Octane render

959 views
Skip to first unread message

Emilio Hernandez

unread,
Feb 13, 2013, 9:00:31 AM2/13/13
to soft...@listproc.autodesk.com
Anyone here has used octane render within Softimage?

If so I will appreciate your comments and your point of view of Octane vs other renders you have used.

And if you have used Arnold and Octane, which one you prefer.


I have the Octane experience only from the videos and the demo.  It seems to me that it is like Maxwell but a lot faster.

Cheers!


--

Toonafish

unread,
Feb 13, 2013, 9:57:44 AM2/13/13
to soft...@listproc.autodesk.com
yep, been playing with it the last few days and I'm really impressed by the speed and quality of Octane.

But comparing both Arnold and Octane, all I can say is for most stuff Arnold is still better suited and production proven, but Octane is a nice addition. Some scenes I tested that took forever to render in Arnold render in just a few minutes in Octane.

Can't wait for the interactive viewport plugin for Softimage though.

- Ronald

Ed Manning

unread,
Feb 13, 2013, 10:08:47 AM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
mind sharing your system config?  I'm debating what card to put in a box for Octane

Toonafish

unread,
Feb 13, 2013, 10:24:03 AM2/13/13
to Ed Manning, soft...@listproc.autodesk.com
an I7 3930 with a Geforce GTX 960 with 2048 MB, but I'd go for more memory on the card if possible.

Jimmy Marrero

unread,
Feb 13, 2013, 10:27:51 AM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
If you have the time, can you explain the workflow with using both octane and softimage without an interactive viewport in soft??


Thanks

Toonafish

unread,
Feb 13, 2013, 10:37:27 AM2/13/13
to soft...@listproc.autodesk.com
Just watch the video in this thread and install the addon: http://render.otoy.com/forum/viewtopic.php?f=29&t=24365

Make sure you also download and replace the dll that is mentioned later on in the thread, and if your SI camera does not export correctly, try changing the default units in the Octane prefs to meters.

- Ronald

Emilio Hernandez

unread,
Feb 13, 2013, 11:14:34 AM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Is the interactive window in softimage plugin available yet?  Personally I don't like switching between apps to see how the render is coming.


2013/2/13 Toonafish <ron...@toonafish.nl>



--

Toonafish

unread,
Feb 13, 2013, 11:23:22 AM2/13/13
to Emilio Hernandez, soft...@listproc.autodesk.com
Nope, this is the best there is for SI at the moment as far as I know.

- Ronald

Emilio Hernandez

unread,
Feb 13, 2013, 11:37:04 AM2/13/13
to soft...@listproc.autodesk.com
Thanks Ronald.  Too bad developers are only aiming at Maya and Max from AD "suite".   I admire the job that independent developers are putting into this.  I think I'll pass until the viewport is integrated in Softimage.  Hope it will not take as long as V-ray...

Even that Maxwell for rendering uses it's own.  The workflow beween Softimage and Maxwell is great with Fire. No fooling around or tweaking things outside the software.  I hope that Next Limit will add the GPU processing and perhaps it will be as fast as Octane.


2013/2/13 Toonafish <ron...@toonafish.nl>



--

Toonafish

unread,
Feb 13, 2013, 12:56:12 PM2/13/13
to soft...@listproc.autodesk.com
They are working on an integrated version for SI as well, donno when it's going to be released but from the video previews things looks promising so far : http://render.otoy.com/forum/viewtopic.php?f=29&t=23700

- Ronald

Toonafish

unread,
Feb 13, 2013, 1:41:59 PM2/13/13
to soft...@listproc.autodesk.com
just to show off the speed ;-)

stopped after 10 minutes rendering: https://dl.dropbox.com/u/4271217/octane_classroom_10min.png




On 2/13/2013 17:37, Emilio Hernandez wrote:

Simon van de Lagemaat

unread,
Feb 13, 2013, 1:51:31 PM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
My experience was that it took a loooong time to reduce fine noise.  It got to the level you have fairly speedily but after that the curve really began to flatten out.  Maybe it was just my vanilla 500 series nvidia card was too slow, not sure.

wireframex Gmail

unread,
Feb 13, 2013, 1:55:16 PM2/13/13
to soft...@listproc.autodesk.com
Octane is a very good renderer

I own 2 GTX 580 with 3Go RAM each with EVGA SR-2 motherboard

More card u have less time render you have the only bad thing is the power consumption ...

I hope to use the integrated version soon :)

Phil

Toonafish

unread,
Feb 13, 2013, 2:51:39 PM2/13/13
to soft...@listproc.autodesk.com
yep, getting rid of the fine noise can take long, just like with all "unbiased" renderers. I'll do another test tomorrow to see how long it takes to get a nice, clean render in Octane.

Would be nice to compare Maxwell, Octane and Indigo.

- Ronald
-- 
Ronald van Vemden
-----------------------------------------------
3D Graphics & Animation
Cyberfish Laboratories | www.cyberfish.nl
Toonafish | www.toonafish.nl
tel. +31(0)20 5289291
fax  +31(0)20 5289292
email: ron...@toonafish.nl 

Steve Pratt

unread,
Feb 13, 2013, 5:33:42 PM2/13/13
to XSi Mail List
Been using Octane for about 18 months to create hi-res stills for advertising. Mostly product vis stuff. It's fantastic, I love it. Especially with the recent post effects they've added, glare and bloom.
Scene and lighting setup is a breeze.
I have it running on a box with a GTX580 3gb for Octane and a Quadro FX1800 for Soft. From my experience with the GTX580 fine noise is no issue. I'm due to upgrade the workstation this year and looking to get one with 2 x GTX590's. That would be 2048 cores all up instead of the measly 512 I currently have.  ;-)

I use Mental Ray now exclusively for animations and now hate the struggle it is to get photo-realistic images out of it. I've also used LuxRender and Cycles. Neither come close to Octane.

