PDA

View Full Version : GI Render Settings


STRAT
08-08-2005, 10:10 AM
Right fellers, Ernest B was thinking about this thread i think, but i thought i'd post it anyway.

GI Render Settings. What about them? i though we'd have an informative learning thread which will hopefully give us all a better insight and use out of our renderer.

Im a proffessional architectural visualiser and switched to C4D as my main render weapon from max about 3 years or so ago, and have always been lucky to have dual xeons provided for me as my render processors, so maybe i've picked up more bad habits about the understanding of what i'm actually doing than i should.

out of interest, here's a quick example of a wip of the typical kind of thing i work on. this'll give you an idea of the general complexity of a typical stand-alone model (sorry i'm not at liberty to provide test models) -

http://www.nikclark.com/strat/rh.jpg

This particular scene has about 50 internal lights (w/shads) and a few external. Accuracy set to 100%, stoches = 100, min=100 and max=100, with a dd of 1 and a prepass size of 1/1. it rendered in 4 hours at a res of 2500 pixels wide on a dual xeon. (not too bad imo)

I've always gone for as high an accuracy setting as i can, with lower samples. i've always found this to be very visually pleasing as possible (in my case anyway)

I'll typically use an accuracy set anywhere around the region of 95% +, with a stoch sample rate at around 100. my min samples usually range around the 30-50 mark, and the max samples are to suit the model, but again, usually around the 100-200 mark.

(I ALWAYS set a diffuse depth to 1. But this is personal choice - i prefer to keep a nice render speed and can successfully use infill omnis to compensate for lack of bounces, and i'll use odd AA settings to for optomisation. But lets not argue about this, as it almost warrents it's own thread.)

I've always found the combination of accuracy, min/max and stoches easy to use and set up, but not always understandable. particulaly the min settings.

to me, min settings are relevant for larger sparce un-interrupted areas of a mesh and max settings for intersections and complicated bits. Accuracy and general stoch samples for more of the realism side of things.
i.e. i'll set a general min/max settings to get a non-blotchy image, completely indipendant of what i might set the accuracy and stoch samples.

but of course, they all work together too. but how can adding more min/max samples up the realism? all they seem to do is iron out the fiddly splotches. it's the accuracy and stoch samples that determin the realism i thought.

why not use an accuracy over 90% as some peeps recently in here suggest you should never do? whats the point of the option to do it then? i use it to good effect :shrug:

i've been testing yet again with a lower accuracy and higher min/max/general settings and am getting nice results, but not as real as when i slap the accuracy up. also, the render time obviously shoots up the more samples i use.

so what do you think. what would you suggest for that model above there, with in view for realism of lighting.

the more understanding we can get the better.

cheers for your time. hope i havent drivelled on too much :p

mmhnemo
08-08-2005, 02:03 PM
Well on my side, during the various stages of testrenders, i gradually lower the min and max values up to the point where artifacts/splotches get visible - then up it again a little and have my final rendersetting.
Also note a value of 100 for min seems alot. I rarely end up with a min value around 30 - mostly its under that.

About the same procedure with stoch samples.

To Speed up Testrenders i reduce accuracy and the preview size to rediculously low numbers.
I usually end up with an accuracy between 80-90% for the final render as anything above that slows down the render tremendously.

I don't use many fill lights to enhance an image since it takes a good deal of work to match these against the overall GI. If they dont match the GI perfectly the whole image suffers and that is not tolerable.

Also the nescessary tweaks to the materials to get good reflections/surface properties play a major role for the time a render takes to finish so if i can get away without such calculation heavy materials it is very fortunate. However since i mostly do closeup shots these materials are a nescessity and add alot to rendertimes.

With the new release (better area lights/shadows/blurry reflections) i believe we will have lots of new options to discuss. In the meantime this is my take on GI rendering.

regards

nycL45
08-08-2005, 02:13 PM
Strat, thanks for taking the time to keep this discussion going and very nice work. A couple of questions:
1. Is your setting accuracy to 95-100% a part of a tried and true production formula or is there a perceptible quality improvement that others can see if pointed out? If it is perceptible, could you use this rendering and provide an enlarged and rendered small detail with accuracy set at 70%, 85% and 100% (assuming the other render settings would have to be adjusted accordingly)? Me, trying to be scientific. lol

2. What is your opinion about MJV's accuracy and c tag head room comment:"You should always have it set to 70% in order to leave yourself some head room for changing the value in the compositing tags."?

(BTW, thanks for the valuable help you've given me in the past.)

Edited 2.

mmhnemo
08-08-2005, 02:19 PM
I just noticed you used a diffuse depth of 1

Next you likely put over 50 lights behind your windows to illuminate the rooms right?

And an accuracy of 100?

-------

How about a dd of 3, no extra lights, adjusted min/max values and a reduced accuracy of 85% - would you care to try out this configuration and let us see the outcome (time/quality)?

Of course for such a discussion it would be beneficial for all participants to have access to the test scene but since you can't hand that out i'd like to see some of the suggestions that might come up incorporated in testrenders from you - if you have the time of course.

AdamT
08-08-2005, 03:14 PM
I'm thinking this would make a nice mini challenge. Someone would have to supply a scene though. I guess everyone would submit their best settings to optimize speed vs. quality. The downside would be that in order to judge someone would have to test all the settings on one machine. Or maybe there would be some way to standardize render times by figuring in the individual entrants' CB scores?

Ernest Burden
08-08-2005, 03:23 PM
Or maybe there would be some way to standardize render times by figuring in the individual entrants' CB scores?

That sounds like a good idea. A CB score as mulitiplier.

There is always the scene I made for the glass tests. Not as exciting an example of arch-vis as STRATs sample, but the model is available and has an exterior and interior component. Just a thought.

AdamT
08-08-2005, 03:45 PM
That sounds like a good idea. A CB score as mulitiplier.

There is always the scene I made for the glass tests. Not as exciting an example of arch-vis as STRATs sample, but the model is available and has an exterior and interior component. Just a thought.

I guess the CB would a divisor, but I think it would work okay. :) It wouldn't be 100% accurate because CB doesn't cover every situation, but I think it would be good enough for such an informal contest.

I think I still have that scene you posted--will have to check. I'd suggest using only two tex in the scene: default and glass. Default is good for showing artifacts and glass is good for reflection and interior lighting issues. Of course everyone should use the same glass material.

Ernest Burden
08-08-2005, 03:49 PM
I think I still have that scene you posted--will have to check. I'd suggest using only two tex in the scene: default and glass. Default is good for showing artifacts and glass is good for reflection and interior lighting issues. Of course everyone should use the same glass material.

That scene was one of the first things I made in Cinema, so if it would need any work (joining surfaces, etc) I would not be insulted. If you or anyone wants to use it that's fine, since testing is why I created/posted it. It's still a pretty boring little street, unfortunately.

vid2k2
08-08-2005, 04:53 PM
STRAT and all, I'd be wiling to model a simplified version of your scene
and, or you could provide one with just some primitives stacked up and
selection sets to represent the glass.

I feel that having basic lighting and camera angle matching the test scene
to be critical as well.

As per other comments: we're just adjusting settings on our own machines
to test time and variations of render quality.

As a side note: When getting mildew splotches in an interior scene I found that
if I cancelled vol lights, the splotches went away. However, additional fills had
to be added to make up the difference.

STRAT
08-08-2005, 05:12 PM
damn! a busy day. havent had much time to check this thread.

but, an easy day planned tomorrow, so i'll do some test renders and details as requested and post up results.

cheers dudes, keep it comming...

STRAT
08-10-2005, 08:14 AM
dudes, i spent all day yesterday rendering and testing on a live job.

and yes, a lower accuracy and higher samples is a nice combination. faster too. thanks for the advice fellers.

old and usual settings -

strength - 100%
accuracy - 100%
prepass - 1/1
diff depth - 1
stotches - 100
min - 30
max - 100

render time over 4 hours.



new settings -

i'm generally getting lovely realism, slotchyless too, with settings like -

strength - 100% (i'll always use a 100% value)
Accuracy - 80%
prepass - 1/1 (depending on scene type)
diff depth - 1 (again, i always use 1 for fast speeds. i've a good infill light backup to compensate)
Stotch samps - 1000
min - 300
max - 100

the image at the front of the post now renders cleaner, just as realistic, but in 1 hour instead of 4.

here's the new pic, a touch of post photoshop, and a lense flair. yes yes i know, but i thought it worked nice in this particular instance -

http://www.nikclark.com/strat/cam04.jpg

jddog
08-10-2005, 09:59 AM
here's the new pic, a touch of post photoshop, and a lense flair. yes yes i know, but i thought it worked nice in this particular instance -

ok ok, but the kind of lens flare that you choose don't match the kind of camera that you have. The visible ring it must be more trasparent but more visible the white circle on the centere where the window is. Please consider this comment als a personal feeling.

Very interesting thread and I always was suprise about your settings for the renderings... excellent results.

jdd

MPS
08-10-2005, 10:50 AM
Thats a really big step down rendertime wise Strat. The quality looks great. I like the way you've merged the photo in too. It looks almost seamless.


Just wondering what settings everyone uses for stochastic mode radiosity?

JamesMK
08-10-2005, 11:39 AM
Gee, I don't know - using GI at all is soooo five minutes ago *limp wrist*

:D

Seriously though, I like the challenge idea, and using CB score as divisor should be accurate enough me thinks :thumbsup:




.

STRAT
08-10-2005, 11:57 AM
whats CB score mean?

JamesMK
08-10-2005, 12:17 PM
whats CB score mean?
The rendering speed mark you get when running CineBench.




.

seco7
08-10-2005, 01:14 PM
Nice job Strat ... I would still be curious to see how dropping the 50 internal lights and increasing the number of bounces effects the quality and render times. Just in case you had a few free hours ... lol.

Ernest Burden
08-10-2005, 01:28 PM
whats CB score mean?

Somebody not had his coffee yet?

CineBench, as a way of equalizing time results on render tests to a set resolution.

I've not run CineBench since before I even owned Cinema, but its a great idea.

STRAT
08-10-2005, 01:39 PM
Nice job Strat ... I would still be curious to see how dropping the 50 internal lights and increasing the number of bounces effects the quality and render times. Just in case you had a few free hours ... lol.

that render hasn't actually got the 50 internals switched on :thumbsup:

and nope, at that high a realism setting and resolution there's no way i'm adding an extra bounce. you can garantee the render time will massively increase. the difference in render times between 1 and 2 bounces really is significant.

besides, i'm more than happy with adding infill lighting to compensate for extra dd. just my choice :)

dann_stubbs
08-10-2005, 01:45 PM
dudes, i spent all day yesterday rendering and testing on a live job.

and yes, a lower accuracy and higher samples is a nice combination. faster too. thanks for the advice fellers.

old and usual settings -

strength - 100%
accuracy - 100%
prepass - 1/1
diff depth - 1
stotches - 100
min - 30
max - 100

render time over 4 hours.

new settings -

i'm generally getting lovely realism, slotchyless too, with settings like -

strength - 100% (i'll always use a 100% value)
Accuracy - 80%
prepass - 1/1 (depending on scene type)
diff depth - 1 (again, i always use 1 for fast speeds. i've a good infill light backup to compensate)
Stotch samps - 1000
min - 300
max - 100

the image at the front of the post now renders cleaner, just as realistic, but in 1 hour instead of 4.


i like the new render better too - it is good to see that c4d DOES have the tools necessary to do a nice render at a good speed - it is just the settings have to be adjusted to get there

a 75% render time reduction is certainly something to justify a bit of time doing render tests... update us as you see if these "revised" settings benefit your other projects.

good to know, and good that you posted so others can see "it can be done" : ) (not always maxon to blame)

dann

AdamT
08-10-2005, 01:48 PM
I'd suggest using a modest luminance value for your interior wall material (10-20%) rather than multiple lights or increased diffuse depth--at least for exterior shots.

STRAT
08-10-2005, 02:00 PM
yup, i'll certainly be using this new approach from now on. i'll keep you advised. perhaps the general 75% time savings will be enough for me to consider extra light bounces for my deadlines.

I'd suggest using a modest luminance value for your interior wall material (10-20%) rather than multiple lights or increased diffuse depth--at least for exterior shots.

well, another method i'll use is to assign the internal walls their own material, then make that material generate a high gi count in the illumination tab. something like 300% or something.

same for the glass, sometimes i'll stop all gi being recieved, but up the generate figure by a few 100%. works really nicely.

Jorge Arango
08-10-2005, 02:18 PM
new settings -


min - 300
max - 100



You have the min and max values swapped, right?


Jorge Arango

STRAT
08-10-2005, 02:24 PM
no, i havent. min=300 max=100

and i'll tell you why -

traditionally with my ultra high accuracy settings i'd find a min setting of around 20-30 usually was the fastest solution.

but lowering the accuracy to 70% or 80% and keeping the min at 30 gave me horrible un-realistic shading, even though i'd no put my stoches up to 1000.

so the higher i made the min samples the finer and more realistic lloking the shading samples on screen got. so the 'realism' is controlled principly now by the min setting rather than pissinf with the stoch samples (as long as the stoch setting is sufficiently high in the first place).

(i tried keeping the min settings down at the normal 30 mark i used and putting the stoch samples up to 5000 or something, but the render times shot through the window, and the realism just wasn't there either)

ok, the higher the min settings the higher the render times, but the higher the mins the better realism shading i was achieving, and i was saving 75% render times from my old renders even with this.

and the max settings - i set that depending on my mesh. 100 was ample in this building.

Jorge Arango
08-10-2005, 02:32 PM
no, i havent. min=300 max=100

and i'll tell you why -

traditionally with my ultra high accuracy settings i'd find a min setting of around 20-30 usually was the fastest solution.

but lowering the accuracy to 70% or 80% and keeping the min at 30 gave me horrible un-realistic shading, even though i'd no put my stoches up to 1000.

so the higher i made the min samples the finer and more realistic lloking the shading samples on screen got. so the 'realism' is controlled principly now by the min setting rather than pissinf with the stoch samples (as long as the stoch setting is sufficiently high in the first place).

(i tried keeping the min settings down at the normal 30 mark i used and putting the stoch samples up to 5000 or something, but the render times shot through the window, and the realism just wasn't there either)

ok, the higher the min settings the higher the render times, but the higher the mins the better realism shading i was achieving, and i was saving 75% render times from my old renders even with this.

and the max settings - i set that depending on my mesh. 100 was ample in this building.

Thank you for clearing this up Strat. You never stop learning.


Jorge Arango

AdamT
08-10-2005, 02:34 PM
If your Min is higher than your Max, then Max has no effect at all. I'd try swapping the two and see how that works.

Ernest Burden
08-10-2005, 02:46 PM
I'm still not understanding the relationships between the settings. I've been testing these this morning, but since I was already rendering at 78% accuracy then I'm not seeing much difference as I play with the other settings. Trying, trying...

STRAT
08-10-2005, 02:51 PM
If your Min is higher than your Max, then Max has no effect at all. I'd try swapping the two and see how that works.

correct. (obvious really)

my min set at 300 was doing the job of both (uncorrectly though)

i've now set the min back to the model's optimum, about 80 in this case, and the max to 300.
it's rendering with the same high quality, but much faster now. much faster.

Ernest Burden
08-10-2005, 02:57 PM
What's interesting is that while testing I am getting rendertimes on a test frame varying from 4:27 (my original settings) to 5:42, but the image isn't changing much at all. There is a GI artifact that remains unchanged regardless.

STRAT
08-10-2005, 02:59 PM
What's interesting is that while testing I am getting rendertimes on a test frame varying from 4:27 (my original settings) to 5:42, but the image isn't changing much at all. There is a GI artifact that remains unchanged regardless.

can you remind us of your gi settings again m8?

Ernest Burden
08-10-2005, 03:17 PM
This is for a 720x540 animation, AA = none

I have been working from:
camera anim. = recompute always
accuracy = 78%
pre-pass = none (though I may have to use pp, but it ups times a lot)
dd = 1
stoc. samp. = 360
min = 12
max = 240

This yeilds 4:27 on a particular frame.


Going to stoc/min/max of 720/120/240 yeilds 5:42 but the artifact is exactly the same. (And I can get away with the artifact, so I'm not worried, just pointing out the quality issue).

Now I'm seeing if prepass gets a better result...

STRAT
08-10-2005, 03:28 PM
the pre-pass will certainly help minimise your artifacts.

try upping just the min to 20 or 30 or something. see what that does.

