PDA

View Full Version : fR-2: the old teapot test


AdamT
11-19-2005, 08:08 PM
I came upon the famous teapot scene when searching through my old GI test folder. I was quite shocked that fR-2 was able to crank this out with next to no artifacts in under two minutes (1:56).

This was using my AMD X2, Xeon dual 2.4, and P4 3.2.

DynamicRealism
11-19-2005, 08:23 PM
looks like ur harnessing the power of the skyportal =) so it seems

AdamT
11-19-2005, 08:38 PM
Yes indeedy.

Out of curiousity I tried the same scene with AR2. I'm at around 8 minutes and the pre-pass hasn't even started to render.

[EDIT: 12 minutes and still no prepass]

[EDIT2: 20 minutes and still no prepass]

JoelOtron
11-19-2005, 08:42 PM
Would it be helpful to do a Maxwell comparison as well--or is the current version of mr not good for benchmarks? (I know theres another version on its way).

Interesting.

Could someone briefly explain the skyportal concept? (Or point me to a buried thread?)

AdamT
11-19-2005, 08:45 PM
Yeah, I think it would be better to wait for the new M~R to do benchmarks. It should be much faster than the old beta.

As I understand it the sky portal acts like sort of a funnel to concentrate rays from outside into openings in a structure. For practical purposes you cover an opening with a Cinema area light and give it an fR light tag with the sky portal option enabled.

AdamT
11-19-2005, 08:56 PM
Tada! At 25 minutes the AR prepass has begun to render.

Continuumx
11-19-2005, 09:04 PM
I came upon the famous teapot scene when searching through my old GI test folder. I was quite shocked that fR-2 was able to crank this out with next to no artifacts in under two minutes (1:56).

This was using my AMD X2, Xeon dual 2.4, and P4 3.2.

These are nice tests, but the speed issue is a moot point when you are using 5 as opposed to a single processer for the speed test?!?!? What is the point? I am sure if I networked 5 rather speedy machines together and purchased netrender, I could crank out some nifty quick GI renders using AR or even Maxwell. I mean with 5 very high level processors I would expect faster times. 1 min 56 seconds, it should be fast!

Most of the speed results are using multiple processors so it does not help to evaluate the results in a unbiased manner. I mean I am interested in FR but I cannot do a unbiased consideration when all the test results are using 4-8 processors networked together.

I apoligize if I sound a little skeptical here, but it just appears as apples to oranges unless you conduct these speed tests in similar manner when comparing different engines for speed.

In the maxwell comparisons, they use a low end single processor for speed comparisons which is very conservative and practical.

Nice image by the way!

xeno3d
11-19-2005, 09:10 PM
Would it be possible to share that scene file?

AdamT
11-19-2005, 09:26 PM
Hehe, well I didn't say it was unbiased--wrong forum. :) But for comparison purposes, it seems like all of the computers working together are about 1.7-8x faster than my X2 box alone. So roughly speeking say it would take 4 minutes on that machine.

Xeno: I'll put up the scene in a little while. I've got to go out for a bit.

MJV
11-19-2005, 10:31 PM
Considering that Cinema can't even bucket render on a network, I'd say that's pretty impressive. I might consider someday getting FR but only if the speed and quality is dramatically better than what I can get from AR.

AdamT
11-19-2005, 10:50 PM
IMO speed is unquestionably better. Quality is another question. It's way too early to give any conclusions, but it *looks* like fR gets a lot of its speed advantage by cheating on quality, i.e., you can get very fast and artifact-free GI because it's blending the snot out of the samples. It'll take some time to figure out how fast it is when high quality settings are used.

Here's the teapot scene for anyone who wants to try it: http://www.3danvil.com/fR-2/teapot_fR-2.rar

It has the settings for my 2 minute fR render.

Continuumx
11-20-2005, 01:04 AM
IMO speed is unquestionably better. Quality is another question. It's way too early to give any conclusions, but it *looks* like fR gets a lot of its speed advantage by cheating on quality, i.e., you can get very fast and artifact-free GI because it's blending the snot out of the samples. It'll take some time to figure out how fast it is when high quality settings are used.

Here's the teapot scene for anyone who wants to try it: http://www.3danvil.com/fR-2/teapot_fR-2.rar

It has the settings for my 2 minute fR render.

Thanks for the unbiased information. I can only tell that the real test will come from a scene with a lot of detail to see how it really stacks up. No solid forms, intricate detail like a heavily greebled mix of geometries.