The ice is all Soft and Octane, comped in Photoshop.
All Soft and Octane.

Cheers.
--
"Twenty years from now you will be more disappointed with the things that you didn't do than by the ones you did do. So throw off the bowlines. Sail away from the safe harbour. Catch the trade winds in your sails. Explore. Dream. Discover." - Mark Twain

Emilio Hernandez

unread,
Feb 13, 2013, 6:06:39 PM2/13/13
to soft...@listproc.autodesk.com
If you can provide the scene I can test in Maxwell in six Core i7 Extreme to compare.  Of course my guess is that Octane will beat it in time.  But at least is a good parameter.  The other will be if someone with Arnold can do the same.

James De Colling

unread,
Feb 13, 2013, 7:01:14 PM2/13/13
to soft...@listproc.autodesk.com

What are the limitations with the gpu renderers?  Last time I looked at it,  subdivision / displacement was unsupported,  I assume that's changed now?

Gene Crucean

unread,
Feb 13, 2013, 7:02:01 PM2/13/13
to soft...@listproc.autodesk.com
Sign me up for the Arnold render. I'll snap those two in halffff. aarrRRAGHAAHH!!! KILL MEEE. I'M RIIIGHT HERRREE. WHAT ARE YOU WAITING FORRRR!!!!!


...hehe. Sorry about the caps. The quote isn't the same unless you say it loud... even in your own head.



On Wed, Feb 13, 2013 at 3:06 PM, Emilio Hernandez <emi...@e-roja.com> wrote:
If you can provide the scene I can test in Maxwell in six Core i7 Extreme to compare.  Of course my guess is that Octane will beat it in time.  But at least is a good parameter.  The other will be if someone with Arnold can do the same.




--
Gene Crucean - Emmy winning - Oscar nominated CG Supervisor / iOS-OSX Developer / Filmmaker / Photographer
** Freelance for hire **
www.genecrucean.com

~~ Please use my website's contact form on www.genecrucean.com for any personal emails. Thanks. I may not get them at this address. ~~

Toonafish

unread,
Feb 13, 2013, 7:08:59 PM2/13/13
to soft...@listproc.autodesk.com
Here's the classroom scene I used. It's not mine, I think it was created
by Alex Sandri. I can do an Arnold test as well tomorrow, but I suspect
it will take muuucho longer to get this quality without a lot of optimizing.

https://dl.dropbox.com/u/4271217/Classroom_SI.rar

- Ronald

Rob Chapman

unread,
Feb 13, 2013, 7:09:11 PM2/13/13
to soft...@listproc.autodesk.com
did you sir, just throw your glove before you?  http://en.wikipedia.org/wiki/Duel#Offense_and_satisfaction

there is only one way to settle this....        

:D 

Toonafish

unread,
Feb 13, 2013, 7:16:25 PM2/13/13
to soft...@listproc.autodesk.com
yep, Octane does not support subdivisions or displacements, also only camera motion blur.

- Ronald

Gene Crucean

unread,
Feb 13, 2013, 7:22:07 PM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
One down one to go. NEXT!!! aaarrghahrahgh!!!

...hehe. Ok I'll stop now.









GET TO THE CHOPPPAAAAA!!!!!

Toonafish

unread,
Feb 13, 2013, 7:31:32 PM2/13/13
to Gene Crucean, soft...@listproc.autodesk.com
yes, better start preparing for some....humiliation !  ;-)

- Ronald

Simon van de Lagemaat

unread,
Feb 13, 2013, 7:40:07 PM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
He'll just subdivide everything ;-)

Gene Crucean

unread,
Feb 13, 2013, 7:46:32 PM2/13/13
to Simon van de Lagemaat, soft...@listproc.autodesk.com
Exactly. I'll just provide a scene that's 100% organic :)

And it was going to use 64GB of memory ;)

Massimo Galluzzo

unread,
Feb 13, 2013, 7:49:55 PM2/13/13
to soft...@listproc.autodesk.com
Time to smoke you all with Mental Ray

Rob Chapman

unread,
Feb 13, 2013, 8:00:55 PM2/13/13
to Massimo Galluzzo, soft...@listproc.autodesk.com
excellent...think we are missing Monsieurs Vray and 3delight.. and I do believe Mitsuba san is due to arrive into the fray very soon.. or is this a GPU versus CPU duel only.. clearly some conditions have to be met. the classroom scene, is this to be the Field of honour? 

Toonafish

unread,
Feb 13, 2013, 1:37:19 PM2/13/13
to soft...@listproc.autodesk.com
just to show off the speed ;-)

stopped after 10 minutes rendering:



On 2/13/2013 17:37, Emilio Hernandez wrote:

Andreas Bystrom

unread,
Feb 13, 2013, 9:10:03 PM2/13/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
prman can render that classroom in 7.5 minutes with better quality
--
Andreas Byström
Weta Digital
octane_classroom_10min.png

Toonafish

unread,
Feb 14, 2013, 4:38:36 AM2/14/13
to soft...@listproc.autodesk.com
That would be swell. But I think we'd also might have to include setup time then. I'm sure a with a lot of hours tweaking, testing and some smart trickery you can get faster render speeds in a lot of renderers, but the whole point with these unbiased renderers is that you hardly have to send any time tweaking to get a realistic looking render.