MJV
08-10-2005, 04:11 PM
and the max settings - i set that depending on my mesh. 100 was ample in this building.

Max setting is irrelevant when lower than min. 1, 20,100,200. All will give the same results here. In other words, if you minimum setting is higher than your max, it means you're adding no extra details at corners, which is perfectly ok and often the proper thing to do, especially if you need realistic shadows throughout.

Nice render!

Ernest Burden
08-10-2005, 05:01 PM
the pre-pass will certainly help minimise your artifacts. try upping just the min to 20 or 30 or something. see what that does.

I had tried many combos, upping the min did not help, just raised the rendertime a bit.

Trying pre-pass options was interesting. None of them changed the look at all (except 1/3). So my original setting were still fast, but went from 4:27 with no prepass to:
1/4 = 5:26
1/2 = 7:07
1/1 = 11:36
all looking exactly the same

Ah, but 1/3 = 5:09 and with much improvement in artifacts. I tried it on several other frames in the piece and they, too, visibly improved without too much added time. I'm goin' with it.

STRAT
08-11-2005, 09:10 AM
definately having better results, having fun now :) . and of course, night time rendering needs even less samples.

http://www.nikclark.com/strat/test.jpg

nycL45
08-11-2005, 01:11 PM
See the light and it will set you free! :bounce:

jondoe0ne
08-11-2005, 01:27 PM
here is a nice radiosity tut that i have found over the web and i find it very helpful in undestanding it's settings:

http://mvpny.com/RadTutMV/RadiosityTut1MV.html

STRAT
08-11-2005, 01:35 PM
here is a nice radiosity tut that i have found over the web and i find it very helpful in undestanding it's settings:

http://mvpny.com/RadTutMV/RadiosityTut1MV.html

yup, i think everyone knows about that.

but actually understanding relationships between settings is something else.

Ernest Burden
08-11-2005, 01:54 PM
but actually understanding relationships between settings is something else.

Indeed. Stephen, did you test my prepass = 1/3 results? I would like to know if that seems 'best' to you, as well.

artemesia66
08-11-2005, 03:04 PM
min = 12
max = 240


Did you try upping the min here? I know strat suggested it. General wisdom seems to be that the min:max ratiio should be btwn 1:5 to 1:10, or cinema spends more time interpolating than it would just casting extra samples. So I'm curious to see if that's true in your scene.

STRAT
08-11-2005, 03:30 PM
Indeed. Stephen, did you test my prepass = 1/3 results? I would like to know if that seems 'best' to you, as well.

i'll generally go with a 1/2 prepass. my final imagery is usually at a res of 2000x1500 pixels, and they're day time shots too, so a 1/3 or a 1/2 prepass suits me. anything higher res then i go for a 1/1.

i find that larger open faces or planes with minimal intersections or corners tend to show up artifacts quite noticably at a lower prepass more often than not, so again, depending on the architecture of the model i'd say a 1/2 or even 1/1 prepass.

it's a strange set of artifacts, sometimes it's almost as though the gi looks 'mucky' or 'weathered'. upping the prepass to max will iron it out.

but for your work i'd say a 1/3 would be ample.

Ernest Burden
08-11-2005, 03:32 PM
Did you try upping the min here? I know strat suggested it. General wisdom seems to be that the min:max ratiio should be btwn 1:5 to 1:10, or cinema spends more time interpolating than it would just casting extra samples. So I'm curious to see if that's true in your scene.

Yes. When my min = 12 I got 4:27, min = 24 got 4:54 but no visible improvement is the render. But I could go with the 1:10 ratio (24/240) if there is good reason for it.

But it was pretty much linear, the higher the min, the higher the rendertime.

Ernest Burden
08-11-2005, 03:36 PM
2000x1500 pixels, and they're day time shots too, so a 1/3 or a 1/2 prepass suits me. anything higher res then i go for a 1/1.

but for your work i'd say a 1/3 would be ample.

What I am interested in here is if there is anything 'special' about 1/3. As I said, I had this artifact that would only go away with stochastic mode, otherwise, regardless of min/max/prepass it was there. EXCEPT for prepass 1/3 with which it was gone.

Does this hold true for others, or could it just be something about my file?

AdamT
08-11-2005, 03:37 PM
Instead of increasing the min setting you might try lowering your max to bring it within the usual ratio, e.g., 12/120 - 24/240.

STRAT
08-11-2005, 03:43 PM
Did you try upping the min here? I know strat suggested it. General wisdom seems to be that the min:max ratiio should be btwn 1:5 to 1:10, or cinema spends more time interpolating than it would just casting extra samples. So I'm curious to see if that's true in your scene.

why these settings? i been test rendering with and without those ratios and these ratios dont give me any advantagious relative speed or quality increases.

jondoe0ne
08-11-2005, 04:04 PM
all these settings are still puzzleing for me.... allthough i have got good results with radiosity i still think it's all about luck... i have tested it with 2 different types of sample settings: the first one is with min>max an rendered in 26 mins and the second one is with min<max and it is rendered in 52 mins... look at them and tell me what u think... my oppinion is that both are good but the one with min>max is somehow odd cuz the shadows made by radiosity are wider and more diffuse than in the other one

render settings
strenght: 100%
accuracy: 85%
prepass size:1/1
diffuse depth:4
stochastic samples:500
min resolution-in the picture's name
max resolution-in the picture's name

AdamT
08-11-2005, 04:18 PM
... look at them and tell me what u think... my oppinion is that both are good but the one with min>max is somehow odd cuz the shadows made by radiosity are wider and more diffuse than in the other one
Both have huge GI artifacts, but the Min<Max render is much better in that regard. You'll need much higher settings to get an artifact-free render. I'd lower the diffuse depth to 1 or 2 and increase samples to around a 1000-1500. Then adjust Min/Max.

artemesia66
08-11-2005, 04:29 PM
why these settings? i been test rendering with and without those ratios and these ratios dont give me any advantagious relative speed or quality increases.

that's what i was wondering. I've read it a couple of times, and not really been able to verify it, so i wondered if you guys who do more GI renders than i do had any input.

jondoe0ne
08-11-2005, 04:30 PM
Both have huge GI artifacts

i can't understand.... where do u see those HUGE artifacts....???

AdamT
08-11-2005, 04:50 PM
You can see that your sample dots appear to be around 1-2' in circumference. They need to be much much smaller and much more dense in order to blend properly. After a adjusting levels to bring them out:

STRAT
08-11-2005, 04:56 PM
you missed......

http://www.nikclark.com/strat/room.jpg



:p

AdamT
08-11-2005, 04:57 PM
I got the chair, but I was just pointing out the most obvious examples. :)

Ernest Burden
08-11-2005, 05:17 PM
By the way, here is an overview of what I'm animating. I never actually get above the rooflines, so the roof surfaces don't need to be correct, and you don't see the X-trees as such. Notice how some of my cars and trees are SPD objects? That part is fun, especially the trees.

My camera sees everything from wide shots to close-in details of this scene. Some cars drive around.

http://www.oreally.com/temp/UHovervw.jpg

STRAT
08-11-2005, 07:45 PM
looking cool m8 so far. can you show us a render of the spd cars and trees a little closer please, and maybe ellaborate on their creation methods a tad more maybe?

Ernest Burden
08-11-2005, 09:55 PM
can you show us a render of the spd cars and trees a little closer please, and maybe ellaborate on their creation methods a tad more maybe?

Sure. But it's getting off the topic, what the heck...

The scene is very large as a ground surface with one big imagemap, as seen in the post above. The SPD is done with a copy of the ground set down about 6" below the street. Then a map, again fullsize at about 5000x3000, has blobs for cars and trees, then a SPD map with an alpha to soften out the edges, also to add some transparency to the SPD trees. All of these are for far background and 'just passing by the edge of the view' sort of use. I don't have to bother with placing all this stuff as objects (including setting the intersection heights), just draw shapes over an aerial photo to locate cars and trees, even buildings if I like. The tallest tree part would be white, set to 'intensity' and 38 feet tall, so some math tells me what shade of gray cars need to be.

I need to do a better job of getting rounded blobs for trees in Photoshop--they keep coming out as cones. So up close, fairly ugly, but as a background they work fine:

http://www.oreally.com/temp/UH-SPD-1.jpg

http://www.oreally.com/temp/UH-SPD-2.jpg

http://www.oreally.com/temp/UH-SPD-3.jpg

http://www.oreally.com/temp/UH-SPD-maps.jpg

AdamT
08-11-2005, 10:15 PM
Nice idea, but I wonder if the poly count of the spd ground might be puttin' the smack down on your render times.

Ernest Burden
08-11-2005, 11:25 PM
I wonder if the poly count of the spd ground might be puttin' the smack down on your render times.

Uh.... I think you have a point.

However, it was easier than putting in the time to place all that stuff, which would carry its own polygon overhead.

jondoe0ne
08-12-2005, 08:56 AM
You can see that your sample dots appear to be around 1-2' in circumference. They need to be much much smaller and much more dense in order to blend properly. After a adjusting levels to bring them out:

well u are talking about the picture with min>max... i think it's normal to appear artifacts in this one don't u???
never the less i think that c4d's AR is not able to get a picture in less time that is both realistic and spotless...

STRAT
08-12-2005, 09:03 AM
the second images min/max settings are obviously allot nicerer, but what's a massive drain on your render times is that diffuse depth of 4. why not try it on 1 or 2 with some infills?

i personally think adding a high dd is an easy lighting cop-out allot of the time. but look at the massive render times you must endure. you really can get an identical image with a low dd and infill lights and a much much lower render time.

jondoe0ne
08-12-2005, 09:18 AM
in this scene i don't have any interior lights... i would have liked it to be as natural as i could get it with less work as i have many projects and i don't have time to fiddle around with every one of them so the infill lights right now don't seem to be an option... as far i think that a easy set up would be with a 20% yellow envionment obj, a sun coming thru the window and i have also put outside teh room a hdri image... now i am trying what adam sugested i should do as in lowering the dd and upping the samples... in around half an hour i'll tell the difference...

jondoe0ne
08-12-2005, 10:23 AM
well it over an hour since it started rendering with adam's suggestions and here is the result

STRAT
08-12-2005, 10:55 AM
well, your other images obviously look nicer, but i wouldn't take Adam's settings litterally.

i personally use settings closer to Adams suggestions in my own work rather than yours, and i can produce some excellent imagery out of c4d with them. but as you've described your working process in the above posts, i'll obviously take allot longs with these settings fiddling about to make them work.

as i say, dont take Adams settings litterally, they're a guide line. c4d WILL produce beautiful imagery with these settings, with appreciatable render times/gi quality, but it needs the experience of the end user to manipulate them correctly.

i'm currently testing an interior scene right now as we speak.

jondoe0ne
08-12-2005, 11:45 AM
i have tried also something in between with dd 3 and ss750... and the result is this
render time 1h 09m 59s

STRAT
08-12-2005, 12:26 PM
i got another gi render question -

in the prepass, whats the difference between the red dots and the green dots? i though one was for min samples and one for max, but it isn't.

jondoe0ne
08-12-2005, 12:26 PM
but to my surprise this one which has a dd=5 and ss=350 rendered in just 36 mins compared to the ones before... now i am really puzzeled... why are you guys telling me to lower the diffuse depth and increase the stochastic samples when in fact increasing the dd and lowering the ss gave much better results and also faster ones...

jondoe0ne
08-12-2005, 12:28 PM
the red dots are the new gi information and the green dots are for already calculated information and they appear mostly in camera animation renderings...

STRAT
08-12-2005, 12:30 PM
the red dots are the new gi information and the green dots are for already calculated information and they appear mostly in camera animation renderings...

well i'm rendering fresh stills with brand new gi solutions each time and getting a mixture of both :/

STRAT
08-12-2005, 12:37 PM
but to my surprise this one which has a dd=5 and ss=350 rendered in just 36 mins compared to the ones before... now i am really puzzeled... why are you guys telling me to lower the diffuse depth and increase the stochastic samples when in fact increasing the dd and lowering the ss gave much better results and also faster ones...

render that exact same scene with the exact same settings, but turn the dd depth back down to 3 and check the render times.

the dd has absolutely no influence on samples and splotches. it's the much lower ss which overulled the dd setting in determining the lower time in this case.

STRAT
08-12-2005, 02:03 PM
http://www.nikclark.com/strat/pool1.jpg


http://www.nikclark.com/strat/pool2.jpg


http://www.nikclark.com/strat/pool3.jpg

handige_harrie
08-12-2005, 02:38 PM
Tbh I don't see the difference in artifacts. I really tried :). It shows how much textures/materials (and jpeg compression) can do to cover up artifacts. The problem with the slight darkening is solved easily in post (curves in PS).

Btw, very nice scene and faked caustic (have you tried other noise variants? the voronoi is quite recognisable ;))

STRAT
08-12-2005, 02:56 PM
Tbh I don't see the difference in artifacts. I really tried :). It shows how much textures/materials (and jpeg compression) can do to cover up artifacts. The problem with the slight darkening is solved easily in post (curves in PS).

Btw, very nice scene and faked caustic (have you tried other noise variants? the voronoi is quite recognisable ;))

you're right, the splotchyness and artifacts are minimal, (they show up more when you overlay each image in photoshop). that and the texturing, and the dark scene make all 3 pics look most acceptably similar. i think the slight darkening might be an error on my side, i think i probably lowered a light setting by accident. naughty boy :p

the caustics aren't a c4d noise. they're actually a water ripple noise bitmap i created separately for another job ;)

ok, here's the final rendering -

http://www.nikclark.com/strat/poolfinal.jpg

strength - 100
accuracy - 60
prepass - 1/1
diff depth - 1
stoches - 500
min - 200
max - 500

render time - 44 mins @ 3000x2250 pixels

I'm over the moon with that time and res. much much MUCH faster than i'm used too. and this gives me the scope to up a few more settings. although, would it make that much of a difference?

but i can confirm i'll still be staying with my dd of 1 for the time being though.

nycL45
08-12-2005, 03:06 PM
What is the DPI in the final rendering?

BTW, thanks for the textbook/professional quality approach.

STRAT
08-12-2005, 03:10 PM
dpi is the standard 72. unless i got specific printing purposes in mind my renders stay at 72 as there's no point in changing it.

AdamT
08-12-2005, 03:32 PM
but to my surprise this one which has a dd=5 and ss=350 rendered in just 36 mins compared to the ones before... now i am really puzzeled... why are you guys telling me to lower the diffuse depth and increase the stochastic samples when in fact increasing the dd and lowering the ss gave much better results and also faster ones...
I think you have to reconsider how you consider good and bad. When it comes to GI I consider a good render one that can be cleaned up in Photoshop or AE with relatively little trouble. Levels and contrast can be adjusted quite easily, but GI artifacts are often quite time consuming to fix--if not downright impossible.

When I took your three recent renders into PS and corrected for levels/contrast, I found that they were surprisingly similar, and all of them still have significant artifacts. Way too many IMO. They appear on the walls and on the sofa/loveseat in the foreground. What I would do in this case is apply compositing tags to those objects, specifying higher GI quality, while lowering scene GI quality to compensate for the increased render time. In the case of the sofa you only really need to worry about the side facing the camera--so I would make that a separate object if it isn't already.

While it's possible to generalize about GI settings, no two scenes are exactly the same. You have to experiment to get the best results. Now, if you have a big render farm you can just jack up the settings and not worry about it too much. If render times are an issue, however, there's no substitute for a little trial and error. I would suggest making good use of region render to speed up the process. In your scene, as in most, you can quickly identify the problem areas (sofa, clock, under the TV, over the door). There's no need to render the whole scene each time as you test your settings.

handige_harrie
08-12-2005, 03:34 PM
Great render and excellent rendertime (at that resolution: wow!) :thumbsup:. What's your rig; Dual Xeon 2.8 or something like that right?

the caustics aren't a c4d noise. they're actually a water ripple noise bitmap i created separately for another job ;)

I thought I recognised it :scream:.

STRAT
08-12-2005, 03:55 PM
Great render and excellent rendertime (at that resolution: wow!) :thumbsup:. What's your rig; Dual Xeon 2.8 or something like that right?



I thought I recognised it :scream:.

yup, a couple of dual 2.8's (linked up if needs be).