So far, I am impressed with the quality of FR but I am also impressed by the improvement in lights from AR2.5 and C4D 9.5. I am for the first time doing a few images with just the ray tracer and some ambient occlusion which can look a lot like GI.

Keep the images coming for FR. We need a test scene with incredible detail for a good comparison.

JoelOtron
11-20-2005, 01:08 AM
IMO speed is unquestionably better. Quality is another question. It's way too early to give any conclusions, but it *looks* like fR gets a lot of its speed advantage by cheating on quality,

Whats quality?
If it looks awesome and your client likes it--thats quality enough for me. I know a lot of us split hairs on the minute details here--which is good, makes everyone aim higher.

But so far I am very happy with what I see from FR2. Maybe its not accurate based on ligth physics, but the reality is--actual light physics might not always create the most pleasing image. I'm very impressed with what I'm seeing.

DynamicRealism
11-20-2005, 01:38 AM
DynamicRealism Tweak 05 (C) lol...just kidding.. 6 minutes 23 seconds, w/ optimized gi


http://members.cox.net/dbxs/potversion.jpg

Continuumx
11-20-2005, 01:41 AM
Whats quality?
If it looks awesome and your client likes it--thats quality enough for me. I know a lot of us split hairs on the minute details here--which is good, makes everyone aim higher.

But so far I am very happy with what I see from FR2. Maybe its not accurate based on ligth physics, but the reality is--actual light physics might not always create the most pleasing image. I'm very impressed with what I'm seeing.

The discussion is different because on CGSociety, our peers are mostly artist and professionals and in this realm quality means a lot. I do agree, I am so far impressed by FR.

Per-Anders
11-20-2005, 02:05 AM
The quality to be honest is better in fR, you can use really low settings and get great results. It uses much lower settings with GI to get better results with far fewer artifacts in much less time than AR. And in my book is you can get as good quality or better with lower settings in less time then the quality is better.

Just on a side note, you guys should turn of "Store in Ambient Tree" in the portal, as that's loosing all of it's definition. That's only if you want things to be faster and defined shadows/lighting aren't such an issue. It's also worth noting that storing in the ambient tree often leads to you needing higher quality gi settings. And on top of that the scene has far too high AA settings, the min could be taken down to -3 and the max down to 1 and it will still look fine (and render a lot faster).

AdamT
11-20-2005, 03:33 PM
Just on a side note, you guys should turn of "Store in Ambient Tree" in the portal, as that's loosing all of it's definition.
Yes!! Thank you! Turning off "Store in Ambient Tree" improves the definition of GI shadows *dramatically*.

JDP
11-20-2005, 04:33 PM
DynamicRealism, any chance of posting the glasss shader you used. I'm not finding the manual too helpful when it comes to the shader tree, some parameters are completely omitted and there aren't very many examples. I'm getting some nice things to happen but it's a bit more of a struggle than it needs to be.

lllab
11-21-2005, 09:09 AM
oh yes please post your testscene:-)

cheers
stefan

Janine
11-21-2005, 12:44 PM
Hmm, so fR kicks ARs ass? :D

http://www.3dfluff.com/files/teapots.jpg

Does it? :D

Hehe, I know I know, the gi engine in fR is superior to AR, no question.

The biggest issue with Adam's pic is the lack of defined shadows. But I'm sure it's just a question of finding the right settings. Especially the skyportal should really help with getting nice shadows. So while I think that fR is definitely more advanced in the GI area than our AR, there's still some life left in the old AR. :buttrock:

lllab
11-21-2005, 12:48 PM
:-)

i' d love to see this Ar settings. is it psd'd a bit?

cheers
stefan

Janine
11-21-2005, 12:56 PM
:-)

i' d love to see this Ar settings. is it psd'd a bit?

Nope, straight out of the renderer, no cheating. :)

The GI settings:
Accuracy = 70%
Diffuse Depth = 2
Stoch. Samples = 450
Min = 25
Max = 250

And the area shadow settings:
Accuracy = 70%
Min = 15
Max = 150

lllab
11-21-2005, 01:00 PM
hmmm, not bad..wanna post that file:-)
where is the omni placed, in the middle of the room? the area light i guess is in the opening, or...?

would intrest me much.

cheers
stefan

Janine
11-21-2005, 01:05 PM
I might post the file later. The omni is just for the direct sunlight, its on the outside, with a hard shadow. I was too lazy to make it a parallel light. ;) The area light is in the opening.

vid2k2
11-21-2005, 01:16 PM
LOL, seems that I had to try a AR2 render to see what I was able to come up with.
Settings are very basic and very low. imac G5 / 2GB/ standard card
100
70
1/3
1
300
20
100

1 area light no shadows

AdamT
11-21-2005, 02:13 PM
All right, don't *make* me come over there!