I think I spent about 30 minutes on the classroom scene adding and adjusting materials and tweaking the lighting and camera before hitting the render button. The only render optimization I did was switch from direct lighting to path tracing.

- Ronald

Stephan Woermann

unread,
Feb 14, 2013, 5:34:14 AM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Try to use the PMC kernel. You should get better results in lesser time, especially on interior scenes...

Stephan


2013/2/14 Toonafish <ron...@toonafish.nl>

Ciaran Moloney

unread,
Feb 14, 2013, 5:52:56 AM2/14/13
to soft...@listproc.autodesk.com
You know the rules, Andreas. Renders or it didn't happen.

Toonafish

unread,
Feb 14, 2013, 6:16:13 AM2/14/13
to Stephan Woermann, soft...@listproc.autodesk.com
That were my thoughts as well, but I it looks worse after 10 minutes then with pathtracing.

https://dl.dropbox.com/u/4271217/octane_classroom_pmc-10min.png

- Ronald

Stefan Kubicek

unread,
Feb 14, 2013, 6:27:35 AM2/14/13
to soft...@listproc.autodesk.com
Define "quality"!

> prman can render that classroom in 7.5 minutes with better quality
>
> On Thu, Feb 14, 2013 at 7:37 AM, Toonafish <ron...@toonafish.nl> wrote:
>
>> just to show off the speed ;-)
>>
>> stopped after 10 minutes rendering:
>>
>>
>>
>>>>>> --
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>>
>>>
>>>
>>
>>
>> --
>>
>>
>>
>
>


--
-------------------------------------------
Stefan Kubicek Co-founder
-------------------------------------------
keyvis digital imagery
Wehrgasse 9 - Grᅵner Hof
1050 Vienna Austria
Phone: +43/699/12614231
--- www.keyvis.at ste...@keyvis.at ---
-- This email and its attachments are
--confidential and for the recipient only--

Toonafish

unread,
Feb 14, 2013, 7:11:09 AM2/14/13
to Andreas Bystrom, soft...@listproc.autodesk.com
show me the pudding !

- Ronald



On 2/14/2013 03:10, Andreas Bystrom wrote:
prman can render that classroom in 7.5 minutes with better quality

On Thu, Feb 14, 2013 at 7:37 AM, Toonafish <ron...@toonafish.nl> wrote:
just to show off the speed ;-)

stopped after 10 minutes rendering:


Toonafish

unread,
Feb 14, 2013, 7:28:34 AM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
This time I let the classroom bake for 45 minutes, still a tiny bit of
noise in the unsharp dark areas because of the DOF. But I don't mind.

https://dl.dropbox.com/u/4271217/octane_classroom_pathtracing-16bounces-45min.png

So......where's all them 5 minute high quality Prman and MentalRay
renders. I suppose you guys are still fiddling with knobs and sliders ;-)

- Ronald

Ed Manning

unread,
Feb 14, 2013, 10:03:55 AM2/14/13
to soft...@listproc.autodesk.com
Will someone PLEASE put some edgeloops or hard edges on those blackboards so they don't shade like domes? ;-)

Steffen Dünner

unread,
Feb 14, 2013, 10:06:30 AM2/14/13
to soft...@listproc.autodesk.com

2013/2/14 Ed Manning <etmt...@gmail.com>

Will someone PLEASE put some edgeloops or hard edges on those blackboards so they don't shade like domes?

+1 ;)

My first thought as well.

Stefan Kubicek

unread,
Feb 14, 2013, 10:09:26 AM2/14/13
to soft...@listproc.autodesk.com
LOL, I thought so too this very second :-)
Still love the overall lighting though.


> Will someone PLEASE put some edgeloops or hard edges on those blackboards
> so they don't shade like domes? ;-)
>
> On Thu, Feb 14, 2013 at 7:28 AM, Toonafish <ron...@toonafish.nl> wrote:
>
>> This time I let the classroom bake for 45 minutes, still a tiny bit of
>> noise in the unsharp dark areas because of the DOF. But I don't mind.
>>
>> https://dl.dropbox.com/u/**4271217/octane_classroom_**
>> pathtracing-16bounces-45min.**png<https://dl.dropbox.com/u/4271217/octane_classroom_pathtracing-16bounces-45min.png>
>>
>> So......where's all them 5 minute high quality Prman and MentalRay
>> renders. I suppose you guys are still fiddling with knobs and sliders ;-)
>>
>> - Ronald
>>
>


Toonafish

unread,
Feb 14, 2013, 10:17:05 AM2/14/13
to soft...@listproc.autodesk.com
...that's not a blackboard, it's a workout bench hanging from the wall :-)

- Ronald

Eric Lampi

unread,
Feb 14, 2013, 11:29:45 AM2/14/13
to soft...@listproc.autodesk.com
That's priceless.

You can't slip anything by Ed.
--
Freelance 3D and VFX animator

http://vimeopro.com/user7979713/3d-work

Massimo Galluzzo

unread,
Feb 14, 2013, 11:44:38 AM2/14/13
to soft...@listproc.autodesk.com
Didnt have much time to play with FG Occhiolino
Here is Mental
 
9 mins and 30 seconds
4 mins fg precalc
 
 
From: Eric Lampi
Sent: Thursday, February 14, 2013 5:29 PM
Subject: Re: Octane render
 
wlEmoticon-winkingsmile[1].png

Toonafish

unread,
Feb 14, 2013, 12:34:43 PM2/14/13
to soft...@listproc.autodesk.com
Arnold took 2 Hours and 10 minutes with 5 Diffuse bounces and AA set to
10. Setting the diffuse bounces to 16 as in Octane was just slowing
Arnold down too much.

https://dl.dropbox.com/u/4271217/Arnold_Classroom_AA10_DiffSamples-7_DiffRays-5_2h-10min.png

Still some fine noise in the DOF, and weird aliasing artifacts in the
high contrast areas around the windows and the lights.