JondoeOne - it really is a case of experimenting and testing. alot of it. your scene looks like it has good potential, but as things stand the lighting rig you've chosen is in much need of improvement. things in that room hover or dont sit correctly. thats not necessarily a simple job for gi settings to rectify, but a job for you and your general lighting scheme to figure out. again, by allot of test rendering. you'll find these things come with experience.

i'm doing it the other way around - i can always get a spot on lighting rig and gi render (well, imo anyway :p ) but only now am i really beginning to understand the relationships of the gi settings i'm using. this is enabling me to optomise my render times even more.

but you must know and be able to achieve good lighting in the first place before messing around with complex gi settings imo.

i always advise peeps to learn and master lighting your model in the raytracer (or scanline) renderer first before looking into gi. master scan line rendering with an impressive fake gi method then your well on your way. real gi is just the next stage up.

jumping straight into gi rendering without mastering basic and advanced normal lighting is a common error allot of people make. it makes you lazy and assuming. gives you bad habits you dont know or understand how to break. but as i say, it all comes with time and practice.

remember my luxury housing image that once got selected for the cg talk select gallery a few years back? that was all scanline rendered with no real gi.

***dad mode off***

right, setting up another test scene of a shopping mall internal next. probably early next week. stay tuned.

AdamT
08-12-2005, 04:01 PM
Excellent point Strat. Aside from artifacts, the main issue with this render is that there's no direct lighting! Where is all that light coming from? The window? If so it would very dark in there, and there would be much more contrast. Add some recessed lighting and/or some free-standing lamps. Turn on the overhead light too. :)

govinda
08-12-2005, 05:46 PM
Strat, that is one handsome render to my casual, occasional GI user eye. I also recognize that noise pattern on the ceiling :), but simple to fix. The part where the rail comes down into the pool and down to the sweet blue sky tone on the green ripples is worthy of staring at and is enhanced greatly by full-res blowup. There's a 'bloom' on the steps that I'm curious about, but damn is this thing gorgeous!

AdamT
08-12-2005, 05:58 PM
Did I forget to say that's an awesome render? Damn Strat, that's an awesome render. :thumbsup: :thumbsup:

Per-Anders
08-12-2005, 06:23 PM
very nice render, my one comment is that the water surface needs much less diffuse/color, that type of water doesn't have much in the way of particulates nor a dusty surface and so where the light strikes it you don't get any more than a reflection/specular, no shadows on the surface itself.

nycL45
08-12-2005, 06:41 PM
I'm not a pool afficianado. With that said, Strat, what is the pole gizmo at the far corner of the pool?

The overall view and details of your rendering are really impressive. :thumbsup:

MJV
08-12-2005, 06:45 PM
Yes, excellent render. The lighting isn't purely realistic but nonetheless has a great surrealistic quality.

AdamT
08-12-2005, 06:56 PM
I'm not a pool afficianado. With that said, Strat, what is the pole gizmo at the far corner of the pool?
Just guessing, but it could be a lift for handicaps. Either that or some kind of decapitation device ... which would be sooo inappropriate.

STRAT
08-12-2005, 09:01 PM
well cheers fellers for the comments. that funny looking hoist thing is as Adam spotted - for disabled peeps. the whole model is a hydro theropy pool in a hospital. the building is being constructed as we speak, and this image is on a public billboard about 30ft long on site (i'm still yet to see it!)

i never use real caustics, like with stoch mode rendering - all tooooo slow to tight deadline production rendering. pehaps next time i'll use a c4d noise (as someone earlier suggested) rather than a bitmap.

mdme - i cant dissagree with your comment about the water surface. water is always a labour of love to create. and i so much enjoy rendering it. but yes, we'll never individually get it right and someone will always see it differently. also, i must have done a zillion test renders on it so by the end i probably lost sight of my target :o

govinda - that bloom on the steps you talk about is decieving. it's actually on the pool surface, separated by the handrail/door frame shadows. the acute angle of the strong sunlight is giving the water's surface this bloom effect, not the submerged steps.

cant wait to test render my mall scene next week now with various gi settings. i'll keep you all well documented. what a learning thread this is for me. cheers fellers. but of course, please keep on posting your gi insights up.

praps i should submit this to the main cgtalk gallery maybe.

nycL45
08-12-2005, 09:30 PM
that funny looking hoist thing is as Adam spotted - for disabled peeps.

:thumbsup:

...the building is being constructed as we speak...

Congrats! Kudos.

what a learning thread this is for me.

Terrific! :thumbsup: :thumbsup:

jondoe0ne
08-13-2005, 07:28 AM
loool
thanx for the advices... you know strat... u may be right... i'll try and look into this problem more closely (god knows i have been messing up wit this scenefor a long long time...)

adam - don't get me wrong but what u consider artifacts in fact are thought to be like that (the shade under the tv and that greyness on the couch) some of them are made wit ambient occlusion to add to the realizm of the scene, also with that clock something is wrong for sure cuz it makes splotches no matter what settings i use...

strat - i really like your picture when i look at it for the first time.... but i we were to look more closely those extra lights put by you in the scene are showing up too much and unrealistic and the textures are not really realistic, at least the glaze tiles... the water is soooooo great!! the best i have ever seen.... but that light u have put in the window by the pool i think it ruins a great deal of the scene. also u could see more indirect lighting on the ceiling coming from the sky lights, while the light coming from thr lockers is just wonderful and very well chosen....

i like these discussions because i think they make us more avare of what we are doing and seeing what others do is so helpful and makes this the most fun way of learning... thank you guys u are just great!!

STRAT
08-13-2005, 08:22 AM
loool
thanx for the advices... you know strat... u may be right... i'll try and look into this problem more closely (god knows i have been messing up wit this scenefor a long long time...)

adam - don't get me wrong but what u consider artifacts in fact are thought to be like that (the shade under the tv and that greyness on the couch) some of them are made wit ambient occlusion to add to the realizm of the scene, also with that clock something is wrong for sure cuz it makes splotches no matter what settings i use...

strat - i really like your picture when i look at it for the first time.... but i we were to look more closely those extra lights put by you in the scene are showing up too much and unrealistic and the textures are not really realistic, at least the glaze tiles... the water is soooooo great!! the best i have ever seen.... but that light u have put in the window by the pool i think it ruins a great deal of the scene. also u could see more indirect lighting on the ceiling coming from the sky lights, while the light coming from thr lockers is just wonderful and very well chosen....

i like these discussions because i think they make us more avare of what we are doing and seeing what others do is so helpful and makes this the most fun way of learning... thank you guys u are just great!!


well, i kinda rushed this scene specially for this thread, i'm sure i could pollish it off nicely if i chose to ;)

the tiles need more work i agree. to me they look a too new and clean and the grout line is a tad chunky. i'd have loved to render in a denser reflection to them and have the reflection blurred out. (i'l re-render when 9.5 comes along for that bit ;) )

totally agree with the lighting in the window by the pool. i'll either turn the light down allot, or even take it out completely.

i messed with trying to get more lighting from the skylights using omnis, but i wasn't happy with the shadows cast. perhaps some gi casting 'invisible' object would have done a better job.

and about your own model - as mentioned, realism and a cleaner image that sits nicely will only come with experience. if it doesnt happen with this job move on to the next job. then when you think you've finally made head room come back to this one and apply your new knowlege.

but the problems lie in your general light set up, not your gi settings. all archi scenes should be lit using direct lighting as they are in the real world (ie, sun light, internal light fittings etc etc). the gi mode is only there as a fill in tool. you shouldn't rely on the gi mode to 'make' the model.

i'll light my scenes nicely first with normal direct lighting before i put gi mode on, then when i'm hapy with the spots and omnis i've set up i'll switch on a general gi emmitting sky dome and turn down my normal lights to suit.

artemesia66
08-13-2005, 01:02 PM
great render, strat. the only thing that stands out for me is that the plants look really dark--maybe a bit of luminance or backlight to simulate translucency?

nycL45
08-13-2005, 02:10 PM
Relying on 9.1 radiosity settings to eliminate artifacts may be a lost cause or waste of time. I finished a fairly large run of tests on one scene varying accuracy, stochs, min, and max and found the run with the second fastest time and ordinary settings was the best. By ordinary, I mean accuracy=80%, stochs=400, min=80, max=100, dd=1, and ps=1/1.

In this case, best means least amount of artifacts and least negative effect on the rendering in general. The bottom line: at the end of every test there were still artifacts requiring removal in post. Make friends with the inevitable.

STRAT
08-13-2005, 03:22 PM
9.1 is perfectly good for eliminating all artifacts, but at a render speed penalty.

i dont think the current version of c4d is particulaly speedy for interiors. for interiors i find the best solution to get rid of the common splotch factor is usually to play with upping the max samples and using a pre-pass of 1/1.

try your exact settings again but double or even treble the max samps, and use a 1/1 prepass, as your accuracy against stochsamps looks fine. (you could prolly even lower your accuracy down by 10-15% even) .the trouble is though is when you double the max samps the render time will inevitably shoot up.

and thats another suprising thing i discovered during these interior tests too - the prepass size. a pre-pass of 1/1 really did make a significant point of minimising the splotchiness in certain cases. whereas rendering exterior shots i'll rarely use a 1/1 prepass. usually 1/3.

but yes, a nice clean interior can take an age too render. whereas exteriors are another matter. c4d is BLISTERINGLY fast at rendering a nice clean slotch free exterior.

but i'm also blessed with my boss buying me a couple of dual xeons, which most people struggle to even think about. that 3000 px pool interior only took 44 mins to render, and imo it's fairly splotchless and clean. i dread to think what time it would have taken one your average single proc pc that most people use. (another reason why i'll use a half fakiosity half real gi lighting rig in most cases in my work)

nycL45
08-13-2005, 05:00 PM
The attached matrix gives you an idea of the range of settings I tested (thanks to the two threads on this subject).

With Mac, we have Grab, a scrn shot utility, which allows for brain-free overlays. I overlayed and flipped the different settings/images and the red line test yielded the most useful results.

Strat, recognize line 18? It was good, but not a winner with my little interior test.

Ernest Burden
08-13-2005, 07:27 PM
All this raises a question: Will adding a compositing tag that forces GI accuracy to __' override the render settings?

If a render is looking really good at 72% accuracy, except for one wall, could we add a 'force GI to 90%' and it would work? Or can it only force UNDER the global seeting?

STRAT
08-13-2005, 07:40 PM
If a render is looking really good at 72% accuracy, except for one wall, could we add a 'force GI to 90%' and it would work?


absolutely m8, thats why it's there. i use it constantly. very handy indeed :)

dann_stubbs
08-13-2005, 07:45 PM
absolutely m8, thats why it's there. i use it constantly. very handy indeed :)

i thought at one time you could only force down? is it able to force up too now?

dann

JamesMK
08-13-2005, 08:07 PM
i thought at one time you could only force down? is it able to force up too now?
Unless it has recently changed in R9.5, it can force both up and down.




.

AdamT
08-13-2005, 08:14 PM
i thought at one time you could only force down? is it able to force up too now?

dann
I think you're confusing it with the AA setting which can only be forced *up*. GI can go up or down.

dann_stubbs
08-14-2005, 01:01 AM
I think you're confusing it with the AA setting which can only be forced *up*. GI can go up or down.


ah, yes... it was AA so many apps - so many settings - so little sleep

: )

dann

STRAT
08-17-2005, 05:12 PM
Back again :thumbsup:

Ok, i been testing again today. This time it's a shopping mall interior.

C4D isn't particulaly uber fast at getting interiors looking real, as far as a clean and accurate GI solution is concerned, so hopefully these settings will shed some more light on the issue.

This is an old scheme i've bought back to life just for this topic. I've optomised the AA settings somewhat, and heavily optomised using various material and render tags too. There's 1 distant hard shadowed light for the sun, and 6 low light emitting omnis in there for in-fill lighting (set at 30% brightness and soft shad with the lowest samples i could muster).

The AA settings aren't ideal, but for a deadline render the quality is more than acceptable.



First off my old settup -

strength - 100%
accuracy - 99%
prepass - 1/3
Diff Depth - 1
stoches - 400
min res - 75
max res - 200

@ 800x600 i switched this render off after 2 hours rendering because i was getting fed up. It had rendered about 1/2 way through, and the image quality, GI wise, was looking fairly average.


So to my new settings. The following images show the subtle differences resulting between the settings used, and again, i'm over the moon with the render times.

http://www.nikclark.com/strat/mall01.jpg

http://www.nikclark.com/strat/mall02.jpg

http://www.nikclark.com/strat/mall03.jpg

http://www.nikclark.com/strat/mall04.jpg

http://www.nikclark.com/strat/mall05.jpg


Here is the best solution i chose for the final image -

strength - 100%
accuracy - 40%
prepass - 1/1
diff depth - 1
stoches - 600
min res - 100
max res - 500

http://www.nikclark.com/strat/mallfinal.jpg

3000 x 2250 pixels @ 50 mins. it also had 3 or 4 mins in post touching up a few splotches, but this was a very minimal job i assue you.

over and out for now :)

nycL45
08-17-2005, 05:38 PM
Strat, the two joints in the fascia panel at the far end of floor opening, this side of the escalators, that appear through the glass of the guardrail are broken and almost gone. Is glass a problem in this regard?

Thanks for the enlightening thread. Will you be updating it with 9.5?

BTW, is there a site where we can gawk at your work?

STRAT
08-17-2005, 07:26 PM
thats the low aa settings. all i need to do is set a render tag to the glass upping the aa. and yes, i will be upgrading as soon as 9.5 is shipping. render times hopefully will atleast double if not treble then :)

you want to gawk at my imagery? not a good idea m8 :eek:

nope. i dont have a gallery site :sad:

nycL45
08-17-2005, 10:08 PM
One sad gawker here. :cry: :)

Looking forward to your 9.5 update. BTW, you said "...render times hopefully will atleast double if not treble then" Huh? Double? Treble? Yikes! Time to :banghead:

STRAT
08-18-2005, 08:11 AM
mods - can this thread be moved to the c4d lighting and rendering forum or t&t forum maybe?

MPS
08-18-2005, 09:49 AM
Strat, excellent work mate. What AA settings do you use?

ps I saw a post you did on CGArchitect a long time ago, stating your old settings. I remember thinking, this guy's work is awesome, I'll copy his settings.... Glad to see even the best artists are learning new stuff. It's been a really good thread for me as my GI rendertimes are killing me.

jondoe0ne
08-18-2005, 12:02 PM
see strat.... that's why i don't like working with internal fill lights... because they are generating this kind of whiteness where it isn't supposed to be....

STRAT
08-18-2005, 12:33 PM
cheers dudes.

MPS - i used some heavey aa optomisations on all the glass in the render tags, and the general render settings for aa are BEST, STILL threshold - 15% and min/max are 1x1 and 4x4. again, this scene is more for a gi test, which is why i optomised AA. if i was submitting to a client i'd obviously use a much more pleasing aa

jondoe - thats the words of inexerience ;)

this scene was relatively rushed in the little time i had free, so minor things like light bloom on the balconys and AA errors are miniscule things that can EASILY be solved with more carefull testing.

in my experience infills are almost a must. if you rely on the gi solution alone to render your models without the aid of infills and a carefully thought out lighting rig to take some of the burden of render times off your hands, you'll end up with much longer render times that'll cripple you in deadline situations.

most if not all archi pros use clever use of infills as standard. thats why your image in this post isn't correctly lit, and it shows. clever use of infills in your scene would make a world of wonders m8 ;)

nycL45
08-18-2005, 01:37 PM
Strat, you will probably hatethis idea, but, how about you doing a thread on lighting? I know about "Digital Lighting and Rendering" by Jeremy Birn, but you are in the viz biz and, like the rad settings, your treatment would be focused on that business and not general theory.

How about it? Say it's a :thumbsup:

Srek
08-18-2005, 02:16 PM
mods - can this thread be moved to the c4d lighting and rendering forum or t&t forum maybe?
Hi,
i think as long as it is an active discussion we should keep it here, after that we'll move it for easier reference to the rendering subforum. If i forget that please remind me ;)
Cheers
Björn

AdamT
08-18-2005, 02:40 PM
Personally, in a fully modeled scene like Strat's, I would turn off material specularity altogether.

artemesia66
08-18-2005, 02:51 PM
Personally, in a fully modeled scene like Strat's, I would turn off material specularity altogether.

And just use reflectivity?

STRAT
08-18-2005, 03:36 PM
specularity is almost at 0 in the scene. the marble floors have a modest specular value. some of the soffits have a miniscule value, but the glass has a high value, but it's a pin-prick bloom setting.

