Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Nice.
Very interesting comparison, thanks. Overall very nice work.
Very nice looking. I was under the impression (I'm still not sure), that if you wanted that GI bounce thing, and the ambient thing, you would need to have one UE2 for each of them?
By the way, an hour is an hour, and is to much time to wait for some things, lol. Final runs is a different story.
So it's a trade of quality vs not waiting bloody ages to see changes from adjusting something. That's understandable. That is curious about that 6x6 bit, as some have suggested going to levels I can't live with on that (I think it was 16x16, tho it was some time ago), so the 8x8 under 'sinc' was a compromise. I'll need to revisit that when I have a few days to sit and stare at the screen as the computer renders away with different settings. It brings up a good question I've simply not put into words yet regarding the samples per pixel (shading rate) vs that filter size. Let me compose my words on that question, and post it separately (It's rather technical, and may require a diagram or two).
And we all know the Daz Default is incredibly fast at some things, and lacking in so many areas at the same time. So I'm not all that surprised that even at 0.2 it's still a tad rough on the edges and only looks good "at or below" 0.1 shading rate, for reflections.
Quite a few of my map sets I've made, I do make a Specular mask for just that. Some times I end up putting it in Diffuse instead, so I can see it in the view field. That also brings up the biggest pitfall for AoA and Omni when setting up stuff, you can't see the tiling in the view field without doing a spot render, and there kind of slow at that. I don't need to do a spot render of the daz default to see the patron density tiled on a shirt (for example), and it renders without any delay at all for spot renders. The Omni appears to have a three second (minimum) pause every single time for the most simple of surfaces (No velvet, No SSS, etc), lol.
Shading Rate vs Sinc filter size.
The blue lines represents the actual pixel borders in the rendered image, the gray lines represent the subdivided pixel samples (1 / Shading Rate). so a Shading Rate of 0.25 would be four samples per pixel in a 2x2 array, or four samples wide and four samples height for 16 samples per pixel total (That's Q1, what is it)? And if it is not for each direction, how dose it divide up each pixel for the sub pixel samples (It can't be a simple Cartesian grid, unless some samples straddle borders between two pixels)?
We need not go into what a Sinc filter is, the code that makes them work, etc. I already know what a Sinc filter is. The question is with that X and Y width and height of that filter (the Filter "Window size"). Is it for taking all of those sub pixel samples and averaging (of a sorts) all of them out for each pixel in a render, or is it only applied to the final pixels completely independent of the number of sub samples for each filter?
These are very important questions, that I was not able to find a straight answer for, and are critical to know for setting up these filter settings without wasting an enormous amount of CPU power in the process.
This chart was a 'Guess' based on DSP filters used in my Amateur Radio Equipment. I know it is not accurate, tho it dose show how a simple Low-Pass filter can get out of hand very quickly with the amount of math needed to perform the filter operation. BTW, those "Total calculations per pixel" is just for the Low-Pass filter, and nothing else, lol.
A shading rate of 1 at 4.x4 is 16 S/p... it's total pixels (resolution) * filter/shading rate . So that's (800 * 1000) * (8*8)/0.01 = 5,120,000,000 samples. Just dropping that back to 4x4 means 1/4 the samples/time.
And for going above 6x6 in Sinc, the only time you should have to is when doing DoF or motion blur, and even then, 16x16 is way beyond insane.
Personally, I'd prefer to do stuff in post than try to get it to render just right. A .5 gaussian blur can fix some of those issues...
As for UE2, yeah, if you want bounce and ambient I guess that would be 2, but I have, so far, not found it necessary to have both. And the nice thing is I can shut off the UE and even shut off the arealights and use a simple distant light when setting up a scene and trying to get everything worked out. Then, for lighting, I usually hide all the complex stuff to get a quicker render to make sure things look roughly right, lighting-wise.
Which is mostly what I did with Iray, too, so the skills transfer. ;)
Here are a couple of renders....
UberEnvironment2:Occlusion w/Directional shadows: SoftBox Quality:4. Sphere set visible in render. There is also a single spot with raytraced shadows.
The cylinder has UberSurface applied....not going for fresnel or anything else just raw reflections.
The first image: Regular 3DL render, default everything except max bounce, sinc filter 4x4, shading rate 1. Surface settings, reflection strength 50% and blur 0% Time:24 seconds. Yes, it has jaggies...
The second image: Only change...10% reflection blur and doubling the blur sample to 8...Look Ma! No jaggies!
PS: Render time...26 seconds!
So, the default render settings are not crap...it's just that 99% of the time the surface shaders are not being used optimally. Fixing a reflection problem in the render settings makes absolutely no sense at all, unless one wants insansely long render times. But fixing it in the affected surfaces, usually has a very minimal impact on the over all render time and usually results in a better looking item, any way...
IPR.
Make adjustments on the fly and see them quickly in the render. Only old school SSS (precompute) and moving things around will trigger a re-render.
That's kind of a nonsolution. It's bad enough when trying to tuck a shirt in to a pair of pants, when smoothing is making the adjustment dial stutter. IPR just makes it impossible, lol. I honestly don't know if it is any better in Studio 4.8 or not, tho the last time I tried using it, the track ball stopped responding, and winamp crashed with a 'Time Limit Exceeded' error (followed almost instant by a blue screen of death). So, unless that IPR has been put on a Parallel-process leash, it just is not as good as advertised.
And that's the ridicules thing. I can watch vids while doing spot renders, with few problems (some times the video will stutter, not crash), yet when IPR goes into it's refresh progress bar thing, all hell breaks lose.
(EDIT) and before the silly question is asked. A lot of tutorials are in video form.