- Ronald


Gene Crucean

unread,
Feb 14, 2013, 1:01:57 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Will you send me that scene so I can look at the settings? I'm curious how you set it up. The sampling can make a huge difference.

Christian Keller

unread,
Feb 14, 2013, 1:01:16 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
looks like a filtering issue. lanzcos ?
you might get away with less bounces, without noticabel quality loss.
with some tricks you could get a better and faster result

cheers,
chris
-- 
christian keller
visual effects|direction

+49 179 69 36 248
chri...@me.com
http://vimeo.com/channels/96149

Steven Caron

unread,
Feb 14, 2013, 1:11:10 PM2/14/13
to soft...@listproc.autodesk.com
mind sharing that scene ronald? did you set the AA to 10 because of DOF? cause on a still image i wouldn't use that high of AA with that high of diffuse samples.

as you knew, with more optimizing you probably can get better result. but octane is pretty awesome to be able to do what its doing for you, and with the graphics card having sooo many little gpu processors i think it will always outperform arnold in a scene like this.

so, how many threads did arnold use? cause if you only have 2 core proc then i dont see that being entirely fair. i see the value of octane and a graphics card purchase over an arnold license and a machine with the same number of processors as in your graphics card.

s


On Thu, Feb 14, 2013 at 9:34 AM, Toonafish <ron...@toonafish.nl> wrote:

Paul Griswold

unread,
Feb 14, 2013, 1:24:36 PM2/14/13
to soft...@listproc.autodesk.com
It'd be interesting to add a 2nd comparison scene that wasn't an arch-vis interior type shot.  

Arnold always struggles with interiors with lots of bounces.  

-Paul

Christian Gotzinger

unread,
Feb 14, 2013, 1:45:37 PM2/14/13
to Massimo Galluzzo, soft...@listproc.autodesk.com
Nice job there! I find that it doesn't match the Octane version in terms of quality (not your fault):
a) It looks "floaty". That's just the problem with FG. You'd need insane settings to get the flawless contact shadows of an unbiased renderer. People like to combat that by using ambient occlusion, but usually the result looks terrible to me, with dark corners everywhere.
b) Sampling noise of a biased renderer looks unnatural (see the reflection on the table at the very left). Unbiased renderers otoh have very organic noise, so often it's fine for there to be some.

I've used Octane many times for still shots and even for animations. Setup times are short, render speed is good and the quality is great. Plenty of features are still missing (such as motion blur), but once the Softimage integration is ready it should become a real option for those who don't like to use an external renderer.
wlEmoticon-winkingsmile[1].png

Toonafish

unread,
Feb 14, 2013, 2:06:56 PM2/14/13
to Christian Keller, soft...@listproc.autodesk.com
I used Mitchell filtering, I like the sharpness. But it could be that's
causing some problems with the high contrast areas in this render.

- Ronald

Toonafish

unread,
Feb 14, 2013, 2:09:30 PM2/14/13
to Gene Crucean, soft...@listproc.autodesk.com
sure, you can download it here: https://dl.dropbox.com/u/4271217/Arnold_Classroom.rar

- Ronald

Toonafish

unread,
Feb 14, 2013, 2:26:37 PM2/14/13
to soft...@listproc.autodesk.com
scene file : https://dl.dropbox.com/u/4271217/Arnold_Classroom.rar

I'm rendering on a 6 core i7 3930 overclocked to 4 Ghz, so Arnold is using 12 threads. With AA set lower the DOF is very noisy. But you're right, maybe I could lower the diffuse samples a little.

- Ronald

Toonafish

unread,
Feb 14, 2013, 2:31:03 PM2/14/13
to Massimo Galluzzo, soft...@listproc.autodesk.com
Nice, but FG is very blurry still. Maybe play some more to get a little closer to the Octane quality if you have the time ?
I'm really curious to see how long MR takes to get that kind of quality, and how long it takes to get there.

- Ronald

Emilio Hernandez

unread,
Feb 14, 2013, 2:37:14 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Any Maxwell attempts?


2013/2/14 Toonafish <ron...@toonafish.nl>



--

Steven Caron

unread,
Feb 14, 2013, 2:39:54 PM2/14/13
to soft...@listproc.autodesk.com
thats a nice machine!

you said you had a geoforce GTX 960. do you mean 690?


if so, that is over 3000 cores!

Christian Gotzinger

unread,
Feb 14, 2013, 2:47:45 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
When using high AA samples (which is necessary for DOF or motion blur) I believe you can pretty much keep diffuse samples down to 1 or 2. You have to oversample a lot for the DOF, and this AA oversampling takes care of diffuse areas as well. Also, for this scene with the large windows you may get away with 2 or 3 bounces.

Toonafish

unread,
Feb 14, 2013, 2:52:02 PM2/14/13
to soft...@listproc.autodesk.com
oops, that was s typo. It's a GTX 680 with only 1536 cuda cores.

- Ronald

Rob Chapman

unread,
Feb 14, 2013, 3:09:22 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
ok these are good conditions of entry for this renderer battle arena one must not spend more than 30 minutes on combined material config / lighting / render knob tweaking.    

but you can let it go for 2 hours...! sheesh, I'd be very unhappy with 5 minutes :) but its a good proof that more time = better quality..?