This is with just the AMD X2 @ 2.6Ghz ... 1 minute 48 seconds. :)

AdamT
11-21-2005, 02:15 PM
Btw, Janine, that's a great AR2 result. I was testing without an area light, which on second thought was totally unfair (since I'm using a portal in fR-2).

vid2k2
11-21-2005, 02:30 PM
All right, don't *make* me come over there!

This is with just the AMD X2 @ 2.6Ghz ... 1 minute 48 seconds. :)

LOL, glad to see render times improving:) Nice to see the comparisons.

Janine
11-21-2005, 02:49 PM
All right, don't *make* me come over there!

This is with just the AMD X2 @ 2.6Ghz ... 1 minute 48 seconds. :)

Haha, ok ok. :D Looks good!

Continuumx
11-21-2005, 02:57 PM
Hmm, so fR kicks ARs ass? :D

Does it? :D

Hehe, I know I know, the gi engine in fR is superior to AR, no question.

The biggest issue with Adam's pic is the lack of defined shadows. But I'm sure it's just a question of finding the right settings. Especially the skyportal should really help with getting nice shadows. So while I think that fR is definitely more advanced in the GI area than our AR, there's still some life left in the old AR. :buttrock:

I like it a lot Janine! AR2.5 is no slacker.

Continuumx
11-21-2005, 02:59 PM
All right, don't *make* me come over there!

This is with just the AMD X2 @ 2.6Ghz ... 1 minute 48 seconds. :)

Nice AdamT!

JoelOtron
11-21-2005, 05:44 PM
Anyone tried the kitchen scene?

Also, it would be nice to see some renders of a room with lots of stuff in it --lots of variety of materials textures applied as well.

The sparse white rooms look great--but the average job will have more going on in it.
Mdme Sadies stone room was nice for that reason, refraction/reflection and detailed bumpy textures.

soccerrprp
11-21-2005, 06:24 PM
That does it for me! AR more than holds its own and at over $500 for FR, I'm happy with AR alone! Thanks folks for the AR/FR comparisons. It was extremely helpful!

dann_stubbs
11-21-2005, 06:45 PM
Anyone tried the kitchen scene?

Also, it would be nice to see some renders of a room with lots of stuff in it --lots of variety of materials textures applied as well.

The sparse white rooms look great--but the average job will have more going on in it.
Mdme Sadies stone room was nice for that reason, refraction/reflection and detailed bumpy textures.

joel,

check out the depth the FR settings here in this thread - i opened the kitchen scene but am not anywhere near to grips with all the settings of FR yet.

http://www.cebas.com/forums/cebas/viewtopic.php?t=3231&start=15

fluffy posted a bunch of the interface screen shots.

dann

AdamT
11-21-2005, 06:49 PM
Didn't mdme_sadie post an fR kitchen render a while back? Anyway, I'll give it a shot when I get a chance.

JoelOtron
11-21-2005, 08:22 PM
@Dann--thanks--will check out.

Adam--cool--will keep my eyes open. I'll check the kitchen thread--I probably wasnt paying attention to anything other than AR in that thread before now.

Janine
11-21-2005, 08:27 PM
That does it for me! AR more than holds its own and at over $500 for FR, I'm happy with AR alone! Thanks folks for the AR/FR comparisons. It was extremely helpful!

Oh crap, hope Edwin doesn't hate me now! :p No seriously, fR is great! Maybe if I find the time I'll try some stuff with it, if they don't kick me off the betateam. ;)

mustardseed
11-22-2005, 01:53 PM
I never could resist having a go at a ready-modeled scene, so for the fun of it here's my contribution.

Rendertime 20mins 28s. 1 area light, lots of omnis. C4d 9.1 core(no AR) on an iBook G4 800mhz (with mail and safari running!).

andrew

maikukai
11-23-2005, 08:16 PM
I must be doing something wrong. I tried Janie's settings on my 2.4 Dual Xeon here at work. After 10 minutes it's only about 1/4 done with the pre-pass. Janie, how'd you render that in 5 min.?

