problem. And you say people complaining: “this is slow, or this render so bad” in both fR, MR and many others. Just because they are used to find ready solutions, and when they have to play a bit, they don't know what to do. Or they starts looking for a tutorial.
Now I'm a bit drastic, But I'm amplifying it in purpose to make it clearer.
If I would have to interview a guy for an opened surfacing position, and he would show me an image telling me, "there is the mia shader in all parts"... I would say “ok see you in a few years, you aren't ready to work here”. But as I said, visualization and film have completely different needs!

I'm not criticizing, I'm describing different worlds.


so, i'll probably never work in a VFX studio, actually i'm working only with this shader :D eheh




This is why almost all VFX houses write theyr own shaders, because commercial shaders/free shaders usually don't meet requirements for this job. And mia shaders are that kind of shaders.
They are cool to use "as-is", you forget about the complexity, you don't even need to know how they works. But if you need wrapping them, you have to understand and spend time to implement the wrap.. probably, at that, point is better to implement your solution. If you frequently want to use the scissor from a swiss knife, you don't create more confortable handles to attach to the scissor, you just pick up a true scissor.


now you have to "compose" your simple shaders in networks and these networks can output many passes, ok, that's good
with the mia you have to understand how it work, create 1 time your wrapper and use this one
you probably lose a bit of control, but now the work can be done by a less technical user :) like me :scream:



So why are there occlusion control in the mia shader? It may have sense for centain people that doesn't know exactely how to use occlusion, or that knows, but they don't have time to deal with.
If I need to apply an external occlusion the controls for occlusion in mia shaders loose sense to exists.


One of the first principle of software development is that in your software there should be one entity and only one (shader in this case) responsible for a specific task (occlusion). Well you can have many if they do different thing in different ways to obtain different results. But it sounds like different entity then ;). Sincne occlusion in mia shader is just a generic occlusion, it shoudn't be there at all, instead it should offer a slot to connect my favorite occlusion shader, and this only.
But it's my personal opinion on how shaders should be designed.


yeah, i think you're right, the concept is surely better in the way you describe... there's just a problem, actually the occlusion in the mia shader is not a generic occlusion ;)



I'm writing shaders now for fR. If you love mia shaders, you may hate mine... or you may appreciate a rather different approach. I'm trying to push fR as harded as I can into the VFX/animation industry, and in this thread I can feel more "archi & desing" artists.. so maybe I'M in the wrong spot here.

Max

maybe.. or maybe i'll love it... i dont know if i can't try ;)

mat

sacslacker
07-03-2007, 01:59 AM
trying to push fR as harded as I can into the VFX/animation industry, and in this thread I can feel more "archi & desing" artists.. so maybe I'M in the wrong spot here.

Naw,you're in the right spot and this is a great discussion! I agree with your stance on shader but I do understand the need for a more packaged approach for those that need rapid surfacing.

As for FR, I'm a fan. Will the FR SDK be released soon? Also, will we see some sort of hair/fur support? That is one thing I look forward to in FR. Rendering curves would be very nice.

Keep up the great work Max. I didn't realize you were working full time in VFX and writing the translator. That gives me a whole new perspective and respect for ya! The FR translator is a fast one and well integrated. Great work man.

MaxL
07-03-2007, 02:17 AM
now you have to "compose" your simple shaders in networks and these networks can output many passes, ok, that's good. With the mia you have to understand how it work, create 1 time your wrapper and use this one. You probably lose a bit of control, but now the work can be done by a less technical user :) like me :scream:

To design the wrapper you need a techical person. This guy can easily put togheter the shading network for that specific task. An average surfacing artist can redo those kind of shading networks in less than 10 minutes, starting form smaller building blocks.
The effort to design the wrapper can be similar, but in this second case the "less technical people" will find in his desk only what it really needs. Not a lot of disabled stuff. Shading networks can be easily reused and maintained in libraries. But every time you will resue them you will have full control over them, to change/add/remove.