Massimo Galluzzo

unread,
Feb 14, 2013, 3:20:21 PM2/14/13
to soft...@listproc.autodesk.com
Damn! then i’m disqualified with MR... A bocca aperta
 
Sent: Thursday, February 14, 2013 9:09 PM
Subject: Re: Octane render
 
wlEmoticon-openmouthedsmile[1].png

Sven Constable

unread,
Feb 14, 2013, 3:20:42 PM2/14/13
to soft...@listproc.autodesk.com

I use 2011… so can anyone share a dotxsi version of the scene?

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Rob Chapman
Sent: Thursday, February 14, 2013 21:09
To: ron...@toonafish.nl; soft...@listproc.autodesk.com
Subject: Re: Octane render

 

ok these are good conditions of entry for this renderer battle arena one must not spend more than 30 minutes on combined material config / lighting / render knob tweaking.    

Toonafish

unread,
Feb 14, 2013, 3:38:17 PM2/14/13
to Rob Chapman, soft...@listproc.autodesk.com
Or, maybe we could use a different scene then the classroom that becomes available for download at the moment the contest starts, and see how fast the results come in with the various renderers. That would be fun :-)

- Ronald

Toonafish

unread,
Feb 14, 2013, 3:41:48 PM2/14/13
to soft...@listproc.autodesk.com

Sven Constable

unread,
Feb 14, 2013, 4:00:51 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com

Thank you.

Toonafish

unread,
Feb 14, 2013, 5:50:11 PM2/14/13
to Christian Gotzinger, soft...@listproc.autodesk.com
That was some good advice, thanks. I Reduced the diffuse samples to 4, diffuse rays to 3 but had to increase AA to 20 to get an acceptable amount of noise. The quality is about the same as before but the render time was reduced to 1h:33minutes. Also got rid of the sampling errors by clamping the sampling values.

But still much more fine noise overall then the Octane render, so I'm rendering one with a diffuse sample setting of 6.

https://dl.dropbox.com/u/4271217/Arnold_Classroom_AA20_DiffSamples-4_DiffRays-3_1h-33min.png

So far Octane is still the winner by a landslide. It's very quiet all of a sudden,  Maxwell, Vray anyone ?!

- Ronald

Steven Caron

unread,
Feb 14, 2013, 6:20:55 PM2/14/13
to soft...@listproc.autodesk.com
i think we all knew that comparing straight ray tracing speed, without textures, displacement, subdivisions, and deformation motion blur... octane was going to win. i mean thousands of processors that are great at doing this type of work compared to ~12 more generalized processors, octane takes advantage of the gpu and arnold doesn't... winner octane!

Leoung O'Young

unread,
Feb 14, 2013, 7:41:24 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Interesting thread, is Octane optimized for the GTX 600's series of card?

Leonard Koch

unread,
Feb 14, 2013, 7:50:20 PM2/14/13
to soft...@listproc.autodesk.com

This might not be 100% correct, but octane is I think simply optimized for cuda, and NVidia optimizes all their cards at least partly for cuda.

Rob Chapman

unread,
Feb 14, 2013, 8:25:47 PM2/14/13
to soft...@listproc.autodesk.com
ok heres 1/2 an hour in the classroom with Mental Ray. bit longer actually because tried to go down the importons irradience route but did not have much luck in the fiddle allocation so had to resort to GI. its architectural materials just trying to get it to work with one big portal light out of the window.


oh and yeah 10 minutes to render on an i7.. cpu.

Steven Caron

unread,
Feb 14, 2013, 8:41:17 PM2/14/13
to soft...@listproc.autodesk.com
gawd i dislike mental ray

Leoung O'Young

unread,
Feb 14, 2013, 9:51:52 PM2/14/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Would the gtx be twice as fast than?
Is there a sweet spot for these cards?

I am think of adding another gtx 560 to my existing 560 to do some testing


On 2/14/2013 2:52 PM, Toonafish wrote:

Steve Pratt

unread,
Feb 14, 2013, 10:07:45 PM2/14/13
to XSi Mail List
Octane - 3hrs on a GTX580
I used the inbuilt daylight system and portal nodes on the windows. Never used the portal nodes before and something doesn't feel quite right. I think there should be more intensity coming from the windows, but I ran out of tweak-time.




On 14 February 2013 01:00, Emilio Hernandez <emi...@e-roja.com> wrote:
Anyone here has used octane render within Softimage?

If so I will appreciate your comments and your point of view of Octane vs other renders you have used.

And if you have used Arnold and Octane, which one you prefer.


I have the Octane experience only from the videos and the demo.  It seems to me that it is like Maxwell but a lot faster.

Cheers!


--




--
"Twenty years from now you will be more disappointed with the things that you didn't do than by the ones you did do. So throw off the bowlines. Sail away from the safe harbour. Catch the trade winds in your sails. Explore. Dream. Discover." - Mark Twain

Massimo Galluzzo

unread,
Feb 14, 2013, 11:31:51 PM2/14/13
to soft...@listproc.autodesk.com
So, i decided to pause Mental Ray (so damn frustrating) a bit and give iRay a go. with xsi 2013
I’m using a crappy nvidia geforce 570 with 1.3gb ram.
 
First render 8 minutes
 
Second render 19 minutes
 
Third render with dof let it go for 26 minutes
 
Too bad we dont have this engine with a proper integration by AD, still cant believe they didnt implement it.
Not like people arent asking, maybe there are too few asking Sorriso
 
Anyway keep in mind my gfx card is quite bad, i might be able to test maxwell but i have to go to a friend studio since i dont have it.
 