AdamT
08-18-2005, 03:59 PM
And just use reflectivity?
Yep. Even if you have your specs set very tight they can cause a nasty glossy look on a flat surface--like a window or mirror.

artemesia66
08-18-2005, 04:02 PM
Good to know. :thumbsup:

STRAT
08-18-2005, 04:08 PM
Yep. Even if you have your specs set very tight they can cause a nasty glossy look on a flat surface--like a window or mirror.

i'd have to dissagree with substituting the ref for the spec in this case. even though a ref isn't has harsh as a spec, it does significantly effect render speeds, and will show up a reflection when ur not expecting it.

a tight spec will also cause probs, but it can be played with. most of the specularity settings on the materials other than glass and marble have an extremely high width and a minimal height. just to seggest a slight shean. i'd go with this over the ref method in architecture personally.

AdamT
08-18-2005, 04:18 PM
Good points, and if speed is a big concern specs might be the only way to go. But ... if there's time I'm going with pure raytraced reflections, especially in 9.5 where blurry reflects are actually feasible.

Ernest Burden
08-18-2005, 05:37 PM
Ah, speed. Gotta have speed.

I went away for a few days (Cleveland!) and left my machines re-rendering that exterior animation I posted samples from. One machine went down due to a power outage (the others are on UPSs and were fne) but even taking that into consideration, it went way too slowly. I use camera animation and the setting tweeked a bit from reading this thread, but it's still just nowhere near done--the frame rendering time-leak problem, as STRAT has made clear, has doomed me. I'm going to have to pay a farm to render it, and also the interior portion for my project.

Let's think about having to spend $1000 per animation project, or $10,000 for additional CPUs just to get to where you can actually get you work done. That really raises the cost of Cinema, now doesn't it?

OK, most of my work is stills, so the GI settings are also important since we still cannot use multiple machines to render a still (not effectively, certainly).

Oh, and STRAT--you didn't test your interior scene with 1/3 prepass. I have foolishly decided this is a key to good, fast results. Please try it on your scene to help show how right or wrong I am.

AdamT
08-18-2005, 06:17 PM
I'm thinking/hoping that fR-2 will be the savior for GI animations.

MPS
08-18-2005, 07:00 PM
I agree with you Adam and Ernest, it costs way too much too use a Renderfarm unless you have alot of financial clout behind you, and it takes far to long to do a 800 x 600 size animation with the quality that Strats shown here. Sadly clients see the still image and they don't understand why we can't offer that kind of quality at a mid-large screen res, all of course for a small charge and within a narrow deadline (I have between friday to next wednesday to arrange 5 images and an 30-60sec archi animation, whilst waiting agesfor architects to confirm whether planned features are gonna change or not and whether they are happy with the model/camera angles etc...)

I really hope that FR2/AR2.5 sorts the speed of animation rendering out. Any money saved on using renderfarms can be spent on new software like FR2 and R9.5 etc, so we can keep taking our work to another level, which keeps bosses happy ;)

dann_stubbs
08-18-2005, 07:51 PM
I agree with you Adam and Ernest, it costs way too much too use a Renderfarm unless you have alot of financial clout behind you,

i don't really agree with that perspective. all it takes is decent planning, i routinely have open subscriptions - most users i have don't wait until the last minute to start that huge render either - i guess it also takes communication to the client to understand there is a cut-off to their changes to allow time for the rendering.

last minute *anything* is usually the most expensive - that is how traditional renderfarms justify their high cost.

dann

Ernest Burden
08-18-2005, 08:20 PM
i don't really agree with that perspective. all it takes is decent planning, i routinely have open subscriptions - most users i have don't wait until the last minute to start that huge render either...

Whether its client changes, an under-performing freelancer, or my own unproductive workflow (or a sick kid, power failure, Monday, whatever) I rarely get to hit 'render' when i should. Its life as I know it, and it appears to be fairly common among commercial artists.

'All it takes is decent planning' suggests that it is possible to know in advance how long its going to take to render a 4000 line GI architectural interior, or 2400 frames at DVD res using at least decent settings. So is it? For me it isn't.

Problem 1 is my own lack of understanding of how different GI settings affect rendertimes. Seems that's going around.

Problem 2 is this render-time creep-up on the one flicker-free animation choice we have under AR2. How do I plan for test frames at 3:00 that, after 25 frames under NetRender are actually rendering at 20:00?

Decent planning? How?

MPS
08-18-2005, 09:22 PM
Your right Dann about the planning phase, except when work buddies are off work unexpectedly, leaving the workload to be shared between everyone else. In my case there is only myself left to work on this particular project, which we had 2 weeks to start from scratch and finish, and the deadline got moved forward two days to wednesday, you can't plan for that kind of thing. Now don't get me wrong, these type of things happen and I'm not saying the work can't be done and that renderfarms aren't useful, because its the really the opposite (renderfarms can be last minute lifesavers). It's just I'd rather spend cash on more 3d software toys to play with than to pay a renderfarm. And the quicker the renderer the less I have to rely on renderfarms and the more of the budget that gets spent on c4d updates etc. Sadly bosses don't really want to pay for new stuff like FR2 and you've got to argue the case to get them.

FredSpeaks
08-18-2005, 10:04 PM
I have yet to find GI to be useable. Beside it taking too long, there is the annoying jump in solution from Net client to Net client. I attempted once on a small interior that had a few camera cuts tom minimize the lenght of indiviual paths to render, but it was completly unusable. Instead, I have just taken to using many weak omni lights with inverse falloff to fill the scene. I provides the about the same effect for interiors as a light dome for exteriors, and provides a constant rendertime for frames.

If baking in 9.5 is all that it is cracked up to be, then a baked GI solution may be the answer.

dann_stubbs
08-18-2005, 10:39 PM
'All it takes is decent planning' suggests that it is possible to know in advance how long its going to take to render a 4000 line GI architectural interior, or 2400 frames at DVD res using at least decent settings. So is it? For me it isn't.

Problem 1 is my own lack of understanding of how different GI settings affect rendertimes. Seems that's going around.

Problem 2 is this render-time creep-up on the one flicker-free animation choice we have under AR2. How do I plan for test frames at 3:00 that, after 25 frames under NetRender are actually rendering at 20:00?

Decent planning? How?

all i can suggest is when you get a walk through take your "average" frame time (maybe an average of all previous renders to date) and multiply that by number of frames - put that on the quote as the needed x amount of time to render project from LAST change. then the client cannot hold you responsible as they will sign off on the quote i would hope. if they question it - just use logic as to every other profession - it takes time. painting, catering etc... all need a fair time to produce the requested result.

while my renderfarm is not the largest out there, $395 is pretty low cost - not the $1000 or $10,000 options you mention. yes i only have a few subscription spots - that is to prevent users from the last minute render messing with other users access. i don't think i've ever had you use renderking so your slight exaggerations as to how to accomplish this job are just on the high side to me.

a 2400 frame render at 20 minutes per frame (average) at DVD res will finish in 33 hours on my small farm. plus with planning you can render early low res tests and re-render as needed hopefully to meet your deadline easily... there are options out there, i do the best i can to offer a very fair service. i had a couple empty subscriptions a week or two ago - it would have been very easy for you to have utilized the farm or planned ahead for this weeks render which i think would have been a help considering your complaints posted about time.

nobody will disagree that faster rendering is a desire - but in the meantime there are options and decisions that must be made to accomplish projects now. if all options are not being evaluated then that is maybe the thing to complain about.

over my work years i've seen a few examples of the "no time to do it right - but there is time to do it over", and that is the same with rendering... i've been very interested in STRAT's posts and render times - it just goes to show that a couple hours tweaking (even by a seasoned pro) can save potentially hundreds of hours on the render side. that is nothing to be dismissed, and i myself try to offer suggestions to users when i notice a job is rendering much longer then the result image implies.

9.5 looks to be a nice addition and optimization - i will have it on the farm as soon as it ships - but no amount of speed increase by maxon will overtake the improper use of render settings by someone in a rush. : )

i am actually curious now how you handle the render side of your projects? i've always tried to educate a bit on that and the time to process... is that not common in arch vis work?

dann

STRAT
08-18-2005, 11:16 PM
Sadly bosses don't really want to pay for new stuff like FR2 and you've got to argue the case to get them.

exactly!

i want FR2 after reading recent posts. it could be the answer to gi animations. but then, so could AR2.5

my boss will happily upgrade me to 9.5, but if i want extras like FR2 i must make a damn good case, which you'd have thought would be easy, but to him it's just spending more money. why should he buy FR2 aswell, if he's spending money on 9.5. isn't 9.5 good enough? why has he just wasted his money? you cant win. trying to weedle money out of ur boss, who technically is a complete layman, is a nightmare. he just thinks i want more gizmos and gadgets. i hope FR2 will be on demo. i really do.

EBIII - i'll try that exact interoir mall scene, with all the same settings but use a 1/3 prepass to check out render times and get back to you.

but, i must admit, i used to use a 1/3 pass constantly in my early days as i couldn't see the gi quality difference, but these dyas i can. it's subtle, but noticable imo. i've upped my usual 1/3 setting to a 1/2 setting these days. and occationally for certain renders i'll use a 1/1.


nycL45 - me doing a thread specifically on archi viz lighting? i guess i already am ;) one day i plan on making a niel blevins style archi viz website. but then i been thinking about that for over 5 years now.

nycL45
08-19-2005, 12:22 AM
Shake a leg, Strat. We're champing at the bit. :bounce:

:thumbsup: :thumbsup:

Ernest Burden
08-19-2005, 02:37 AM
while my renderfarm is not the largest out there, $395 is pretty low cost - not the $1000 or $10,000 options you mention...i don't think i've ever had you use renderking so your slight exaggerations as to how to accomplish this job are just on the high side to me.

a 2400 frame render at 20 minutes per frame (average) at DVD res

I wasn't clear. I was making a very general statement on what I was going to have to spend to get around the 'run-up' flaw in AR2. Without it, I could render my own frames in the case of my current project (the first time I've used C4D for a client animation project). However, with it, I cannot. So I did get a quote from a farm (not yours) and if am honest and admit that my average time is going to be 10-20 per frame for images that test at 3 minutes per frame, then I would spend about $750. Ah, but this is only the exterior portion with a single SUN light. I will need to render an interior with many lamps. Probably less frames overall but likely higher rendertimes than the exterior. So overall, I think I will have to consider myself luck if I can get this rendered for $1000.

This is not a client-billable item. I priced a completed piece, the cost of rendering is mine to bear. I simply didn't know that my target average time wasn't going to be possible for animation with AR2. I still have to render the thing.

I cannot get my head around booking up a farm for a month when I need only a day or so of actual hours. You make a good point about render tests. I didn't talk to you about this project because I am not sure when I will be ready to run the interior and because I didn't know for sure I would need a farm for a piece of this length.

The $10,000 number refers to the cost of putting in more computers, whether 8 single CPU machines, four duals, or that quad/dualcore AMD BOXX is building. OK, the BOXX may be $12K - $14K, but its only one machine to mess with. I come up with '8' since I have bought an '8' licence to Maxwell, and somehow the number applies to FinalRender2 (though I cannot figure that out yet, either).

i've been very interested in STRAT's posts and render times - it just goes to show that a couple hours tweaking (even by a seasoned pro) can save potentially hundreds of hours on the render side. that is nothing to be dismissed

I have followed this thread with interest (even helped inspire it, I think). I had already done alot to get my quality/speed optomized with the anim. I keep talking about. I still hope to learn more about how the major setting work in practice.

i am actually curious now how you handle the render side of your projects? i've always tried to educate a bit on that and the time to process... is that not common in arch vis work?
dann

I started with advice from STRAT and others here and at cgarchitect.com, and built as I went. I am not new to digital rendering, quite the opposite. But every project is better optomized.

MPS
08-19-2005, 09:40 AM
I think that the problem (challenge) for us as artists will be that we are always looking to improve the work we do. Clients see better and better Images/animations and we are getting better and better at using GI solutions for our work. I think this is proven by the improvements seen on this thread by Strat, who as we know is a seasoned veteran at 3D, and by his own admission has learned a lot from his own GI setting tests. I'm lucky to have also benefited from these test via this thread. Clients want as good a job done as possible and they have every right to want it. It just means that we as artists have to find the optimal workflow to get this output. With time some clients via experience might understand how long rendering takes and be a little bit more patient but in general most clients will expect the work ASAP. (Thanks again Strat and everyone for sharing your knowledge and experience).

STRAT
08-19-2005, 09:57 AM
done some prepass testing.

this scene has NO direct lighting. purely IBL with a white sky object. this is because the prepass generally effects the pure gi settings, gi lighting and samples more than direct lighting gi, or a mixture. (ie, the more omni and spot infills you have the less the prepass generally effects the image)

settings for both -

strength - 100%
accuracy - 70%
diff depth - 1
stoches - 600
min res - 200
max res - 0


image 1, prepass of 1/1, rendered @ 800x600 in 8 mins -

http://www.nikclark.com/strat/rh1.jpg




image 2, prepass of 1/3, rendered @ 800x600 in 7 mins -

http://www.nikclark.com/strat/rh3.jpg


if you overlay these 2 images in photoshop you can see a significant sample quality, particulaly behind glass (where the gi is trying to work even harder) and under hard edges like balconies and window recesses. obviously a higher general rate of samples would help, but as stated, this is a prepass excercise.

EBIII - i tested that mall image with a 1/3 prepass size, but the image quallity was near identical as the 1/1 tests, and the render speeds reduced only by literally a minute or so. but every scene will be different - the pool image for example only rendered nicely at a 1/1 size. a 1/2 size was noticably different to the point of unnacceptable.

So, for me personally the prepass size is determined by testing. generally a 1/2 is a nice size for me these daysble imo.

Ernest Burden
08-19-2005, 01:16 PM
EBIII - i tested that mall image with a 1/3 prepass size, but the image quallity was near identical as the 1/1 tests, and the render speeds reduced only by literally a minute or so...
So, for me personally the prepass size is determined by testing. generally a 1/2 is a nice size for me these daysble imo.

Thanks for that. Interesting that the times didn't change much. With my tests (just one model, though) I found the times were quite improved with 1/3.

Stephane Bernie
08-19-2005, 05:43 PM
all i can suggest is when you get a walk through take your "average" frame time (maybe an average of all previous renders to date) and multiply that by number of frames - put that on the quote as the needed x amount of time to render project from LAST change. then the client cannot hold you responsible as they will sign off on the quote i would hope. if they question it - just use logic as to every other profession - it takes time. painting, catering etc... all need a fair time to produce the requested result.

dann

I must vouch for Dan's service. I've only needed to do two walkthrus for my projects ( I rarely offer the service to my clients, they are usually happy enough with stills.) but on two occasions we proposed a quick 2 minute walk thru of a retail store concept to "wow" the client's investors and the suppliers. I checked with Dan if he had a free spot ( he did), I calculated the number of frames and factored in an average render time ( example 5 minutes/frame or 10 minutes/frame.) That gave me my render time budget ( i think it was 75 hours). We then told the client when was the drop dead date and added the render farm subscription as a production fee on top of our billable hours. The client accepted and that was it.

BUT, changes happened nevertheless ( change this wall color to blue, add product, blablabla). We did not stop the renderfarm... what we did, is that we would freeze the animation at certain points and fade in/out a separate rendered still with the changes.

That worked well enough and were able to keep our sanity....

Regarding GI in animations, i have one rule until proven otherwise: DON'T. I have yet to hear a client specifically ask " please make me a global illumination render" or saying " hmm... i don't know... seems as though your stochastic samples are too low....."

I render GI stills of different views along the walkthru, then use it as a benchmark to place fill lights to achieve a close enough effect. Also, I add a dash of Post effect glow in Cinema to soften the lighting even more. Result: good enough lighting, no artifacts, no stress, short render times.

Cheers,

Stephane

handige_harrie
08-19-2005, 09:23 PM
Ah, speed. Gotta have speed.
I went away for a few days (Cleveland!) and left my machines re-rendering that exterior animation I posted samples from. One machine went down due to a power outage (the others are on UPSs and were fne) but even taking that into consideration, it went way too slowly. I use camera animation and the setting tweeked a bit from reading this thread, but it's still just nowhere near done--the frame rendering time-leak problem, as STRAT has made clear, has doomed me. I'm going to have to pay a farm to render it, and also the interior portion for my project.
That sounds bad, have you tried my suggestion (http://forums.cgsociety.org/showpost.php?p=2529498&postcount=287)?
Considering the amount of money you're mentioning, I'd give it a try. To get the standard-gi flickering acceptable it would mean upping the settings but I think despite that you can catch up compared to animation-gi after a couple of frames.