What's the point to have a mix8. If you have 9 layers you need a mix20, and if you have 21? Maybe you need to stack a mix8 and a mix20... Ok, this is not that common, but it's the overall concept that doesn't convince me.

there's just a problem, actually the occlusion in the mia shader is not a generic occlusion ;)

Ok that's not a problem (don't focus on occlusion, it was just an example). That specail occlusion should ba a shader a part, which could be used connected to the mia shader, as well as connected to other materials. Maybe you wrote your custom BRDF function and you would like to use it with the cool occlusion coming with mia library... but you can't.
And if mia library comes with an occlusion shader and the architectural shader has the same occ algorithm built in, well that's wrong, don't you see redundancy on this?
Please remember we are talking about Maya, where shading networks are not only possible but they are the suggested workflow since its first release. Other 3d pakages may have other probelms/solutions.

Max

MaxL
07-03-2007, 02:27 AM
Naw,you're in the right spot and this is a great discussion! I agree with your stance on shader but I do understand the need for a more packaged approach for those that need rapid surfacing.

Yes! Yes! Yes! And that why they call it architectural shader, or bal-bla-bla whatever shader! Because they are magnificent if you need to do that kind of stuff. End even it they could be used as general purpose materials, the are not that ideal for every kind of job.

-"Will the FR SDK be released soon?" -> Yes indeed. It's 100% fully working since quaite a long time. All the shaders we have implemented has been written with that SDK, no second calss SDK here.
-"Will we see some sort of hair/fur support?" -> Yes.

-"The FR translator is a fast one and well integrated" -> Thanks, you should try the one in the comins SP3, it's almost N times as faster as many cores you have ;)

Max

coccosoids
07-03-2007, 06:55 AM
Oh lord!
The rendering revolution in maya has started! ;)
Anyway, I'm just hoping to see something like the mia in upcoming releases...
I wanted to translate yesterday one of my interior scenes but I gave up on it... :(
It's just too much fuss creating a ramp with different gray values for every material in the
scene (more than 30!!!).. and that's just the tip of the iceberg.

One suggestion: if you could come up with a script that automatically translates MIA
networks to FR you'll have a lot more 'followers' !! ;)

dagon1978
07-03-2007, 08:57 AM
To design the wrapper you need a techical person. This guy can easily put togheter the shading network for that specific task. An average surfacing artist can redo those kind of shading networks in less than 10 minutes, starting form smaller building blocks.
The effort to design the wrapper can be similar, but in this second case the "less technical people" will find in his desk only what it really needs. Not a lot of disabled stuff. Shading networks can be easily reused and maintained in libraries. But every time you will resue them you will have full control over them, to change/add/remove.

What's the point to have a mix8. If you have 9 layers you need a mix20, and if you have 21? Maybe you need to stack a mix8 and a mix20... Ok, this is not that common, but it's the overall concept that doesn't convince me.

and what's the problem to have a mix100layer if you want? :) i mean, for a skilled man is not that hard...
the effort to design the wrapper can be similar, but you have to do it just 1 time :) not every time you use it
and you can use presets to save your libraries
again, you lose a little bit of control, but you can think much more on the "artistic" point of view, not only of your "nodes" and how to make it work :)
it's a matter of points of view... i think your workflow is powerful, but a bit overdated... even your example it's not convincing me at all...

"If I would have to interview a guy for an opened surfacing position, and he would show me an image telling me, "there is the mia shader in all parts"... I would say “ok see you in a few years, you aren't ready to work here”. But as I said, visualization and film have completely different needs!"

now lets start thinking on the "artistic" part of your work, you are judging an artist work, but you think about how it make the work and not of his ability/sensibility to make a good lighting/shading work, it's a point of view, but i think is a bit "old-style"
i know, it's an "industry", but ehy, a good artist can surely learn how to make shader nodes work, a technical man difficultly can learn how become a good artist ;)


Ok that's not a problem (don't focus on occlusion, it was just an example). That specail occlusion should ba a shader a part, which could be used connected to the mia shader, as well as connected to other materials. Maybe you wrote your custom BRDF function and you would like to use it with the cool occlusion coming with mia library... but you can't.
And if mia library comes with an occlusion shader and the architectural shader has the same occ algorithm built in, well that's wrong, don't you see redundancy on this?
Please remember we are talking about Maya, where shading networks are not only possible but they are the suggested workflow since its first release. Other 3d pakages may have other probelms/solutions.