Max
wlEmoticon-smile[1].png

Christian Keller

unread,
Feb 15, 2013, 5:14:57 AM2/15/13
to soft...@listproc.autodesk.com
you should always use the gaussian filter for unclamped images. all those sharpen filters
have a spot where the curve is going under zero and cause negative pixelvalues, surrounding areas
with a lot of light. that´s the black borders around your windows and lights

cheers,
Chris
-- 
christian keller
visual effects|direction

+49 179 69 36 248
chri...@me.com
http://vimeo.com/channels/96149

Thomas Volkmann

unread,
Feb 15, 2013, 5:19:06 AM2/15/13
to soft...@listproc.autodesk.com
Here is a quick Blender/Cycles test:
 
ca 35min on a Dual 6Core(HT) @ 2.66GHz ... I saved 2 cores for doing other stuff in the meantime though.
There are some normals fucked up and I haven't tested cuda rendering yet, but you get the idea.
Basically I just wanted to see if I ever have to use mental ray again (if we hadn't our Arnold licences) ...I don't think so :)
 
cheers,
Thomas
 
 
Massimo Galluzzo <mas...@massimogalluzzo.it> hat am 15. Februar 2013 um 05:31 geschrieben:

Arvid Björn

unread,
Feb 15, 2013, 5:29:50 AM2/15/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Here's 5 minutes 25sec MentalRay render, using only physical sky and portal lights. I took the Arnold scene linked above and switched all the materials to architectural with my usual MR settings, obviously no textures. Removed the Arnold stuff, took a few minutes of setting up in total. Looks pretty clean to me, no knob-tweaking.


Inline image 1
image.jpeg

Rob Chapman

unread,
Feb 15, 2013, 5:36:53 AM2/15/13
to soft...@listproc.autodesk.com
ah good, I knew MR wasn't as bad as my 1/2 hour fiddle demonstrated,  so it was physical sky I was missing on top of the portal. as well as was gonna say that if someone who had rendered any interiors before would have made a much better job on the render with MR if the lights were setup better. this is like the 2nd time Ive ever used a portal light so am not familiar with how to set it up or when to use it for render with photons or importons whatever.
image.jpeg

Toonafish

unread,
Feb 15, 2013, 6:19:27 AM2/15/13
to Arvid Björn, soft...@listproc.autodesk.com
Wow, that looks very good !! Looks like MR is back in the race ;-)

Could you share some of the render settings  ?

- Ronald

Arvid Björn

unread,
Feb 15, 2013, 6:45:53 AM2/15/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Sure, use unified sampling, min 1, max 100, quality 2 (in this case), enable FG (default settings in this case), 16px tilesize, spiral pattern.
Use Architectural material on everything, set all samples everywhere to 1. Enable AO in Architectural, set AO shadow color to black.

Use the quality slider to adjust quality / reduce noise.

I have a script that defaults Architectural to turn off reflections and set the above settings, and another script to set up MR as described, so it's pretty quick to set up.

Other than that I just added pysical sky and portal lights and tweaked the intensities a bit.

Stefan Andersson

unread,
Feb 15, 2013, 6:52:15 AM2/15/13
to soft...@listproc.autodesk.com
It has always been easy to tweak mental ray to render a decent image. The trick is when you add more things to make the image complex, and then.... move the camera :) Add some hair and motion blur into the mix, with an animated character.

Then report back the render times. And this goes for all renderers, not just mental ray.

The limitation with GPU render engines is ram. They just can't load everything you need into ram. And if you get a GPU raid that can handle it, you have spent a lot more money than what you could have bought a CPU farm for.

With that said, I see a lot of possibilities for GPU rendering and visualizations or previz.... I'm going a bit off-topic.... as you were people. :)

regards
stefan




On Fri, Feb 15, 2013 at 12:19 PM, Toonafish <ron...@toonafish.nl> wrote:
Wow, that looks very good !! Looks like MR is back in the race ;-)

Could you share some of the render settings  ?



--
Stefan Andersson | Digital Janitor
blog | showreel | twitter | LinkedIn | cell: +46-73-6268850 | skype:sanders3d


Sven Constable

unread,
Feb 15, 2013, 7:09:01 AM2/15/13
to soft...@listproc.autodesk.com

Since we comparing gpu to cpu renderers… let the cpu renderer use at least all cores we can throw at them :)

 

4 satellite nodes (8cores each) and 1 workstation 6core.

mentalray 3.8

16 min per frame+ 5 min irradiance precalc.

Way to go, octane!  ;)

 

http://www.imagefront.de/tmp/classroom_mentalray3_8.png

 

Shading, lighting and tweaking took me much longer than 30 minutes but I'm on 2011. No unified sampling and I also had some crashes with satellites that seems to be fixed in newer version of mr.

I guess the rendertime on just the workstation would be around 1hour. The workstation is faster than the nodes on the farm and there is also some network overhead when using satellite rendering but anyway, rendertimes aren't bad. I always liked mental ray.

 

cheers,

sven

Toonafish

unread,
Feb 15, 2013, 7:28:41 AM2/15/13
to soft...@listproc.autodesk.com
Great one, how did you get those sharp shadows ? Did you use a lightsource ?

I noticed the window intensity with portal nodes also depends on the position and power of the sun.