Back again :thumbsup:
Ok, i been testing again today.
(...)
over and out for now :)
This time I disagree slightly with you the other way around. For me the first render is on the verge of acceptability. Of course it's very personal when you talk about what is acceptable in terms of (gi)quality, so don't see it as crits. But what strikes me is that in all the tests the only place where I can see artifacts is the walkway ceiling in the upper left corner of the image. The rest seems clean and more than acceptable in all the renders. I would put on a / change the compositing tag on that part and up the gi-settings. Apart from that you're right in saying that you can clean it up in post anyway. I also found the final highres render a lot cleaner but that's the postwork I guess (?).

Anyway, great tests and very helpful results:thumbsup:

Ernest Burden
08-19-2005, 10:00 PM
That sounds bad, have you tried my suggestion (http://forums.cgsociety.org/showpost.php?p=2529498&postcount=287)?

I did. I ran a test pass of frames at 1/2 size, but the flicker made them unusable. I figured I could get away with it, but evan after my post effects, it was very noticeable, windows going from dark to light and right back again, that sort of thing. Standard GI doesn't work for my animation.

The funny thing is, if using any GI in animation will cause a 20:00 render (that should be 3:00) I am tempted to use stochastic, if I could only get it down to a decent time. At least that wouldn't creep-up or have flicker.

jondoe0ne
08-19-2005, 10:10 PM
I did. I ran a test pass of frames at 1/2 size, but the flicker made them unusable. I figured I could get away with it, but evan after my post effects, it was very noticeable, windows going from dark to light and right back again, that sort of thing. Standard GI doesn't work for my animation.

The funny thing is, if using any GI in animation will cause a 20:00 render (that should be 3:00) I am tempted to use stochastic, if I could only get it down to a decent time. At least that wouldn't creep-up or have flicker.

but what are u animating...? i am very curious, could u post some stills?

Ernest Burden
08-19-2005, 10:22 PM
but what are u animating...? i am very curious, could u post some stills?

Kinda gets lost, doesn't it?

post 53 (http://forums.cgsociety.org/showthread.php?t=265726&page=4&pp=15)

STRAT
08-19-2005, 10:41 PM
But what strikes me is that in all the tests the only place where I can see artifacts is the walkway ceiling in the upper left corner of the image. The rest seems clean and more than acceptable in all the renders. I would put on a / change the compositing tag on that part and up the gi-settings. Apart from that you're right in saying that you can clean it up in post anyway. I also found the final highres render a lot cleaner but that's the postwork I guess (?).

yes, that part is the obviouse offender. and it perhaps wouldnt suprize you to learn there is already a separate gi tag on that object in the renders bumping up the gi. obviously not enough. because the general gi looks good, i think i can bump up that particular setting and have scope to play around.

i'll test it out on monday. could be interesting.

jondoe0ne
08-19-2005, 10:49 PM
Kinda gets lost, doesn't it?

really sorry, i knew the picture but i haven't really associated the pictures with your name....
and what differences did u have between the two methods... and which do u prefer?

handige_harrie
08-19-2005, 10:56 PM
STRAT: Yeah I'm curious to see the results :).

I did. I ran a test pass of frames at 1/2 size, but the flicker made them unusable. I figured I could get away with it, but evan after my post effects, it was very noticeable, windows going from dark to light and right back again, that sort of thing. Standard GI doesn't work for my animation.

The funny thing is, if using any GI in animation will cause a 20:00 render (that should be 3:00) I am tempted to use stochastic, if I could only get it down to a decent time. At least that wouldn't creep-up or have flicker.

Ah ok, then I rest my case for standard-gi ;). It's a pity though that it takes so long.

Have you thought about doing a separate gi/ao pass composed together with a regular-lighting pass? Or do you need the colorbleeding effect?

Ernest Burden
08-19-2005, 11:26 PM
Have you thought about doing a separate gi/ao pass composed together with a regular-lighting pass? Or do you need the colorbleeding effect?

I would prefer, in an all perfect world, to use stochastic mode to render a GI animation. That's just going to take too long per frame, even for a relatively short piece like mine.

So that only leaves camera animation. I have cars that move and the sun (plus its color-changer material) but the cars are set to not be seen by GI (though I am able to have them recieve GI). Camera animation does the trick. It's just with that nasty time creep, even with setting it to 'recalulate always'. Oh, well.

I did consider going a seperate pass for GI, but wouldn't it take about as long to render? I tested a 1/2 size GI laid over a scanline render, but the GI was visibly fuzzy as it affected the base.

FredSpeaks
08-20-2005, 12:45 AM
This rendered in 5 min on a 1.8 G5. 14 low brightness omnis with low res soft shadows and inverse square falloff. It isn't GI, but it is pretty close. Plus this is 5 min for every frame, and the client doesn't know the difference.:)

http://atechstudio.com/clients/mike/FillLights.jpg

jondoe0ne
08-20-2005, 10:11 AM
This rendered in 5 min on a 1.8 G5. 14 low brightness omnis with low res soft shadows and inverse square falloff. It isn't GI, but it is pretty close. Plus this is 5 min for every frame, and the client doesn't know the difference.:)

my oppinion is that is not even close to gi.... it's just a render with many (maybe too many) infill lights...
why don't you try this: put a sun light through the window (a distant or a paralel light), on the spot where the sun light hits the floor put a no shadow omni, put another light in the window maybe an area light with inverse square falloff, and then slowly light the darker places in the room, and don't forget to play with the light's color and shadow density...
i am not an expert, ...strat says i'm inexperinced :PpPp, but maybe this works better...

handige_harrie
08-20-2005, 01:26 PM
I would prefer, in an all perfect world, to use stochastic mode to render a GI animation. That's just going to take too long per frame, even for a relatively short piece like mine.

So that only leaves camera animation. I have cars that move and the sun (plus its color-changer material) but the cars are set to not be seen by GI (though I am able to have them recieve GI). Camera animation does the trick. It's just with that nasty time creep, even with setting it to 'recalulate always'. Oh, well.

I did consider going a seperate pass for GI, but wouldn't it take about as long to render? I tested a 1/2 size GI laid over a scanline render, but the GI was visibly fuzzy as it affected the base.

Yeah stochastic mode would be perfect, considering the grainy look you're after (right?). Although it's true that it takes to long to render for animations if you want acceptable quality.

As for separate passes, I believe it does save you rendertime because in the gi-pass:
-you only use plain grey materials (of which nearly all should be set to receive gi only) and np heavy (raytrace) materials or high-res textures.
-you can use proxy models (or just keep it simple and lower spd settings and hN-subdivision values)
-you can get away with more artifacts because of the control you have in post (mix modes and opacity).

It does depend on how much time the plain-lighting scene would take on average per frame. You need to know that first. As well as the time it takes now (with gi).

If you want, I'd be happy to do some testing :).

Ernest Burden
08-21-2005, 02:55 PM
in my experience infills are almost a must. if you rely on the gi solution alone to render your models without the aid of infills and a carefully thought out lighting rig to take some of the burden of render times off your hands, you'll end up with much longer render times that'll cripple you in deadline situations.

I'm bumping my head on what should be an obvious answer--

OK, so you advocate infill lights to save on GI time. But if you wanted a light to illuminate without being considered in the GI solution, how do you do that? Would you simply drop a compositing tag on the light and uncheck 'seen by GI'? Lights have a settings pane for caustics, but not radiosity.

confused...

AdamT
08-21-2005, 03:20 PM
The infill lights are considered in the GI solution. They save time by giving the GI algos more samples to work with, thereby cutting down on artifacts. In other words, they're used in order to avoid increasing diffuse depth.

Ernest Burden
08-21-2005, 03:40 PM
The infill lights are considered in the GI solution. They save time by giving the GI algos more samples to work with, thereby cutting down on artifacts. In other words, they're used in order to avoid increasing diffuse depth.

Interesting.

Interiors lit by all downlights, for example, with a dd of 1 render exremely slowly.

I understood STRAT to mean he was using a non-GI infill, but there's no such animal. I suppose the falloff helps, though, by limiting the sphere of influence of a light.

I'm lighting the interior of my hospital project now...

STRAT
08-21-2005, 08:52 PM
Interesting.

Interiors lit by all downlights, for example, with a dd of 1 render exremely slowly.

I understood STRAT to mean he was using a non-GI infill, but there's no such animal. I suppose the falloff helps, though, by limiting the sphere of influence of a light.

I'm lighting the interior of my hospital project now...

yup, what Adam says. i use infill omnis not only to easy the pressure on the gi calculations, but because i use a low dd (always because i work to tight deadlines) the infills also add to the bounced light effect, ergo adding to the realism of the image.
and besides, even if i were using higher dd, i'd still use infill omnis to add to the realism.

EBIII - imagine a higher dd than 1 with downlighters though. interiors with c4d's latest AR (ver 2) still take an age to render. personally my saving grace at the mo is a couple of fast rendering machines.

definately keep us up to date with those hospital interiors.

Ernest Burden
08-21-2005, 09:19 PM
definately keep us up to date with those hospital interiors.

The fun never ends!

Right now I'm testing GI settings on the interior. I found that I almost have to use a dd of 2 to get light to show on the ceiling.

One problem with onteriors is that the designer sets up fairly even lighting, they need to avoid bright and dark areas. But for a rendering that makes for a boring picture.

I have a series of ceiling panels that are currently set to use a luminous mat to generate light. But its way too even, so I will have to put some sort of light (spot probably) onto each polygon, which will take a while. Besides, the lum. mat is slow to render.

I have not gotten away from big GI atifacts, but using ideas from this thread its getting better.

jondoe0ne
08-21-2005, 09:31 PM
yeah! i've also learned some pretty cool tips around here... nice thread!

handige_harrie
08-21-2005, 09:32 PM
I've done some gi interior tests myself. 36 Actually; 18 with and 18 without infill lights.

Made a quick&dirty page for it:
http://www.shademaster.nl/peter/radiositytest.htm

The ones I think are the best (least noticable artifacts) are no. 1, 6, 16 and 19.

http://www.shademaster.nl/peter/radiositytest/jpg/026.jpg
(no. 6; with the same settings as the scene that's up for downloading)

Ernest Burden
08-21-2005, 09:52 PM
What I do to set up lighting is isolate each light type, so I can better see what its doing:

the panels, luminous mat:
http://www.oreally.com/temp/UH-int-panels-00.jpg

panels again, with settings:
http://www.oreally.com/temp/UH-int-panels-01.jpg

accent spots. How they get aimed I am not yet sure...
http://www.oreally.com/temp/UH-int-spots-00.jpg

There will be some sunlight coming in from a few big windows at the ends of the hall. That will use a distant light set very bright plus an area light to simulate environmental general light from outside.

STRAT
08-21-2005, 10:32 PM
i personally in architecture never use (or rarely use) a strength of over 100. probably just my own preference, but i find it help further on down the line with keeping a constant. i find anything over 100 unnatural too. use infills to compensate if it's too dark. even though they're part of the gi solution they dont take hardly any time to render (as long as they're properties have low shad samps)

looking nice EB, but with all that detail, and when aa is turned on you may find a dd of 2 will kill render times. i'd put in some more infill omnis, all instanced up of course, with either ranges or fall off on, with a very low soft shad setting (say 30% maybe, with lowest samps).

also, see what it does to render times if you increase the min samps to 30 or 40.

have you tried playing with the saturation levels too maybe?

Ernest Burden
08-21-2005, 11:24 PM
have you tried playing with the saturation levels too maybe?

No, never have. I would like you to explain how much over 100% you use for GI saturation.

I am being careful to use falloffs and distance stps on the lights. I have put in all the down spots for the ceiling panels. There are 58 panels, so 58 lights. dd-1 is working better now, but dd-2 looks better, but at about twice the render time, so obviously I will use dd-1 if I can. I've almost never used dd over 1 before.

Ernest Burden
08-21-2005, 11:39 PM
and when aa is turned on you may find a dd of 2 will kill render times.

AA is not always used because of my post stuff.

OK, so dd-2 looks better, but not really. Here is the same render with dd-1 (the darker portions) and dd-2. At first dd-2 looks nocer, but it actually lacks some contrast that I'm going to want later. If you brighten up the dd-1 version, they are similar. Also note how the pink color bleeds in dd-2, the other color is the accurate one:

http://www.oreally.com/temp/UH-int-panels-02.jpg

panchopantera
08-21-2005, 11:42 PM
Ernest I like your work with all the post you do to it but I don't understand why you would be worried about getting clean photorealistic gi if you are going to draw all over it to have your final particular quality. A quick question - the people on your renderings, are they drawn or are they part of the same effect you you do to your renderings? I mean are they photographs before you do your magic?


JC

Ernest Burden
08-22-2005, 12:57 AM
I don't understand why you would be worried about getting clean photorealistic gi if you are going to draw all over it to have your final particular quality.

Because good input yeilds good output. However, some Gi artifacts actually help my work look hand-done. Just not too many.

the people on your renderings...are they photographs before you do your magic?

Some are, and some aren't. Sometimes I am using painted figures (re-using, really) and more recently I have been using photos that are altered in Photoshop. I also have been experimenting with 3D rigged and animated figures with non-photoreal textures for animation pieces, though I have, so far, gotten away without people in architectural animation...but I won't for much longer.

panchopantera
08-22-2005, 01:50 AM
Lovely work, specially the Columbus Centre and the image on the middle of the page with the girl in the foreground drinking the cocktail. Ok sorry for the OT carry on with your tests.

Ernest Burden
08-22-2005, 02:59 AM
Now we're getting somewhere.

I had to have a light for each ceiling panel, but just a down-pointing light would not get the glow near the panel that I wanted. I tried a number of things, including non-rendering/non-shadowing planes to bounce light back up. That caused bad GI problems, since its a GI effect. Simpler to add STRATs beloved fills. (Damn you, muttonman, damn you)! So I copied the down light to be an up light. Since I had used Instances for all these lights I only have to modify the master, and play with settings on it, to apply to the whole model. Nice!

The accuracy is still low, at 33% here, but the lights are about right:
http://www.oreally.com/temp/UH-int-panels-03.jpg

another view, plus how I will treat the end product:
http://www.oreally.com/temp/UH-int-panels-04.jpg

gettin' there...

AdamT
08-22-2005, 03:24 AM
Looking good!

STRAT
08-22-2005, 08:10 AM
definately starting to sing now! keep going

STRAT
08-22-2005, 08:37 AM
But what strikes me is that in all the tests the only place where I can see artifacts is the walkway ceiling in the upper left corner of the image. The rest seems clean and more than acceptable in all the renders. I would put on a / change the compositing tag on that part and up the gi-settings.

Anyway, great tests and very helpful results:thumbsup:

ok, i did have one already in place at 80% accuracy, but obviously not high enough. this looks better, with a compositing tag gi accuracy forced up to 98% -

http://www.nikclark.com/strat/mall06.jpg

handige_harrie
08-22-2005, 12:24 PM
STRAT: That certainly looks better, although now I also notice there are some large gi-artifacts (circles) at the very left of the image (behind glass). Although for a second I thought it was a texture; so if it's just a still you could easily get away with that.10mins :thumbsup:.

Ernest: That looks wonderful! Although I feel the post-processed image is a bit overexposed. Maybe this is what you want, to my taste it takes away just to much detail and color.

dann_stubbs
08-22-2005, 12:31 PM
Now we're getting somewhere.

I had to have a light for each ceiling panel, but just a down-pointing light would not get the glow near the panel that I wanted. I tried a number of things, including non-rendering/non-shadowing planes to bounce light back up. That caused bad GI problems, since its a GI effect. Simpler to add STRATs beloved fills. (Damn you, muttonman, damn you)! So I copied the down light to be an up light. Since I had used Instances for all these lights I only have to modify the master, and play with settings on it, to apply to the whole model. Nice!



that really makes the lights come to life... nice work!

dann

nycL45
08-22-2005, 12:45 PM
After all the great info flowing to and fro here, is it safe to say, as an elementary starting point, that extremely simple interior scenes require low min/max and complex scenes need much higher min/max? (Assume acc% 70-80+/- and ss 200-600+/- and final settings should be the result of testing.) I know this is a generalization but, it's based on handige_harrie's renderings/matrix and strat's tests here.