Max

i can't see redundancy at all max ;)
this is a problem of ergonomy, and this is part of my knowledge/work
the AO in the mia_mat is a matter of 1 click
for an ergonomic point of view this kind of occlusion has to be included in the core of a renderer (as in turtle or some other engines)
you're thinking with a VFX mind... but you're wrong in this point
you dont want to use it? u're not forced
this is not rendundancy, you're providing an easy setup for non technical people... if you are developing shaders for archviz people keep in mind this point ;)


the are not that ideal for every kind of job.


every? in the world? :)
eheh Max, let start to see your "better designed" shader for archviz, and then we can start to talk of what's "ideal" or not :)

MaxL
07-03-2007, 09:48 AM
Who said I'm going to write shaders for archviz? :)
My shaders will be mainly designed for films people. There is already plenty of solutions for archiviz, and other people (not me) are writing fR shaders for it.
But I don't want to uncover all my strategies ;)

It's better if we talk about what's already in the game, and not about what will come.


- "i know, it's an "industry", but ehy, a good artist can surely learn how to make shader nodes work, a technical man difficultly can learn how become a good artist ;)" -> This isn't that true. I saw a lot of good artists struggling and starving in front of tiny technical hurdles. Actually I saw several during Happy Feet too. Therefore they were unable to express their art (I'm always talking about the film industry).
In teams with good leaders and good look development artists/art directors, it's easier to make a technical people to deliver good images, rather than an artist to learn pipeline tools.
Why do you think in the last years recruiters ask more and more for scripting and programming skills. Don't trust me if you want. But this is the industry trend.

Even if this could be interesting, we should really go back to the track, I'm 110% OT now.

Max

dagon1978
07-03-2007, 10:34 AM
Who said I'm going to write shaders for archviz? :)
My shaders will be mainly designed for films people. There is already plenty of solutions for archiviz, and other people (not me) are writing fR shaders for it.
But I don't want to uncover all my strategies ;)

It's better if we talk about what's already in the game, and not about what will come.


so your shader is better designed for your tasks, not for all, otherwise you'll use it also for archviz ;)



- "i know, it's an "industry", but ehy, a good artist can surely learn how to make shader nodes work, a technical man difficultly can learn how become a good artist ;)" -> This isn't that true. I saw a lot of good artists struggling and starving in front of tiny technical hurdles. Actually I saw several during Happy Feet too. Therefore they were unable to express their art (I'm always talking about the film industry).
In teams with good leaders and good look development artists/art directors, it's easier to make a technical people to deliver good images, rather than an artist to learn pipeline tools.
Why do you think in the last years recruiters ask more and more for scripting and programming skills. Don't trust me if you want. But this is the industry trend.

Even if this could be interesting, we should really go back to the track, I'm 110% OT now.

Max

the mainly problem here is ergonomy

the industry trend is to use technical skilled people because you have the wrong pipeline (in my point of view), so you have to adapt the people to your work, and not (as a good job would be) the work to the people

BTW yes, we are OT ;)
nice to talk with you Max, keep up the good work

Mat

Michael-McCarthy
07-03-2007, 12:09 PM
Seems like a bit of back and forth. Just to make it clear:

We will have cool Arch vis and "all in one" Shaders (as soon as SP3 ships) and we will continue to release shaders like this for the comunity.

We will have more very powerful film/FX,and game shader nodes, and pipeline tools as Max is a master of these things :) (that is, more then the many we already do! )

We love to hear your feedback and hear what YOU want :) We are all about making the fastest, high quality, feature rich, and easy to use renderer around!

Thanks
Michael McCarthy
cebas

dagon1978
07-03-2007, 12:47 PM
Seems like a bit of back and forth. Just to make it clear:

We will have cool Arch vis and "all in one" Shaders (as soon as SP3 ships) and we will continue to release shaders like this for the comunity.

We will have more very powerful film/FX,and game shader nodes, and pipeline tools as Max is a master of these things :) (that is, more then the many we already do! )

We love to hear your feedback and hear what YOU want :) We are all about making the fastest, high quality, feature rich, and easy to use renderer around!

Thanks
Michael McCarthy
cebas

thanx :thumbsup:
but please, release a beta version too ;) i think it would be great for all the maya users

Mat

Michael-McCarthy
07-03-2007, 01:27 PM
We do not and will not have an open beta for our products as this does not suit our development pipeline needs.