- Ronald

Ed Manning

unread,
Feb 15, 2013, 7:33:40 AM2/15/13
to soft...@listproc.autodesk.com
On Fri, Feb 15, 2013 at 6:45 AM, Arvid Björn <arvid...@gmail.com> wrote:
Sure, use unified sampling, min 1, max 100, quality 2 (in this case), enable FG (default settings in this case), 16px tilesize, spiral pattern.
Use Architectural material on everything, set all samples everywhere to 1. Enable AO in Architectural, set AO shadow color to black.


Hi Arvid --

Why tile size 16?  I ask because I often use this setting (all the time for previews, often for renders) in order to minimize the amount of idle proc time while the last couple buckets finish, but have been told lately that it's actually a time-waster since the renderer has to resample all the bucket edges.  I'm not sure which helps/hurts more, although I guess maybe there'd be a way to figure out a rule of thumb based on bucket time vs. number of cores...

Massimo Galluzzo

unread,
Feb 15, 2013, 7:44:00 AM2/15/13
to soft...@listproc.autodesk.com
Thats a nice one Steven, did you use FG or irradiance particles?

Toonafish

unread,
Feb 15, 2013, 7:46:57 AM2/15/13
to soft...@listproc.autodesk.com
You took the word rendertweaking to a whole new level to get your rendertimes down ;-)

Not as realistic as the unbiased renders I think, but very nice results. I'd be curious to see an actual rendertime on a single workstation with a decent CPU.

- Ronald

Sven Constable

unread,
Feb 15, 2013, 7:59:24 AM2/15/13
to Massimo Galluzzo, soft...@listproc.autodesk.com

irradiance particles. They are close to fg with portal lights in a overall look, but when it comes to small details, they win. Especially with a lot of diffuse lighting.

Btw my name is sven, not steven :)

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Massimo Galluzzo
Sent: Friday, February 15, 2013 13:44
To: soft...@listproc.autodesk.com
Subject: Re: Octane render

 

Thats a nice one Steven, did you use FG or irradiance particles?

Arvid Björn

unread,
Feb 15, 2013, 7:59:29 AM2/15/13
to soft...@listproc.autodesk.com
Mostly to avoid cores idling, with large tile sizes you could really be wasting time, especially for something that only occupies a small part of the image, or a really detailed patch that's left to last running on a single core after everything else is finished. But it's part of my optimization routine, so occasionally I increase the size. Never heard of the extra sampling issue, can't imagine it would over-shadow the 1:8 ratio of cores rendering simultaneously though, unless it's a scene with extremely even detail level.

Emilio Hernandez

unread,
Feb 15, 2013, 9:52:49 AM2/15/13
to soft...@listproc.autodesk.com
Try MR with this advanced MIA Material.

Haven't tried yet, but it says it speeds things up and the behavior is better.

http://felixgeremus.com/?p=108



2013/2/15 Arvid Björn <arvid...@gmail.com>



--

Michal Doniec

unread,
Feb 15, 2013, 12:07:03 PM2/15/13
to softimage
These are LCD tellys, not blackboards. They don't use blackboards in schools anymore.


On 14 February 2013 15:03, Ed Manning <etmt...@gmail.com> wrote:
Will someone PLEASE put some edgeloops or hard edges on those blackboards so they don't shade like domes? ;-)


On Thu, Feb 14, 2013 at 7:28 AM, Toonafish <ron...@toonafish.nl> wrote:
This time I let the classroom bake for 45 minutes, still a tiny bit of noise in the unsharp dark areas because of the DOF.  But I don't mind.

https://dl.dropbox.com/u/4271217/octane_classroom_pathtracing-16bounces-45min.png

So......where's all them 5 minute high quality Prman and MentalRay renders. I suppose you guys are still fiddling with knobs and sliders ;-)

- Ronald

Nicolas Burtnyk

unread,
Feb 15, 2013, 2:00:15 PM2/15/13
to soft...@listproc.autodesk.com
Hey Sven,

Nice MR renders.  Do you mind sharing the scene?

-Nicolas

Tim Crowson

unread,
Feb 15, 2013, 7:24:33 PM2/15/13
to soft...@listproc.autodesk.com
I downloaded the scene and noticed it was called 'modo_classroom' so I figured why not render it in modo.
The attached image took 14min on my i7930 at 960x540. Didn't go for phsyically-accurate materials, but I'd like to. Still some noise in a few areas.

--

 

Tim Crowson
Lead CG Artist

Magnetic Dreams, Inc.
2525 Lebanon Pike, Building C. Nashville, TN 37214
Ph  615.885.6801 | Fax  615.889.4768 | www.magneticdreams.com
tim.c...@magneticdreams.com

Mihai Iliuta

unread,
Feb 16, 2013, 2:05:25 AM2/16/13
to soft...@listproc.autodesk.com
Hi there,

Hopefully this will be in the right thread....here are some classroom tests with Maxwell Render.

Ronald, I just took your scene and applied Maxwell materials trying to match the look from your tests and for the lighting I just used Skydome.

My machine is an i7 2600K @ 3.4GhZ, and according to our benchwell test (www.maxwellrender.com/benchwell) your machine should run it in about 10 min, mine took 14min, so 40% slower.

Time 15min:
https://www.dropbox.com/s/2uw6akqot5yun4d/15min.jpg

45 min:
https://www.dropbox.com/s/110ec8qwea2dl39/45min.jpg

2h:
https://www.dropbox.com/s/d3ty0h696nxwbuh/2h.jpg

7h:
https://www.dropbox.com/s/yqdclhiwj36a9b6/7h.jpg


The last one was to try and match your 45min Octane render, although some areas looked cleaner sooner (DOF areas especially). It's true that it takes a much longer time to get a super clean image, but usually a small amount of denoise is enough instead of waiting...