STRAT
08-22-2005, 01:07 PM
After all the great info flowing to and fro here, is it safe to say, as an elementary starting point, that extremely simple interior scenes require low min/max and complex scenes need much higher min/max? (Assume acc% 70-80+/- and ss 200-600+/- and final settings should be the result of testing.) I know this is a generalization but, it's based on handige_harrie's renderings/matrix and strat's tests here.

not nessesarily so. i thinks handige's tests are great, but to make it more of a user database i'd have added a couple of boxes or other primatives into the square room just to make the gi calcs slightly more realistic (who ever renders an empty square box? throw in a square stool or table and the gi calcs would be transformed). saying that, i've still printed out a copy of his results pinned on my wall here :)

every scene is different. to me, a simple interior scene is comparable to some of the living rooms we've seen in this thread, where as a complicated interior is similar to my mall or EB's hospital.

the difference is NOT the amount of samples needed to get a clean render (min/max), but the amount of geometry in the scene.

a simple box room with a table and chairs may require the same amount of min/max samples to render them as smoothly as a complicated model like EB's, but EB's model has allot more of those little bits in there, multiplying the time shed loads.

so it's the mesh thats the offender, not the samples. you need similar sample settings, but a heavey mesh will obviously take allot longer to render them just due to the fact there's more detail in there.

handige_harrie
08-22-2005, 05:06 PM
Well it'd be worth repeating the test with some more poly's inside (furniture). Obviously I kept it empty and simple to keep the rendertimes down (as I don't have a dual xeon *jealous* ;)).

If I can find the time I'll look into it. (Too bad you cant batch render 36scenes). Antoher thing that would be handy is something like exif information output to store rendersettings in the rendered image.

Btw, it's funny that the default radiosity settings I nearly always use (rendering detailed models outdoor) are those of no. 6. Fast and virtually no artifacts. So for me they definately are a starting point.

andronikos916
08-23-2005, 12:25 AM
no time to read the thread guys now - just want to subscribe and read it carefully soon. Hope to post some tips too. (if you haven't covered everything)

have fun,
Andronikos

Ernest Burden
08-23-2005, 02:07 AM
no time to read the thread guys now - just want to subscribe and read it carefully soon. Hope to post some tips too. (if you haven't covered everything)

Oh, there's plenty of room for more wisdom. We havenot reached any conclusions, either.

By the way. I had left the exterior portion of my hospital animation running on one PC, even though I know I am going to have to pay a farm to render it because of the time run-up bug. I just had a look at how it was going--the sequence of frames that started out at about 3:00 per frame has now climbed to over an hour per frame. Isn't that lovely? Even with a good farm, this bug is going to cost me a lot of cash. I want to use some bad words now.

andronikos916
08-23-2005, 11:43 AM
Oh, there's plenty of room for more wisdom. We havenot reached any conclusions, either.

By the way. I had left the exterior portion of my hospital animation running on one PC, even though I know I am going to have to pay a farm to render it because of the time run-up bug. I just had a look at how it was going--the sequence of frames that started out at about 3:00 per frame has now climbed to over an hour per frame. Isn't that lovely? Even with a good farm, this bug is going to cost me a lot of cash. I want to use some bad words now.


I totally understan you bug Ernest! ...I hope that in 9.5 this "bug is fixed". Maybe with the baker? Don't know yet - haven't tryied it yet. (don't have 9.5) :scream:

By the way, Which C4D renderfarm are you planning to use?

cy,
Andronikos

STRAT
08-23-2005, 11:56 AM
i dont think it is a bug, i think it's just one aspect of c4d gi rendering that hasn't been fully thought out by the programmers because it's not an obvious problem and not apparent at first glance.

hopefully the more and more peeps that report this issue Maxon will sort out the rendering algorithms.

it's strange why some suffer and some dont. i suffer constantly from it, even with reletively simple models.

andronikos916
08-23-2005, 12:03 PM
yes it is not a bug but it is... lol

cy,
Andronikos

Ernest Burden
08-23-2005, 12:05 PM
i dont think it is a bug, i think it's just one aspect of c4d gi rendering that hasn't been fully thought out

Well it sure isn't a 'feature'.

it's not an obvious problem and not apparent at first glance.

When the camera animation option was added, and the language about how it is there to speed up rendering of animation, you would think they would have tested the point before releasing it, and before that all-too-ironic manual description.

I'm planning to use RenderCore, mostly because they have such a huge farm that regardless of cost (which keeps climbing the longer I let this thing run on my other PC) will still be completed in an hour or so. At this point, rendering 3 minutes of animation will cost about what a full Cinema Studio seat costs. Without the bug (and how can an unexpected negative consequence not be a bug?) I would have rendered on my own tiny farm for no additional cost.

nycL45
08-23-2005, 02:58 PM
Re: #162 above.
Strat, let me approach this from another angle. Two scenes: 1/ a small still life, say a booze bottle and glass closeup and 2/ an EBlll hospital. Would you start your min/max tests for scene 1/ on the low end and for scene 2/ at your mall range? Or, would you use the same min/max for both scenes and test down or up from there?

The point? Running multiple radiosity settings tests for the best combo can =+++ time and the results could add/save time. Handige used 4/12 (#6) and you used 75/200. Isn't there a rule-of-thumb starting min/max combination, based on complexity of the geometry, lighting, etc., one can start with and test below and above these settings to the best? Say, as a starting point, a still life, 6/30 is suggested and an atrium, 60/300 is suggested. Now, run x, y...n tests above/below these for the best.

BTW, AdamT helped with suggested min/max ratios, 1:10, 1:5, 1:1.

Right now, it's a game of infinite scenes, infinite combos and time.

Ernest Burden
08-23-2005, 03:14 PM
Running multiple radiosity settings tests for the best combo can =+++ time and the results could add/save time. Handige used 4/12 (#6) and you used 75/200. Isn't there a rule-of-thumb starting min/max combination, based on complexity of the geometry, lighting, etc., one can start with and test below and above these settings to the best?

Right now, it's a game of infinite scenes, infinite combos and time.

It sure is. I am still hoping to glean some understanding of relationship between the actual scene and good settings.

Also, do the min/max setting refer to the base units? For my work its either 'foot' (typically) or 'meter' (occasional). But many of you use cm, right? would a min/max of 4/12 be the same on a (correctly built) scene in units of cm vs. meters? In one a door is 200 units tall while in the other its 2 units.

STRAT
08-23-2005, 03:55 PM
Re: #162 above.
Strat, let me approach this from another angle. Two scenes: 1/ a small still life, say a booze bottle and glass closeup and 2/ an EBlll hospital. Would you start your min/max tests for scene 1/ on the low end and for scene 2/ at your mall range? Or, would you use the same min/max for both scenes and test down or up from there?


again, depends on geometry. for scene 1, a small still life, i may choose to start my min/max testing at say 30/50. for the mall or EB's model i'd start testing at 30/100. both scenes starting off with say 100 stotch samps.

so, the same min settings, but different max settings, because i can better judge max settings to suit what mesh complexity i got. then, in both scenes i'd do 2 or 3 more renders with radically different min settings to see what was the fastest. anything from 20-60 usually suffices for most scenes i find.

usually then i'll keep the min settings at that chosen level. then i'll up the stoches a couple of hundred together with the maxes, whilst playing with the accuracy. but then, some times, again depending on the model, you may need a very high min sample that supercedes the maxes.

but again, for me personally, it's down to intuition.




The point? Running multiple radiosity settings tests for the best combo can =+++ time and the results could add/save time. Handige used 4/12 (#6) and you used 75/200. Isn't there a rule-of-thumb starting min/max combination, based on complexity of the geometry, lighting, etc., one can start with and test below and above these settings to the best? Say, as a starting point, a still life, 6/30 is suggested and an atrium, 60/300 is suggested. Now, run x, y...n tests above/below these for the best.

BTW, AdamT helped with suggested min/max ratios, 1:10, 1:5, 1:1.

Right now, it's a game of infinite scenes, infinite combos and time.


6/30 for the simple life and 60/300 for the complex scene, as starting points sounds ok. coincidence maybe, but ok. yes, i'd start the testing for a simple scene at 6/30 maybe, but for a complex archi model? maybe 60/100 first off, just to establish my min settings.

as for Adam's min/max ratios? i've heard this before and personally dont see any benefit in them.

corrolations between mins, maxes, and stotches - is there any? i dont think there is a hardfast staunt formula to go by.

as i say, i find a min anywhere up to 100 usually suffices, a max somewhere up to 300-400, stoches around the 500-700 mark, and an accuracy of 50-70% maybe. depending on deadlines.

STRAT
08-23-2005, 04:09 PM
quick example of what i'm working on as we speak. it's a small 3 story external of a building, with a load of windows, balconys and plain white walls. about 20k polys maybe, so nothing spectacular. (sorry, cant show you the exact scheme, but VERY similar in fashon to this building) -

http://www.nikclark.com/strat/gitest.jpg



the final render specs i chose -

strength - 100
acc - 70
prepass - 1/2
dd - 1
stoch - 600
min - 100
max - 400

@2000x1500 it renders at about 10 mins.




i started testing with -

acc - 70
prepass - 1/3
stoch - 100
min - 30
max - 100

i first established the optimum min samps was 100, i then tested the max samps up to 400, the stoches were slowely upped to the 600 mark (which gave a nice quality/speed combo with the max samps), then before final rendering i upped the prepass to clean the render up.

this testing sequence took me maybe a whole of 4-5 mins.

nycL45
08-23-2005, 09:40 PM
Strat, do you think I'm trying to nail you to "a hardfast staunt formula to go by"?. Me? No, no, no. ;)

dann_stubbs
08-25-2005, 05:58 AM
Strat, do you think I'm trying to nail you to "a hardfast staunt formula to go by"?. Me? No, no, no. ;)


ernest - i thought this link from the maya forum (i know blasphemous - but i follow both) may make you feel better

http://forums.cgsociety.org/showthread.php?t=255746

translation - don't belive the hype, or the marketing guys! : )

dann

STRAT
08-25-2005, 08:36 AM
sorry nyc, i wasn't implying that m8 :p

but i too am trying to find a rough link/connection behind the thinking stage when setting up the gi solution to a render.

i got an idea in my head, but i think that just comes from experience. i use a shed load of incidental direct lights in my work (did i mention that yet? :p ) so my gi solutions probably differ from most.

as a guide for myself, by the time most of my final settings are established, my max samps are somewhere in the region of about 10 times higher than my mins, and the stoches are 2-3 times that of the max samps. the dd is always set at 1, the strength always to 100 (any thing more or less i'll control it via the render tags or material props), the prepass is to suit, as is the accuracy (60-90%, depending on the time i have to spare).


the next tests i'll try is with the min V's max V's stoch samps. i want to see when an overiding min sample (higher than max) is beneficial, also i'd like to see how to keep the stoches really low whilst compensating with the min/max settings. and how all this effects render times.

i'll post up again when i got a freee slot to test :)

Pegasus
08-25-2005, 08:42 AM
What do you think about a mini contest about this topic?

destro80
08-25-2005, 08:53 AM
What do you think about a mini contest about this topic?

I've been following this thread but not posting because I haven't done GI renders yet.
I'd be happy for it to continue as it is. Over 170 constructive posts and still moving up at a good pace.

Maxon can upgrade AR which is great, but threads like this are a "user upgrade"...and free.

andronikos916
08-25-2005, 08:57 AM
destro80 : "Maxon can upgrade AR which is great, but threads like this are a "user upgrade"...and free".

best commnet ever.:thumbsup:

cy,
Andronikos

Ernest Burden
08-25-2005, 12:44 PM
ernest - i thought this link from the maya forum (i know blasphemous - but i follow both) may make you feel better...
translation - don't belive the hype, or the marketing guys! : )

The Max forum would turn my stomach, but I have respect for Maya. I'm even letting my own son learn it (wouldn't let him onto C4D, by the way). I have spend most of my time after Siggraph--since discovering the disaster that is C4D GI-based animation--thinking I was going to have to move to Maya.

But Final Render-2 may keep me in the flock. AR2.5 probably does not solve the problem.


No-one has addressed my question about scene units, and whether that afects the min/max numbers. Because I'm looking at STRAT's results as a guide, but his are from a model based in either cm or meters. Does that make a difference?

STRAT
08-25-2005, 01:05 PM
about material saturation -

No, never have. I would like you to explain how much over 100% you use for GI saturation.

this is a great, under-used option. as most peeps know there's a difference between gi and radiosity, even though they are linked and the resulting consequences are similar (but thats for another thread :p ). The saturation spinner is great for aiding in the radiosity effect (ie, the colour bleeding effect).

here is a quick test i did. the model and render settings are rough and ready and only used to demonstrate the saturation option. it's a fantastic option to keep in mind when you have walls or floors that need 'livening' up for instance.

notice the nice colour bleeding from adjacent objects as the spinner is increased. in these tests i've increased the saturation level to all materials except the pink wall and the mirror.


image 1, all saturation levels at the nominal 100%

http://www.nikclark.com/strat/bathroom01.jpg


image 2, saturation upped to 200%

http://www.nikclark.com/strat/bathroom02.jpg


image 3, saturation upped to 300%

http://www.nikclark.com/strat/bathroom03.jpg


image 4, saturation upped to 400%

http://www.nikclark.com/strat/bathroom04.jpg


image 5, saturation upped to 500%

http://www.nikclark.com/strat/bathroom05.jpg


image 6, saturation upped to 1000%

http://www.nikclark.com/strat/bathroom06.jpg


as you can see, saturation can be a hidden gem, (render times stay constant) specially with non-reflective white walls. as i say, it helps simulate the radiosity effect, but also greatly enhances the gi effect too :)

STRAT
08-25-2005, 01:13 PM
No-one has addressed my question about scene units, and whether that afects the min/max numbers. Because I'm looking at STRAT's results as a guide, but his are from a model based in either cm or meters. Does that make a difference?

my standard units are 1 c4d unit = 1 mm (same i use in autocad and max).

in my c4d testing i've found no effect on model sizes or units to the render solution or times. if i scale up one of my typical models 1000X larger, or 1000X smaller, it renders in exactly the same time and quality.

this is good as certain other 3d apps certainly do have scale issues when rendering gi.

i have noticed though, specially on certain c4d shaders and particulaly the latest AO shader (not the new 9.5 one), that the scale of the model has a massive bearing on how that shader or plugin works.

moka.studio
08-25-2005, 02:05 PM
my standard units are 1 c4d unit = 1 mm (same i use in autocad and max).

in my c4d testing i've found no effect on model sizes or units to the render solution or times. if i scale up one of my typical models 1000X larger, or 1000X smaller, it renders in exactly the same time and quality.

this is good as certain other 3d apps certainly do have scale issues when rendering gi.

i have noticed though, specially on certain c4d shaders and particulaly the latest AO shader (not the new 9.5 one), that the scale of the model has a massive bearing on how that shader or plugin works.

I find C4d works best when I work at a scale similar to the primitves ( and a bit bigger rather than smaller). It often means scaling the models when importing from other apps, but I have gotten used to it. It just affects too many parameters, even if the GI is still fineas you say ( I have neveer tried Gi on a x1000 scaled scene). BUt the size of soft +area shadows, the size if procedurals etc etc, not to mention the modelling tools, or dynamics, are all geared to work at a certain size
jp

nycL45
08-25-2005, 04:12 PM
... are all geared to work at a certain size
jp

What is the size? What happens if the size is much smaller/larger than this optimum (?) size?

Strat, I was kidding, too. Helpful illustration of GI Saturation, btw. I've wanted to know about that for some time. I'm surprised it does not add to the render time load. A bonus. :thumbsup:

moka.studio
08-25-2005, 04:21 PM
What is the size? What happens if the size is much smaller/larger than this optimum (?) size?

Strat, I was kidding, too. Helpful illustration of GI Saturation, btw. I've wanted to know about that for some time. I'm surprised it does not add to the render time load. A bonus. :thumbsup:

The computer explodes.
NNo, sure you can work with very large geometries or very small, just the tools are optimized for a certain size.
The size is roughly based on the size of the c4d primitves, meaning a size at which the prmiitves can be used as buillding blocks ofr the scene. If you start working with scenes much bigger or much smaller, say by a factor of x1000, you will see what I mean...
jp

STRAT
08-25-2005, 05:17 PM
as a guide for myself, by the time most of my final settings are established, my max samps are somewhere in the region of about 10 times higher than my mins, and the stoches are 2-3 times that of the max samps. the dd is always set at 1, the strength always to 100 (any thing more or less i'll control it via the render tags or material props), the prepass is to suit, as is the accuracy (60-90%, depending on the time i have to spare).


the next tests i'll try is with the min V's max V's stoch samps. i want to see when an overiding min sample (higher than max) is beneficial, also i'd like to see how to keep the stoches really low whilst compensating with the min/max settings. and how all this effects render times.




i been testing that mall image again today with all kinds of combinations of samples, and was getting some very nice renders, but times varied massively.

in conclusion, the best solutions by far are the types of relationships and settings i posted and tested a few pages back.

no obviouse sample relationships you can shake a stick at, but i'm so far happy with the situation i mentioned above

keep comming fellers! i need to learn more!