Anyone that is interested in beta testing fR for maya please contact me directly at mDOTmccarthyATcebas.com with "fR for MAYA BETA" in the subject. What we have planed will blow your socks off! :) We are always looking for new and talented testers.

A bit about our beta program:

You need to be ether failure with fR or another comparable rendering package.

You need to have time and want to spend time TESTING. Our beta is for new ideas, testing our new features and contributing (not simply playing with or trying to use in your own productions).

Activity. We have some simple tasks and expect users to be active on our beta forums and reporting bugs. Inactive testers just get dropped.

As I said we are always looking for talented and motivated testers.

Thanks
Michael McCarthy

lazzhar
07-03-2007, 02:02 PM
A question to Michael McCarthy and MaxL.
Will the next update support features in Maya that weren't before? like FUR, Maya Hair ..ect ?
And thanks for your wonderful work you did by the way.

Michael-McCarthy
07-03-2007, 02:14 PM
We cant really talk to much about what will or will not be in the next SP. But it will have a bunch of fixes and new features. We are however working on fur and Hair.

No worries though it should hit the streets in a few weeks (barring anything to weird)

Thanks
Michael McCarthy

sacslacker
07-03-2007, 03:48 PM
Thanks, you should try the one in the comins SP3, it's almost N times as faster as many cores you have ;)

Max


I honestly can't wait!

MaxL
07-03-2007, 10:28 PM
Entering the beta test program is a small effort from your side (to accomplish the given tasks) but it's also a privilege be able to ask for features and fixes and see them implemented quickly in a fast releasing cycle.

Our service packs usaually contains more fixes than brand new features, however in this SP3 there is new cool stuff too.

Max

joie
07-11-2007, 04:26 PM
Hi;

I just want to ask something more if I may..., Is Finalrender able to tonemapping the rendered image? like MR does with the physical sky?.

Thankīs.

PS: I have tried to enter to the oficial finalrender forum and a virus came up!, please check that, is the second time is happens and I donīt think is me...

lazzhar
07-11-2007, 07:22 PM
Hi;

I just want to ask something more if I may..., Is Finalrender able to tonemapping the rendered image? like MR does with the physical sky?.
....

Yes it's already there from the begining.

joie
07-11-2007, 10:27 PM
And how do you tonemapping the image?, via camera attributes?, via shader?, via render globals?, is it a postprocess?

lazzhar
07-12-2007, 11:23 AM
It's from Render Global : General>FrameBuffer Options then enable: Color Mapping.

joie
07-12-2007, 01:08 PM
Thank you very much, before I worked without tonemapping at all, but, since MR introduced the arquitectural shaders I canīt live without it!, even if I donīt use those shaders at all..., tonemapping really brings your images to life, if finalrender is capable of doing that Iīm pretty sure I will give it a try when the demo is out.

Thankīs.

techmage
07-13-2007, 01:46 AM
Thank you very much, before I worked without tonemapping at all, but, since MR introduced the arquitectural shaders I canīt live without it!, even if I donīt use those shaders at all..., tonemapping really brings your images to life, if finalrender is capable of doing that Iīm pretty sure I will give it a try when the demo is out.

Thankīs.

But only the priviliged get the demo :D

joie
08-01-2007, 06:02 PM
Hi there;

I have one question..., the distributed rendergin only allows up to 10 CPUs to render one image..., is there any way to render with more than those?, may be more licenses?

MaxL
08-01-2007, 10:54 PM
the distributed rendergin only allows up to 10 CPUs to render one image..., is there any way to render with more than those?, may be more licenses?

Yes, each extra render license gives you access to up to 10 DR (Distributed Rendering) CPUs more, not depending on what or how many machines they are.
For example if you have a full license (that include the maya translator) and you buy three extra render license, you could have 40 CPUs running on your render.

Max

razorback
08-02-2007, 02:16 AM
Ok... I see grown men fight over it, definately alot of speculation, lots of talk about it, alot of hype going all round, people doing their own research and comparisons.... so instead of fighting over whos better looking, whos a retard and whos got the bigger schlong, i have to ask...

can i try it before i buy it (or is this mythical beast even out yet)? and where are those 8 hours of training videos that you guys are talking about?