So to compare

the 7h render would take about 5 hours on your machine.
the 2h render matches in noise your Arnold 1.33min render and would take about 1h20min on your machine.

For the 7h render I turned up the intensity of the ceiling lights using Multilight to match your render (but they would never look that strong in reality if you expose for daylight, unless you want the electric company to come arrest you...).

Quality wise, perhaps in my not so unbiased opinion, there is more life in the Maxwell render, mainly due to not limiting bounces. I don't understand what the point is to have only 2-3 bounces of GI because it starts looking dead and like an AO pass - greyness everywhere. Look especially in the chrome material, radiator area, in between the wood boards on the desk. All these things add up in my opinion and other tests will show it better. I prefer to know I'll get a render that looks and feel like a photo, not a render that looks like yet another....render.

Also most of our users will take up 2GB of RAM just for loading their textures, no matter the resolution of the render and many want to render at at least A4 print size which is about 3500px wide at 300dpi.

Finally, maybe we should maybe do a test with thousands of particles, hair, DOF, MB, sharp caustics??? Bring it ON!!!


330.gif

Massimo Galluzzo

unread,
Feb 16, 2013, 6:37:52 AM2/16/13
to soft...@listproc.autodesk.com
Maxwell is amazing, the quality of light is impressive, thanks a lot for those tests Mihai Sorriso
 
 
Sent: Saturday, February 16, 2013 8:05 AM
Subject: Re: Octane render
 
wlEmoticon-smile[1].png
330.gif

Toonafish

unread,
Feb 16, 2013, 6:54:19 AM2/16/13
to soft...@listproc.autodesk.com
Great ! A Maxwell render. I agree, quality wise 7 hour render rivals the Octane 45 minute render. Rendertimes are higher, that was to be expected, but it's nice to see some proof.  You are right I think, a lot of the renders don't look as realistic due to low diffuse bounces or use of occlusion. I think this Maxwell render is the first one that is better quality then the Octane render.

And the lighting looks better then my green room too :-)

How many bounces did you use ? I left Octane at the default 16.

- Ronald

Toonafish

unread,
Feb 16, 2013, 7:11:08 AM2/16/13
to soft...@listproc.autodesk.com
Nice, a Modo render. Looks very smooth, but less detail in the shadow areas then the Maxwel or Octane renders.

Could you share some render settings ?

- Ronald

Tim Crowson

unread,
Feb 16, 2013, 10:35:17 AM2/16/13
to ron...@toonafish.nl, soft...@listproc.autodesk.com
Yeah, this wasn't using brute force MC. Used IC instead, with 12 bounces, 256 IC rays, with an Irradiance rate of 5.0. I did this at the end of the day yesterday and didn't get around to a purely MC render. AA at 128, shading rates mostly at 0.1.

-Tim
--


Octavian Ureche

unread,
Feb 16, 2013, 10:51:32 AM2/16/13
to soft...@listproc.autodesk.com
Hey guys,

Here's my take on it with vray in xsi on an i7-3770 at 3.5 ghz (in low priority mode cuz i had work to do in the meantime). 
Single dome light with an hdri as color texture, 32 light samples.
Materials are just vray standards with 32 samples for the glossy reflections & no interpolation (it's faster, but i think it is not as accurate). 
Adaptive dmc with lighcache and irradiance map (glossy rays computed from lightcache - faster), and 3 gi bounces (yeah, i was lazy).
10-15 min setup time. Still noisy, but i wanted to keep it in the 10 mins range, just to see how it might hold up agains the other players.
Render is 720p resolution.

Cheers,
Octav
Vray_Classroom.jpg

Alok Gandhi

unread,
Feb 16, 2013, 11:03:44 AM2/16/13
to soft...@listproc.autodesk.com
Looks very nice. The light shades seems wierdly transparent though and the LCD are too green, maybe you changed the colour. But I like the more contrast.

Sent from my iPhone
<Vray_Classroom.jpg>

Octavian Ureche

unread,
Feb 16, 2013, 11:21:28 AM2/16/13
to soft...@listproc.autodesk.com

That was an lcd? Damn, i thought it was a blackboard of sorts and gave it a greenish tint to make it more interesting.
Figures, when u grow up in eastern europe, and you see a classroom...an lcd in front of it is the last thing that crosses your mind. :)

Gene Crucean

unread,
Feb 16, 2013, 12:39:40 PM2/16/13
to soft...@listproc.autodesk.com
What do you guys all say we keep these at the same resolution and camera angle? :)

Btw, LCD's? Really? Maybe in Japan, or umm.... the Samsung factory's internal school. In America we have chalkboards with the occasional projector. Not 200" LCD's. That seems crazy to me. Can someone post links to pics of actual setup's like that?
--
Gene Crucean - Emmy winning - Oscar nominated CG Supervisor / iOS-OSX Developer / Filmmaker / Photographer
** Freelance for hire **
www.genecrucean.com

~~ Please use my website's contact form on www.genecrucean.com for any personal emails. Thanks. I may not get them at this address. ~~

Alok Gandhi

unread,
Feb 16, 2013, 1:02:33 PM2/16/13
to soft...@listproc.autodesk.com
I was also not sure of the LCD, although now in India we do have that but I studied on a chalkboard. Someone in this thread posted before that these were LCD that is why I pointed out.
--

Gene Crucean

unread,
Feb 16, 2013, 1:06:17 PM2/16/13
to soft...@listproc.autodesk.com
Sorry Alok, that wasn't directed at you in anyway. I was just amazed that places on earth have this? I'm still amazed to be honest.
It is loading more messages.
0 new messages