DynamicRealism
11-23-2005, 10:16 PM
i smell the crap of a bull <sniffs around>

AdamT
11-23-2005, 10:34 PM
I must be doing something wrong. I tried Janie's settings on my 2.4 Dual Xeon here at work. After 10 minutes it's only about 1/4 done with the pre-pass. Janie, how'd you render that in 5 min.?
Janine used an area light in the opening and an omni below it, which should speed things up considerably.

BeccoUK
11-24-2005, 09:43 AM
AdamT: Thank you for posting the teapot scene. I gave this a go in Maxwell Render (Beta) and this it what got churned out:

http://digital-depictions.com/maxwell/teapots/teapots01.jpg

Obviously, it took longer than Maxon's Advanced Render and Final Render. Knowing it would be longer I gave the teapots a Maxwell Metal with a low value roughness.

The lighting is a 115 watt emitter used on a single polygon, sized the same as AdamT's scene light.

This Maxwell version has better defined shadows. Final Render seems to oversmooth - is this how it gets those fast times?

Anyway this passed a bit of time and I got to know Maxwell that little bit better.

STRAT
11-24-2005, 09:48 AM
AdamT:

This Maxwell version has better defined shadows.

nice, how long render time out of interest? yup, maxwell got nice shadows, but it's very 'wisshy-washy' and pretty undefined in the room corners where the walls/floor/ceiling meet.

BeccoUK
11-24-2005, 10:00 AM
Strat: I started this rendering yesterday afternoon and stopped it just before posting the image. In that time It reached just under sample level 17.

I suppose with...
1. Another day or so of rendering the corners would have been more defined.
2. More experience with Maxwell on my part.

Whenever MaxwellRC gets released I may re-render this teapot scene.

vesalus
11-24-2005, 10:01 AM
comparing those engines with different settings cannot equal comparing a phong rendering and a radiosity one, its not that simple, they all got their pros and con...

Janine
11-24-2005, 11:01 AM
Janine used an area light in the opening and an omni below it, which should speed things up considerably.

Wrong. :p I didn't. The omni is on the outside. There's an area light in the opening. And a sky object in addition to all that with some blue but that doesn't have much influence on the rendertime. The area light has a colored falloff btw. Right, if people don't believe me, download my scene file. :) click link (http://www.3dfluff.com/files/janineteapot-final.rar)

Chrissyboy
11-24-2005, 12:05 PM
Hi Janine - thanks for providing this file, very interesting. Quick question; the sky with the blue set at 300% brightness, is this just to provide a bit of colour on the secondary bounces? I've tried changing the colour to red, green etc, it does change the scene but the effect is very subtle.

The reason I'm asking is that I've always been advised against using Sky objects in GI scenes, (as opposed to large spheres) as they set an infinite scale for the whole scene. Is this just nonsense?

Cheers for the file - Chris

Wrong. :p I didn't. The omni is on the outside. There's an area light in the opening. And a sky object in addition to all that with some blue but that doesn't have much influence on the rendertime. The area light has a colored falloff btw. Right, if people don't believe me, download my scene file. :) click link (http://www.3dfluff.com/files/janineteapot-final.rar)

lllab
11-24-2005, 12:24 PM
as far as i know a sky object is not infinite, it adjusts itself to scenesize.
i always use skyobjects in GI- fast and no problems.

this is a rumours from very old times, maybe it was true for older versions.

cheers
stefan

Chrissyboy
11-24-2005, 12:29 PM
Ah - that makes a whole lot of sense.
Cheers Stefan - Chris

as far as i know a sky object is not infinite, it adjusts itself to scenesize.
i always use skyobjects in GI- fast and no problems.

this is a rumours from very old times, maybe it was true for older versions.

cheers
stefan

Janine
11-24-2005, 12:57 PM
Hi Janine - thanks for providing this file, very interesting. Quick question; the sky with the blue set at 300% brightness, is this just to provide a bit of colour on the secondary bounces? I've tried changing the colour to red, green etc, it does change the scene but the effect is very subtle.

The reason I'm asking is that I've always been advised against using Sky objects in GI scenes, (as opposed to large spheres) as they set an infinite scale for the whole scene. Is this just nonsense?

Cheers for the file - Chris

It's nonsense. ;) The sky doesnt have geometry, it doesnt influence the scene scale. But a sphere does! Always use a sky, never use a sphere. But you're right, the sky in my case only has a subtle effect. Most of the light is coming from the area light. The area light is also where the nice shadows come from.