From whats been said, it sounds like a great product, and would fit into our pipeline, hell if it does everything its promised to do (and do it well), i'd rebuild our pipeline around it. But its not gonna be easy for me to convince a dozen people to move from something they know just coz it sounds like a good idea...

Peace.

MaxL
08-02-2007, 05:27 AM
razorback: fair enough. if you want to try it you can write to cebas and ask for a try license. But don't not ask it right now. We are preparing for siggraph, there is lot of stuff to and doprobably nobody will have the time to aswer. Just wait two weeks.

I know it's always hard to convince people to use something different. And that's why is important to try yourself, because then you can show what you can do with it.

Max

dailyqumiya
08-02-2007, 08:12 AM
I don't use Maya:)

MaxL
08-02-2007, 10:18 PM
I don't use Maya:)
Good on you! You can have a long life even without it ;) Unless other people in this Maya forum will kill you for this.. ahahahah, muahahahahah

joie
08-03-2007, 01:16 PM
when the help docs say that if you want to render across the whole network (animation frames I mean) you have to use the same way you use with MAYA software..., What does it mean?, MAYA software donīt do network render at all..., please explain that :)

razorback
08-03-2007, 06:43 PM
@MaxL : i look forward to testing it. got a photoreal car job coming in, so it'll be a good place to test it.

@joie : it means whatever network rendering platform you're currently working with eg:muster, royal render, qube. if you don't network render now, then it wont network render for you.

@dailyqumiya : good. go away.

joie
08-03-2007, 07:55 PM
I guess that was all about...

Yes I use one render manager, but it doesnīt support finalrender at all... :(

joie
08-04-2007, 12:22 AM
I guess that was all about...

Yes I use one render manager, but it doesnīt support finalrender at all... :(

razorback
08-04-2007, 03:07 AM
@joie : doesn't your render manager support command line ?

I know this is totally the wrong place to ask, but since someone touched on it anyways....

Regarding the tonemapping workflow ... how isit different from NOT using the tonemapping workflow and rendering out 32bit floats and doing color correction at the end ?

MaxL
08-04-2007, 08:32 AM
when the help docs say that if you want to render across the whole network (animation frames I mean) you have to use the same way you use with MAYA software..., What does it mean?, MAYA software donīt do network render at all..., please explain that :)

The explaination is linear. Whatver you do to render with maya on several computer, you can do that with fR as well. If you use a render farm manager, a dispatcher, to su submit to the farm, you have to do the same with fR. If you use batch files you can do the same with fR.

After investigation we found out that is not that usefull if we would provide a network rendering manager along with fR. For sure you already have maya and MentalRay (because it comes with maya) an probably you want to be able to do network rendering the same way with all of them. If we would implement a propertary solution to dispatch fR jobs on a farm, the most of the people won't use it because they woul probably need to dispatch jobs from other software and render engines using a shared render farm. For this reason it's mandatory that a dispatcher must support several different softwares and render engines. This is provide a unified workflow and to avoid that several dispatchers running at the same time will run concurrent jobs on the same machine, like a 3d renderng and a compositing rendering at the same time, busting the computer memory and generating swap and failures.

There are already a lot of good dispatchers on the market, Royal Render, Deadline, Rush, Smedge, and many more. Several of them already have dedicated settings for finalRender for maya. We don't want to compete with those software houses. They are our friends, not our enemies :)

Maybe the documentation doesn't explain well this tipic. What we want to say is that if you know how to use maya and what you need to do to make maya render on a network, then you know also what to do to render with fR, because it's exactly the same.
To render with maya from a command line you have to run:
render -r sw [flags] [filename]
To render with fR:
render -r fr [flags] [filename]

If you don't agree or if you have other ideas, you are very welcome to explaine your position. We are here also to listen.

Max

MaxL
08-04-2007, 08:44 AM
@joie : doesn't your render manager support command line ?
If your render manager doesn't support fR and doesn't support command line either, ask the developer why :)

Regarding the tonemapping workflow ... how isit different from NOT using the tonemapping workflow and rendering out 32bit floats and doing color correction at the end ?

Absolutely no difference at all, but many people especially in the architectural/design industry doesn't comp to get the final result, they just get the image that comes out of the 3d rendering.

As Lighting TD I never use tone mapping (prman doesn't even have a such feature), I render linear colors and I do my personal mapping in comp using color curvers. This gives you much more control over the grading, but of course it is one step more to do. That can be expensive if for your job you have few time or not time at all.