STRAT
08-26-2005, 08:05 AM
i got an easy day today, so if there's any specific area of testing you want me to carry out pls let me know :)

nycL45
08-26-2005, 02:13 PM
This is more of a (lingering) question than a test: The tutorial, Radiosity Tut1MV.html part 1, starts with diffuse depth (dd) and covers settings 1 to 5, with higher the setting, the higher the quality and r. time. In this thread, dd 1 is mostly used. The AR manual notes settings up to 100 can be used and that 3 is the default and "is suitable for most scenes." (Higher dd can be used with "open scenes".) Is there a combo of the radiosity settings that will give => quality without jacking up the r. times when higher dd is used? Could you lower the ss, min, max and not get the artifacts with higher dd?

STRAT
08-26-2005, 02:33 PM
dd has absolutely no bearing on quality and artifacts or gi splotches

the manual states and average dd of 3 is usuable apt for most scenes, and allot of peeps use that standard, and without the use of incidental or infill omnis then yes, i agree, a dd of 1 (for internals or low lit environments), would be most inadequate for purely indirectly lit models.

but upping the dd neither contributes to enhancing image quality and neither does it create extra artifacts or splotches. it purely establishes indirect light bouncing. the higher the dd the more illuminated your model will be due to more light rays bouncing, but the rendertimes dramatically shoot up too.

ideally i'd always use a dd of 3-5 in most cases instead of compensating with so many infills, but i dont have the luxury of time. and with blood, sweat and tears testing for a light rig, you can get nearly as good results with your dd set to 1, but saving you massive render times.

there is no way around this. you cannot have a higher dd with lower render times. you can fake it and optomise it with infills and a lower dd, but upping the dd means higher render times, no matter how you optomise your samples.

eg - if you optomise the best samples rate for your render with a dd of 1, then you keep all the same settings but up the dd to 3, your image will certainly look alot lighter and maybe more cheerfully, but the quality and artifacts will remain the same. just the render times will rocket upwards.

STRAT
08-26-2005, 03:05 PM
lol, guess what? i did another test :) diffuse depth this time.

i did a rough and ready render of my mall, nicely lit with my infill omnis and a dd=1. it rendered in 1 min and 20 secs.

i then just changed the dd to 2. it rendered in 4 mins and 20 secs, but was obviously too bright.

i then left the dd at 2, and turned down my 6 shadow casting infill omnis to get a more pleasing light balance. render time was again, around the 4 min 20 second mark.

the second image had a more contrasted lighting solution, and the light saturations and radiosity looked richer and fuller (obviously due to 2 bounces and not 1 with infills to compensate) but the render time was nearly 4 times higher than the dd=1 scene.

and the artifacts and image quality in all the renderings remained constant. (if anything, it was better with the dd=1 image as there was less full on gi for the amount of samples given)



http://www.nikclark.com/strat/mall08.jpg



http://www.nikclark.com/strat/mall09.jpg


also, putting the dd=1 image next to the dd=2 image you can obviously see a difference, in radiosity and gi compaired to the faked gi helped image, but for realism? they looked as real as each other to me, and certainly not worth the extra 400% render time to get that extra bit of realism nobody would notice.

(i even tried a dd=3 render with exactly the same low settings, but it took over 5 mins even to render 1 line in the prepass!!!)

at the end of the day i'm sticking with a dd=1 because i cant aford not too.
it's all down to a good lighting rig. if you can develop an excellent direct infill lighting rig to supliment a low dd then you're quids in imo. gi on it's own just isn't the solution.

brammelo
08-26-2005, 04:45 PM
at the end of the day i'm sticking with a dd=1 because i cant aford not too.
it's all down to a good lighting rig. if you can develop an excellent direct infill lighting rig to supliment a low dd then you're quids in imo. gi on it's own just isn't the solution.

You will be one happy camper when you get 9.5 :)

nycL45
08-26-2005, 05:09 PM
Okay, Strat. Since you are still cranking here's another Q. You say "it's all down to a good lighting rig." A lighting rig is something you create with each project? It sounds like there is a method to your rigs, i.e., interior v. exterior, residential, commercial, skylight/no skylight, sky v. cans v. trouffers, etc. Would you like to elaborate on creating "a good lighting rig"?

BTW, thanks for looking at the dd question. IMHO, offering a dd range of 1-100 when a dd=1 is most useful creates unnecessary brain labor. This thread has certainly helped to simplify the radiosity monster. :thumbsup:

STRAT
08-26-2005, 05:24 PM
Okay, Strat. Since you are still cranking here's another Q. You say "it's all down to a good lighting rig." A lighting rig is something you create with each project? It sounds like there is a method to your rigs, i.e., interior v. exterior, residential, commercial, skylight/no skylight, sky v. cans v. trouffers, etc. Would you like to elaborate on creating "a good lighting rig"?

BTW, thanks for looking at the dd question. IMHO, offering a dd range of 1-100 when a dd=1 is most useful creates unnecessary brain labor. This thread has certainly helped to simplify the radiosity monster. :thumbsup:

i cant ever see why a dd of 100 would suit. to me, any dd above 5 (or even 3 in this example) means i may as well use stoch mode gi, and hey, we all know the problems there dont we kids :thumbsup:

as for lighting rigs, ur right, i create a different variation for each render, but based on a successfull similar previous job i might have done to give me a starting point :)

for externals, whatever the subject, i generally go for just a hard shadow casting distant light for my sun, usually around the 75% brightness mark, and i'll use a sky object with a plain white (or slightly blue) illum material at a strength of 100% assigned to it. to get the correct lighting i'll usually then just play with the illum percentages.

internally is another beast all together. I'll still use the same distant light for the sun, and the same sky object to cast light in through the windows (i'll usually crank up the illum settings for this too), but the internal rigs i use here usually differ greatly from job to job. if it's a multi story interior such as this i may use 5 or 6 omnis, in the middle of the model, between each floor level. in this example i have 12 omnis - 6 at more or less eye level, and another 6 between floors 3 and 4. all have a wide fall off, and soft shadows set at 75-100%. (and all shadow samples are the lowest you can muster), their brightness is set at about 20% with a slight yellow tinge. (all instanted of course).

to get any extra lighting i need i'll put in direct omnis or spots where actual lights may exist in the model itself. or i may put omnis in the rear of each actual shop unit to give the impression of light flooding out into the mall. also, i'll turn the shop front glass material to emitt a gi effect of 300-400%. this gives a lovely local radiosity effect. there's also a fair bit of material saturation differences in there to enhance the mood, and a fair few gi forced render tags to help optomisation and sample rates.

but as i say, it's all down to a fair bit of fiddling and experimentation. the more you do the better you'll get a feel for it all. this mall is probably about the 50th model similar to this i've rendered for clients in the past 3 years or so alone. i wish it were easier and quicker.

nycL45
08-26-2005, 05:52 PM
From post #1 of this thread: http://www.nikclark.com/strat/rh.jpg
This particular scene has about 50 internal lights (w/shads) and a few external. :p

This is another "I always wanted to ask" questions. Where are the 50 internal lights? Are they lighting the ceilings in the flats and the interior wall surfaces? What are shads? Shades? Your glass reflections are superb, did you do anything special to achieve the range from the lower floors to the upper floors?

Ernest Burden
08-26-2005, 07:12 PM
You will be one happy camper when you get 9.5 :)

Why's that?

The problems with long renders in AR2, at least in my case, are not a result of blurry reflecions or area shadows. I use neither. I will happily buy 9.5 as soon as I can, but it isn't going to make me a happy camper unless it solves the GI animation time run-up bug, which it isn't. Now FinalRender probably will, so that is more likely to make me happy.

Unfortunately both will come a few weeks too late for me, as I am going to end up paying a hefty sum to render my current project. Maybe 10% of my fee, if observed behaviour of AR2 is any guide. I am going to test the feasability of rendering with stochastic this weekend. Any port in a storm...

STRAT
08-26-2005, 08:34 PM
EBIII - i agree, the GI animation problem solved would really make me a happy chappy, but lets wait and see. if rendering is sped up atleast by 200-300% in the whole, perhaps, until maxon fix the problem, it'll make the slow down more tollerable. we may be able to use the new AO baking facility. if this is true, and it really is as good as it sounds then problem solved, but i got to see it to believe it first. also, i'd still prefer to render gi instead of AO given the choice.
i too am going to start badgering my boss for FR2 after i see reviews about it's animation performance.




This is another "I always wanted to ask" questions. Where are the 50 internal lights? Are they lighting the ceilings in the flats and the interior wall surfaces? What are shads? Shades? Your glass reflections are superb, did you do anything special to achieve the range from the lower floors to the upper floors?

the 50 internal lights are 1 in every room. again, all with soft shads (shadows), a large falloff, and all attenuated. they are 1 light each placed somewhere in the center of each room lighting up everything thats in the room, be it walls, ceilings, floors etc. but also, as with most of my archi glass i'll make it emit a very high gi effect.

ironically though for this image, all lights were not activated as you wouldn't see the rooms lit up in broard daylight such as this. :p

thnx about the glass comments :) nothing special in the glass - just a normal semi opaque material, black in colour, with fresnel switched on in the transparency slot and normal ray-traced refs. there's a huge value in the specular slot which is also controled by the angle of the sunlight. i'll also put in an invisible plain infront of the building with a detailed sky map on it specifically for reflections,

Ernest Burden
08-28-2005, 03:43 PM
EBIII - i agree, the GI animation problem solved would really make me a happy chappy, but lets wait and see. if rendering is sped up atleast by 200-300% in the whole, perhaps, until maxon fix the problem, it'll make the slow down more tollerable.

Let's not wait and see. We all have deadlines, and I'm not meeting mine as a result of this problem, or, it will cost me a lot of money to work around. Maxon has let us down on this, plain and simple. The 9.0 manual says camera animation is there to speed up your work, but it does the opposite.

The other problem with 'camera animation' GI is that a sequrnce gets an initial GI seed and runs with it, but each CPU is starting with a different seed (or so it appears) and therefore you will get a flicker when a NetRender'ed sequence is used. I've got three AMD CPUs from the same core family (two 2400+ and a 2700+) and they produce noticeably different results.

This is bad.

So what to do? Well, I decided to revisit our beautiful but unavailable friend--stochastic mode radiosity.

By using very low settings I am able to get acceptable results, at least for my process. My framerates are up, but using NR, a frame that tests at 15 minutes renders at 15 minutes, frame after frame. NO run-up. I havent compiled a sequence into animation yet, but I am not seeing anything to make me believe it will exhibit the sort of GI artifact flicker that I described above with cam-anim.

Here are how my endproduct frames will look:

camera animation, using my best time/quality settings:
http://www.oreally.com/temp/UHXS-1949-camanim.jpg

with stochastic, set to accuracy 10% dd-1 samples 20
http://www.oreally.com/temp/UHXS-1949-10-20.jpg


cam-anim
http://www.oreally.com/temp/UHXS-2249-camanim.jpg


stochastic
http://www.oreally.com/temp/UHXS-2249-25-18.jpg


The biggest problem is obviously the noise. Even though I'm adding noise, it's too strong for my tastes in anywhere in a shadow. Using higher samples helps, but the rendertimes were no good. Using a quality of 100% also made for a smoother result, but took too long. Interestingly, the smoothness did not change visibly from 0% to 10% to 25% to 50%. Just at 100% did the image smooth out, and at that setting, I could only do little region tests, not a whole frame.

The biggest issue for me is that the shadows in stochastic are shades of black, while the cam-anim brings more color, which is exactly what I want. But do we get what we want with C4D GI animation? No, we don't.

So I will reduce the amount of noise added to my frames and live with the results. Rendering animation in stochastic, can you believe it?

Testing a frame in stochastic yeilded a fairly linear result for settings:
accuracy/samples
5%/72 = 3:21
5%/36 = 1:43
5%/18 = 0:54

raising quality% didn't affect visual results much, or time for that matter. Untill you did 100% and then it was too long to even test.

On another point, AdamT, you were right. My SPD surface to make background cars and trees was adding a lot to rendertimes. I tried using the Jenna-made surface (thank you so much for doing that for me) but it was so high-poly that it would open, but couldn't really be manipulated in any way (2GB RAM in this machine). So I went about making it myself. It took a really long time to do, so write it off to a learning exercise.

I used my SPD map to make a 'relief object', made editable, cut out all the 'flats' between objects, did various poly reducing techniques. The result is a much leaner polygon set than the former SPD surface calculated for each frame.

for example, with frame 700:
cam-anim w/o SPD layer on=2:08 SPD=5:54 newpolys= 3:33
stochastic w/o SPD layer on=7:46 newpolys= 11:02

So while it was too much work to do the poly conversion of my SPD object, it shows that doing so will save a lot of time on the animation. What we obviously need is a tool in C4D to do a SPD>poly object conversion. Or, building scene differently.

lllab
08-29-2005, 02:47 PM
here a small project i just did,

one standard still - GI image:
rendered only 4:14min on 800*500px.
full size @5500*3540 was about ca. 3h 30min

modeled, textured and rendered in c4d 9.1

my settings:
92%
50%
1
25
1
11
always

cheers
stefan

ps. updated with a bit more contrast as suggested:-)

Ernest Burden
08-29-2005, 03:04 PM
here a small project i just did,


A really nice rendering! I would suggest punching up the contrast on the building in post, its noticeably weaker than the site image. Great design, too. Just a sun light, or STRATfills, too?

lllab
08-29-2005, 03:26 PM
thanks Earnest- very nice,

maybe i adjust the contrast a little, will see...

is is just a yellow sun with soft shadow highsamples, slightly blueish shadow at 95% density, some fill lights slightly colored and one of my outdoor hdri for skylight and reflection, slighly blurred, and a handmade sky as background.

i really looking forward for the new update, so i can finally use areashadows and blurry effects:-)

cheers
stefan

ps. the site is not a image, it is rendered too, just alpha trees, grass texture, plaza tiles...

moka.studio
08-29-2005, 04:17 PM
Very nice rendering Illab.
I will be picky though, I find the woman with the back turned, at the left of the image, too static. The ones close to the entrance look god, as well as the one to the front ( she also looks a bit cut out, but somehow it works well, to me, with the image).
- nice design too ( yours?)

Jp

lllab
08-29-2005, 04:56 PM
hi moka thanks,

yes you migth be right with the people, i was a little tired in the end, it was a very tight deadline...

the design is from a young ex-himmelblau partner, who now has his own well running office.
he is one of the best architects here in town imo.

i did only the rendering.

cheers
stefan

nycL45
08-29-2005, 09:13 PM
Good looking rendering, lllab. Thanks for sharing your settings and background on the landscaping. :thumbsup:

STRAT
08-30-2005, 08:38 AM
great rendering stefan, specially with the gi settings almost non existant! you'd probalby greatly reduce the render times even more by upping the min samps from 1 though.

did this whole thread ultimately help with your approach to gi render settings and optomisations?

moka.studio
08-30-2005, 08:48 AM
hi moka thanks,

yes you migth be right with the people, i was a little tired in the end, it was a very tight deadline...
cheers
stefan

I know that all too well.... ;)
jp

brammelo
08-30-2005, 09:25 AM
Why's that?

The problems with long renders in AR2, at least in my case, are not a result of blurry reflecions or area shadows. I use neither. I will happily buy 9.5 as soon as I can, but it isn't going to make me a happy camper unless it solves the GI animation time run-up bug, which it isn't. Now FinalRender probably will, so that is more likely to make me happy.

Unfortunately both will come a few weeks too late for me, as I am going to end up paying a hefty sum to render my current project. Maybe 10% of my fee, if observed behaviour of AR2 is any guide. I am going to test the feasability of rendering with stochastic this weekend. Any port in a storm...

I was perhaps replying to one of Strat's post about a good lighting rig...?