AdamT
11-24-2005, 03:36 PM
Hi Becco_UK,

You're right about the image I posted earlier--very soft shadows. That was because I used the "Store in Ambient Tree" parameter in the sky portal. If I uncheck that the shadows are much much better:

http://www.3danvil.com/fR-2/fR_2m26s_X2.jpg

This was rendered with one computer only--the AMD X2.

vid2k2
11-24-2005, 04:51 PM
The addition of the sky makes a big difference:)
Thanks for your reference file Janine.

My last render on this subject ~~~~~~~~~~~~

one computer: iMac
one test file (thanks for loaning it out AT)
one happy AR2.5 user
Time: 4:28 :thumbsup:

FantaBurky
11-25-2005, 10:02 PM
Hi Becco_UK,

You're right about the image I posted earlier--very soft shadows. That was because I used the "Store in Ambient Tree" parameter in the sky portal. If I uncheck that the shadows are much much better:

http://www.3danvil.com/fR-2/fR_2m26s_X2.jpg

This was rendered with one computer only--the AMD X2.

finalRender seems way more stable then Maxwell :O. Is it a beta version thats out? Is it really that advanced (when talking about the bridging to C4D)? I mean, aint there many bugs/glitches? That render seems fantastic though I dunno enough to decide if its worth it or not.

Per-Anders
11-25-2005, 10:28 PM
Final Render is not a beta, it's the finished product. Yes it's integration is top notch with Cinema, it's more or less seamless. Yes there are a few bugs (just as with any peice of software), but very few, and nothing that will stop you from using this in a production environment (although there do appear to be a few problems with C4D r9.52 for a few users), it's a finished product after all.

FantaBurky
11-25-2005, 10:57 PM
Wow, thanks for the wonderful answer mdme. That was exactly the answer I was looking for. Seems I need to put away some money now, cause I've alwase wanted a different render then AR, just to see the differences and hopefully the possibilites. Wasnt sure about Maxwell since it isnt the finished version, the renders from that software stuns me. Though the latest render from Adam (in the AR VS the rest thread) is more my style. Like Pablo Fransisco says: Clean, soft and smooth, so "faaahaaiiiyen" :curious: . I'm sure Maxwell can achieve the same, but if finalRender is the real deal then its a must have for my self.

dann_stubbs
11-26-2005, 12:55 AM
Wow, thanks for the wonderful answer mdme. That was exactly the answer I was looking for. Seems I need to put away some money now, cause I've alwase wanted a different render then AR, just to see the differences and hopefully the possibilites. Wasnt sure about Maxwell since it isnt the finished version, the renders from that software stuns me. Though the latest render from Adam (in the AR VS the rest thread) is more my style. Like Pablo Fransisco says: Clean, soft and smooth, so "faaahaaiiiyen" :curious: . I'm sure Maxwell can achieve the same, but if finalRender is the real deal then its a must have for my self.

http://forums.cgsociety.org/showthread.php?t=295854&page=1&pp=15

have a look at the depth of the controls - i'm really liking FR2 a lot. so far very stable and i've been cranking the bounces and caustics up on a couple scenes and they look very maxwell but in just an hour or two... lots more testing and learning to come...

(interface shots are a couple posts into that thread)

dann

Opelfruits
11-26-2005, 07:24 AM
if your looking for a simple render engine then i wouldnt get final render, but if you have the time to learn how to use it then go for it.

mustardseed
12-06-2005, 04:25 AM
Very sorry to revive this, and apologies if it's terribly OT, but I just rerendered the scene again on my Quad and it tore through it in 59secs. Again, this is C4d core without AR.

andrew

andronikos916
12-20-2005, 06:32 AM
AdamT ..what threa you just started again? Did you realize that it was exactly 1 year ago whe I started that teapot GI with maxwell.

haha.

Merry Xmas to all! ...maxwell is still under heavy development so I will miss this years crazy party!

cy,
Andronikos

By the way very fast renderings with FrSt2.
I have in my office PC most render results with that sen using all dif. render engines. I will try to upload this week to see if we actually have a speed increase or not?

bye!
:thumbsup:

Ernest Burden
12-20-2005, 02:14 PM
Did you realize that it was exactly 1 year ago whe I started that teapot GI with maxwell.

How's the noise level, it is clearing up yet?

AdamT
12-20-2005, 02:19 PM
How's the noise level, it is clearing up yet?
Zing! Good one. :)

CGTalk Moderation
12-20-2005, 02:19 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.