Max

razorback
08-04-2007, 10:09 AM
Absolutely no difference at all, but many people especially in the architectural/design industry doesn't comp to get the final result, they just get the image that comes out of the 3d rendering.

As Lighting TD I never use tone mapping (prman doesn't even have a such feature), I render linear colors and I do my personal mapping in comp using color curvers. This gives you much more control over the grading, but of course it is one step more to do. That can be expensive if for your job you have few time or not time at all.
Max

Ah allright. Its no wonder this whole tonemapper craze started with the mia_architectural shader. But its kinda risky rendering out tonemapped images, the color levels would totally be destroyed if you pushed it too hard.

I thought I was the only one who found it odd that you'd use a tonemapper to test the shaders, then disconnect it before rendering. Then rebuilding it in comp.

Okay enough. Sorry for hijacking the FRfM thread.

techmage
08-21-2007, 09:29 PM
Service pack 3 is out

http://www.cebas.com/news/read.php?UD=10-7888-35-788&NID=237

Looks like some nice stuff, anyone run tests?

joie
08-22-2007, 11:24 AM
Is FGIMerger.exe working good?, it takes all the files generated of each frame and deletes them! Ž_Ž

MaxL
08-23-2007, 08:23 AM
I should try, it's long time I don't use it, I don't remember the behavior, but if you find any problem or you think a tool should work in a different way, here is the users forum, the best place to report things, ask for features or fixes:
http://www.cebas.de/forums/smf/index.php

For the mia shader lovers, SP3 adds a shader that is pretty much the same (fR-Architectural).

Max

joie
08-23-2007, 01:16 PM
As I said before, the last two times I tried to enter the cebas forum, my firewall detected viruses coming from it and closed my IE explorer..., Do you have any reports like these?, Iīd love to see the forum, but now Iīm afraid of doing so... :(

MaxL
08-24-2007, 12:45 AM
I don't know about IE it's years I don't use it.
I have firewall and antivirus daily updated, I've never got any problem. Maybe your settings are too restrictive about cookies? I'm just guessing, I don't know.

As alternative you can email to me.

Max

joie
08-24-2007, 01:17 PM
Iīll try with Firefox...

But my problem wasnīt the cookies or settings restriction, one virus came up from the site and got my PC down, I had to reinstall everything..., then I visited again the forum and again the same virus came up and needed to reinstall again, so I fed up with it. (strange)

brainspoon
08-29-2007, 02:20 PM
Hi,

Is it possible to set a scene gamma in Finalrender like in Maya?
The only place I where one can specify a gamma is in the AQMC settings which only changes the gamma of the global illumination calculation.

The advantage of the correct gamma is that the falloff of lights looks much more natural.

brainspoon
08-30-2007, 10:17 AM
I found some gamma parameters in the defaultFinalRenderSettings node, but they do not have any effect.
They are called "display gamma", "file in gamma" and "file out gamma".

MaxL
08-31-2007, 12:14 AM
Those controls are not used at the moment, that's why they are not in the GUI.
One question. Do you comp or post process your scenes?
Because applying gamma just out of the render prevents you to be able to reproduce the beauty result by comping together the RE.
You may be aware that if you render diffuse and specular as separated images you cannot apply gamma on them and sum them together to obtain a beauty with gamma on it...

gamma(a+b, gammaValue) != gamma(a, gammaValue) + gamma(b, gammaValue).

This whole tone mapping and gamma thing done in the render is just wrong.
I can add more controls about this, it would just take few hours. But I'm sure then people will start complaining that in comp they cannot reproduce the same result as they got in the beauty.

Think that the principle of using gamma to change the light falloff is not correct. You are using the wrong tool to do a right thing. It may looks fine in your example, but it's actually affecting the textures colors and everything elase. It will bahave in a wrong way when multiple lights are lighting the same point (it happens almost all the times), just look at the formula I wrote before. You may consider using an intensity curve on the light instead.
But if everybody wants gamma controls, then I can turn them back on as it was at the beginning of the alpha testing.

Max

brainspoon
08-31-2007, 03:13 PM
Would be great to have them to work in a linear way.

CGTalk Moderation
08-31-2007, 03:13 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.