STRAT
08-30-2005, 09:46 AM
I was perhaps replying to one of Strat's post about a good lighting rig...?

unfortunately, as the AR stands (and even the new up-comming AR by the sounds of it) it'll still drastically slow down whatever lighting rig you use if gi is active as well. but yes, a 'fake gi rig' is usually the way to go, but it's not the real thing :(

i have my reservations about the new AO baking option too where medium to large complicated archi meshes are concerned too. altho i'm hopfully wrong :)

lllab
08-30-2005, 11:44 AM
thanks all,

strat: "you'd probalby greatly reduce the render times even more by upping the min samps from 1 though."-

thanks STRAT: this even reduced rendertimes from 4:14 to 3:39 (800*500) and this while working on my pc:-) i used 4-min 11-max, rest as above.

" did this whole thread ultimately help with your approach to gi render settings and optomisations?" m- well no not really this thread, but an old vray thread here though.
But Strat just helped to improve the settings, so this tread is starting to help:-)

this is one of my standard GI outdoor-setting. i developed it some time ago, adjusting and improving it from time to time. important are also the lights and a good hdri, not only the gi settings alone.

i think the new update will make it a lot easier to work with light and also GI, and i guess maxon will rework the rendercore so that we have bucktes and new GI in v10. until then there is also FR soon- a thing that makes me very happy:-)

cheers
stefan

c-hri
08-30-2005, 01:28 PM
great rendering stefan, specially with the gi settings almost non existant! you'd probalby greatly reduce the render times even more by upping the min samps from 1 though.

did this whole thread ultimately help with your approach to gi render settings and optomisations?

@strat:
I followed this thread since it started, but now
I'm quite confused : why will upping the min value (==more samples) reduce rendertimes??

@lllab:
nice rendering (speedy rendertimes, too).

lllab
08-30-2005, 01:38 PM
if the min settings are too low c4d has to "think" to long where to place the few samples, so thats why in my case a min of 4 was faster than a min of 1.

cheers
stefan

STRAT
08-30-2005, 01:44 PM
@strat:
I followed this thread since it started, but now
I'm quite confused : why will upping the min value (==more samples) reduce rendertimes??


because a min value too low also has an adverse effect on render times. it can take longer to calculate how to distribute a very low min sample setting compaired to something higher. (ie, a min setting of 1 usually takes allot longer to render than a setting of 10), which is way render testing to find an optimal min setting is advisable. in archi, usually something between the 20-30 mark is applicable.



EDIT

ah, too slow, lllab jumped in first :thumbsup:

EDIT

c-hri
08-30-2005, 01:54 PM
thx to both ;)

good to know that - because I always started with a 1 min setting for a gi scene setup. till today.

nycL45
08-30-2005, 02:29 PM
it can take longer to calculate how to distribute a very low min sample setting compaired to something higher.

Does this longer distribution calculations (cause) leading to longer render times (effect) explanation apply to low Max samples, too?

STRAT
08-30-2005, 02:36 PM
Does this longer distribution calculations (cause) leading to longer render times (effect) explanation apply to low Max samples, too?

nope, because max samps only effect mesh joins and angles/edges, whereas mins effect large flat open faces/surfaces, or the rest of the 'unaffected' model if you please. the lower the max samples the faster the render.

besides, visual wise you can get away with minimal min samps, you wouldn't want to use the same minimal max's as they're more visually obvious.

STRAT
08-30-2005, 03:13 PM
just been testing min samples for render speeds on my mall model, render times are in mins and secs -

min samps @ 1 = 2:06
min samps @ 5 = 2:01
min samps @ 10 = 1:37
min samps @ 25 = 0:46
min samps @ 50 = 0:30
min samps @ 75 = 0:27
min samps @ 100 = 0:28
min samps @ 125 = 0:31

so from this i'll use a min sample setting of 85. imo it's wise to always get your min setting established first during the gi testing stage, as this will then remain an optomised constant no matter what other settings you experiment with.

but of course, some meshes may require a min sample way higher than this. why higher than the max samples need to be set at. but in archi as a general, this isn't the case.

Ernest Burden
08-30-2005, 03:34 PM
so from this i'll use a min sample setting of 85.

What was the visual quality at the lower vs higher values? 85 certainly looks good on time, but is it also best on visual quality?

STRAT
08-30-2005, 04:16 PM
What was the visual quality at the lower vs higher values? 85 certainly looks good on time, but is it also best on visual quality?

i found the visual quality of min samps is usually neglegable compaired to max samps at these lower sample rates.

in this test a min sample of 1 looked rather different (worse) than a min sample of 10, but a min sample of 10 looked similar to a min sample of 100. different, but similar. although, the higher the min samples the more defined and 'real' the general lit feel of the model was, but the render times were starting to creep up steadily.

Also, when i upped the min samps to a certain figure, somewhere near 600 i think it was, the image was getting a bit blotchy, due to the fact i didnt have enough stoch samples to help compensate and go around, but as you can imagine, render times by now were un-usable.

The higher mins were giving a nice image, but silly render times, so the other settings are now used to compensate instead. playing with the accuracy, stoches and maxes will give just as good results with much lower render times.

which is why choosing a comparable min setting for good speed is probably advisable. (it's the first setting during gi settup and testing i always establish first)

the realism is now primarily controlled by upping the stoch samps and max samps. it's then a juggling game with these 2 settings. generally, even though you can get nice results with a lowish stoch setting and a massive max (or min) setting, it's generally best to use a higher stoch setting. (it's like trying to enlarge a small res photo in photoshop - the res isn't there in the first place to get decent results).

so a stoch sample rate of maybe up to twice that (or obviously higher if you prefer) of the max samps is used. then to make it even more real why not up the accuracy a bit. as i say, it's a juggling game of trial and error.

you can always keep upping the accuracy into the 90's (as i used to do) to get even better realism. but again, you need the other samples to aide in this which results in higher render times, but, after rendering with both accuracy levels, i think the higher is still more real than the lower. the main reason i changed was render times - i was getting very similar realistic results (not as real, but very similar) using the lower accuracy and higher settings than the higher accuracy and lower settings. but in both cases, the min setting i'd choose would be identical.

andronikos916
08-31-2005, 12:25 AM
just been testing min samples for render speeds on my mall model, render times are in mins and secs -

min samps @ 1 = 2:06
min samps @ 5 = 2:01
min samps @ 10 = 1:37
min samps @ 25 = 0:46
min samps @ 50 = 0:30
min samps @ 75 = 0:27
min samps @ 100 = 0:28
min samps @ 125 = 0:31

so from this i'll use a min sample setting of 85. imo it's wise to always get your min setting established first during the gi testing stage, as this will then remain an optomised constant no matter what other settings you experiment with.

but of course, some meshes may require a min sample way higher than this. why higher than the max samples need to be set at. but in archi as a general, this isn't the case.

thanx Strat for the info and test!

very much apriciated!

cy,
Andronikos

maikukai
09-16-2005, 10:49 PM
here a small project i just did,

one standard still - GI image:
rendered only 4:14min on 800*500px.
full size @5500*3540 was about ca. 3h 30min

modeled, textured and rendered in c4d 9.1

my settings:
92%
50%
1
25
1
11
always

cheers
stefan

ps. updated with a bit more contrast as suggested:-)


Stefan,

Could you elaborate on how your lighting rig is set up for this scene? It looks incredible, nice work.

lllab
09-21-2005, 12:22 AM
strat, what accuracy do you have with such high min samples?

i have a nice dense grid of red dots with a min of 10 and 90-95%accuracy. amin of 85 would make gi preepass all red??!

i guess i also depends on the size of the model. i always look that i get a nice even density on the prepass dots, but not too dense....

maikukai: well hard to explain, but i try:
it is a yellowish sun with soft shadows and very high shadowmap slightyl colored blue, 2 fill lights fromleft and right of the sun in complementary color, very low intensity fills on ground level, green where is grass, grey where is plaza. some tiny spots-bright where i want it lighter or highlights("light-painting")some light shave negative fall off. one negative light where i wanted it darker.
a lot color comes from a hdri i made-blu from sky green fromsides dark brown from floor.

cheers
stefan

maikukai
09-21-2005, 05:42 AM
Stefan,

Your explanation will do just fine. I am trying to work on my lighting skills and it is great to hear the specifics of what goes into such a great scene. Thanks for sharing.

STRAT
09-21-2005, 08:34 AM
strat, what accuracy do you have with such high min samples?

cheers
stefan

can you quote a specific example i've done? i dont know what instance ur talking about in this case :p

lllab
09-23-2005, 11:52 AM
i mean in your tests:
min samps @ 1 = 2:06
min samps @ 5 = 2:01
min samps @ 10 = 1:37
min samps @ 25 = 0:46
min samps @ 50 = 0:30
min samps @ 75 = 0:27
min samps @ 100 = 0:28
min samps @ 125 = 0:31

do use meters or cm units? maybe your models are bigger??hmmm.

cheers
stefan

STRAT
09-23-2005, 01:00 PM
i mean in your tests:
min samps @ 1 = 2:06
min samps @ 5 = 2:01
min samps @ 10 = 1:37
min samps @ 25 = 0:46
min samps @ 50 = 0:30
min samps @ 75 = 0:27
min samps @ 100 = 0:28
min samps @ 125 = 0:31

do use meters or cm units? maybe your models are bigger??hmmm.

cheers
stefan

gotcha!

first off my basic units are mm. i work in mm across the board from acad to max to c4d.

i dont come up with min samps and accuracy levels to suit each other or to get an even sample render in the pre-pass. whether i get a dense mesh of red samples or not, it doesnt matter. whats important to me is speed.

my accuracy is usually between 70 and 85% these days, with a min sample setting that gives me the fastest render - see tests.

you are correct, an accuracy of 95 with a min of 10 will give a nice coverage of sample dots, because of the high accuracy, and a min of 85 at this accuracy level would give a dense red coverage, but it wont nesesarily be slower. faster infact maybe, depends on the model. it's all relative. -

find your optimum min sample setting for a scene. lets call it 50. if you render with an accuracy of 95% but use a figure higher or lower than 50, the render will take longer than using 50.

it's kinda confusing, but as i mentioned earlier, i try to find the optimum minimum sample setting a.s.a.p. in the render test process. this will then stay a constant (unless the scene dictates it needs to be raised). All other settings can then be found after the min setting has been established.

all the render settings are very dependant on each other, and changing one setting will obiously effect all the other settings, but min samples are pretty constant. get that initial setting and keep it. changing the other settings wont generally effect the minimum sample value's effect on render times or quality.

lllab
09-24-2005, 12:00 PM
hi strat,

thanks for the answers,
it is now clear to me, you use mm, so your modell size is different, i use cm.

as for c4d having no "real cad units" this means your models are 10times bigger than mine.

also your min samples are about 10 times higher-that explained it i guess...

i also have my settings for speed+quality tested, but they where so different to yours thats why i asked.

cheers
stefan

STRAT
09-24-2005, 12:40 PM
that seems most strange.

i wouldn't have thought that size makes a difference to render settings whatsoever. if that was the case, then this thread with my results would only apply to peeps using the scale of mm like me. all other scales would require totally different settings. that cant be right :shrug:

AdamT
09-24-2005, 02:36 PM
AFAIK size is only important in a relative sense. In other words, it comes into play a lot if you have very large objects and very small objects in the same scene, but if you scale the whole scene up/down it shouldn't make a difference.

Ernest Burden
09-24-2005, 02:53 PM
that seems most strange.

i wouldn't have thought that size makes a difference to render settings whatsoever. if that was the case, then this thread with my results would only apply to peeps using the scale of mm like me. all other scales would require totally different settings. that cant be right :shrug:

That is something we've looked at before. Simple to test--just scale your model by 1000 and rerender, compare numbers.

That's why I was asking many pages earlier in this thread about whether or not the GI settings were dependent on scale, noting that most of you work in metric while I mostly work in English.

lllab
09-24-2005, 05:09 PM
i thought its scale dependend, i sometimes loaded in models at 10 or 100 times to big and then all my gi settings didnt work as usual...

maybe its because of use of sky and floor objects?
anyway, the settings above dont produce fast images at6 my scenes, they do if i divide them by 10 though, but maybe its something else.

i a bit confused now, will check when there is more time...

cheers
stefan

STRAT
09-24-2005, 08:02 PM
That is something we've looked at before. Simple to test--just scale your model by 1000 and rerender, compare numbers.

That's why I was asking many pages earlier in this thread about whether or not the GI settings were dependent on scale, noting that most of you work in metric while I mostly work in English.

i have tested this before with no difference if i remember correctly, but yes, now it's arrisen again i'll test it and report back on monday.

STRAT
09-26-2005, 02:21 PM
i have tested this before with no difference if i remember correctly, but yes, now it's arrisen again i'll test it and report back on monday.

scaled the whole lot up and down a zillion time, with no render speed differences here.

andronikos916
09-27-2005, 12:33 AM
according to me to0 scale with AR2.5 does not matter... it is just for visual perposes and modelling reasons...

cy,
Andronikos

lllab
09-27-2005, 10:03 AM
yeas it seems to be just relatively relevant, it must be somrthing else then...hmmm

anyway i have my settings which do fine& fast for me, i was just wondering...

thanks,
stefan

STRAT
11-05-2005, 08:48 AM
Well, since the out come of this thread my regular gi settings have settled down into a more flowing routine. typically -

strength - 100 (i always use 100)
accuracy - 80
pre-pass - 1 or 2 depending on scene
stoches - around 1000-1200
diff depth - 1 (always 1, and use infill omnis if needs be)
min - usually around 50 (again, depending on model)
max - 500 (depending on model, but somewhere around 50% of the stoch value)

it is true that the higher the accuracy the better the render. by choice i'd use the above settings with a 99 or 100% accuracy. or even use stoch mode rendering.

i still go with the ethos that a 100% accuracy with lower samples is better than a 70% accuracy with higher samples, but the 100% option obviously takes allot longer to render.

but, and the nice thing with my 80% method, if any elements render with splotches with the above settings, all i do is plonk a render tag to the object forcing the gi accuracy up to 95% on that individual object, instead of globally upping the samples.

thank the Lord for render tags.

Ernest Burden
11-05-2005, 11:22 AM
thank the Lord for render tags.

Lord be thanked, and you.

What happened to your finding of a good target min of 85? And at much lower accuracy?

I've been setting accuracy to something like 15% and tagging the big stuff to 72-92%. But the single best thing I've found so far is to bake a few big, important surfaces. It's hard to build in the time to do it, and if you are only going to be doing one or a few views it may not be worth it, but for more views and expecially animation, it really helps on rendertime and eliminates flickering.

nycL45
11-05-2005, 11:33 AM
What kind of render time difference (+/-%) are you getting with this recipe vs. settings like:

GI: Strength 100%; Accuracy 40%; Prepass 1/1; Diffusion Depth 1; Stochastics 600; Min Resolution 100; Max Res 500

Has 9.5 made any difference with settings or speed?

Also, do you adjust the Ray Depth or Shadow Depth under the Options tab to improve r. time? One of your exterior renderings had a lot of glass, what kind of depth settings would be appropriate?

STRAT
11-05-2005, 01:27 PM
EBIII - depending on the model a min of 85 is a common figure for me these days. i'm finding mins of 50 and above the norm these days.

i still havent tried baking for anims yet. and probably wont soon yet either. i rarely animate, but i'm certainly placing high hopes with FR2. hopefully this will eliminate the use of c4d baking. if it doesnt, then as you say, selected baking may be an answer.

nycL45 - those render settings seem fine, except i'd change the accuracy to 70 or 80% or something. the render would be so much more realistic. even though the render times would be longer i'd be happy about this because i'd get better gi quality. besides, i got a couple of dual proc pc's, so personally i'm not to concerned with tweeking to this degree if it means faster rendering over quality.

has 9.5 made a render speed difference? it probably has, but it's neglegable, or i dont spot it at least. i was expecting a 200-300% speed increase orininally. how dissapointed was i ? ;) i think the main AA speed increases effect residently built c4d geometry. my meshes are 100% externally modelled.

as for depth settings? again, because i'm not overly concerned about finding every last drop of speed, i usually keep them as the defaults. works for me. but, if i got a scene with zillions of windows or opacity mapped trees for example which need higher depth settings then yes, i'll tweek until i get a correct minimum level.

nycL45
11-05-2005, 06:12 PM
... i got a couple of dual proc pc's, so personally i'm not to concerned with tweeking to this degree if it means faster rendering over quality.

Ha! I've got a full quad white mice rig here. Man, I have to be concerned.

I don't how much longer this thread will last but, it has been amazingly helpful. Post some more of those pretty pics, Strat. They're inspiring. If not pics, how about a lighting thread?

CGTalk Moderation
11-05-2005, 06:12 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.