I'm kind of getting the sensation of translating the "Scientific Papers of James Clark Maxwell" again. I'm making a lot of assumptions of what in that formula is what, lol.
A shading rate of 1 at 4.x4 is 16 S/p... it's total pixels (resolution) * filter/shading rate .
S/p is Samples Per Pixel? in that the Sinc Filter X and Y control the "sub-pixel sample" count? (then what is Shading rate for?)
I had thought the "Shading Rate" controlled how many samples per pixel, and the sinc filter X Y controlled the Filter Window for combining those samples into the value for the pixel. So the optimum filter window would be the same size as the number of samples for that pixel. A smaller filter window would require more steps to combine all the samples, or discarding some. A larger filter window would require getting samples from neighboring pixels (possibly retracing light paths for them again), or repeatedly adding the border values in to fill the excess window width.
In this particular example the Filter window is to small to get all the samples for that pixel, and further steps would be needed to include them. Unrealistically, Studio limits the number of decimal points for Shading Rate, so a Shading Rate of 0.16666666666 (sixes for ever) for six sub pixel samples across is just impossible. Also a Shading Rate of 0.0277777777(sevens for ever) for 36 total sub pixel samples is equally impossible. Whatever method it uses, matching the Shading rate for a six by six sub pixel sample array for the 6x6 filter window is not going to be easy in Studio.
P.S. "16x16 is way beyond insane." was my thoughts as well given the suggested Shading Rate (0.1) in the example, lol. That's either a 10x10 array of sub pixel samples, or 3.16227766017 x 3.16227766017 sub pixel sample array depending on what the Shading Rate actually is.
That is promising, and discouraging at the same time. So it is possible to get smooth looking reflections without going to insane Shading Rate settings, so long as you don't want a crisp clear reflection. I'll look in to this when I get back from a coffee additive (half-n-half) store run. Thanks for the hints, and suggestions.
I've been really enjoying using arealights in 3DL to get some rather nice, realistic-approaching renders. Amusingly, arealights + PWToon is creating a wonderful sweet spot of 'realistic lighting with simplified/fast renders.'
I've been debating getting the AoA light pack, but it's a bunch of money and I'm not sure, at this point, how necessary it'll be -- if I want something fast and don't care about realism, distant light and maybe UE2 for a little touch seems sufficient. If I want really good, soft lighting that reflects, arealight + UE2 Bounce.
Given all that... how necessary are the AoA lights?
I just don't want to spend the money and then have it sit in the crate unused.
Given that the AoA stuff is not compiled shaders (at least the surface shaders are), I have little interest at this time. The AoA surface shader is so slow, it is near imposible to work with on some stuff and figures.
As for you, think of it this way. Is it on sale, and do you 'Know' you 'need' it. Cool toys are cool to get, tho sometimes what you already have may be better (or it may not). The tools you have were free, the other stuff is not. If you are unsure if you 'Need it' or not, then it may be just a case of not having enough experience with using what you have. That ultimately is why I have not gotten the AoA cameras and lights yet, I don't know enough about what I do have to be sure there is stuff I want to do that they cant and the AoA mey be able to.
Here is a good example, the AoA cameras vs what is included in Studio. I have a single time use once and not again need for a special render of my test chamber. I have not figured out how to make the Studio cameras do a longitude and latitude render (not a sphere inside a square, and actual Lat-Long render). I have not read anything anywhere that states that the AoA cameras can do that as well, so I keep plugging away with what I have.
Really, there are very few things that don't blur the reflections to some degree...but upping the blur samples and lowering the percentage of the blur still yields a significant savings and a 1 or 2 percent blur with 16 samples is often enough.
Which sphere has the blurred reflections?
The one on the right. (innocently humming). oh, it has diminishing multiple reflections before the other one.
It dose look good tho, very impressive.
I have the AoA atmosphere cameras. But I find it not super useful... except for the depth camera, which actually turns out to be very useful. So hey.
Yeah, unless someone really sells me on it, I'm inclined to make use of what I have and see if there is a sale some day. Events lately have me working at winnowing down my purchases to must-haves.
AoA lights are good...but are they must haves?
No. Between tweaking the Omnifreaker lights and imported shaders, I'm not finding a real need for them...but they do have some features others don't (although imported/from scratch lights can/do have the category feature that was one of their key selling points when they came out).
I think it's the AoA Advanced Spot...it has so many control features, but at it's heart, it's just a shadowspot. It's the controls that make it much more useful and different.
Hrm. Well, I have one big batch of purchases before I move my purchasing elsewhere... maybe I should. $25 isn't going to break the bank.
I just keep thinking of how AoA cameras got broken and worry.
What controls does the Advanced Spot have that stand out?
Yes, but it's only a tiny amount and upping the blur samples only added a few seconds to the time. It was like 5 seconds going from the default 4 to 16.
As to the samples/pixel, shading rates, etc...back on the old (old forums that no longer exist except in snippets in the Wayback machine), I think it was Richard or Adam that did all the math and great breakdown of it. Basically, at certain point, you are increasing the time to render all of those dramatically for very little gain. And this was before all the improvements to 3DL, which made those high samples/super low shading rates even worse.
The defaults (or a little higher for DoF/motion blur) and a shading rate of 0.2 was basically the 'best'...and yes, there were pixel compared images in that thread and going lower in the shading rate upped the time so much that the 1% or less differences were deemed not worth it.
And in some cases, it is much faster to use a shading rate of 1 and default values and render a larger image and down scale it in post, than it is to tweak the settings for the smaller image.
yep, and saying again, very impressive.
So, the answers to the questions I have, have been wiped from the face of the earth.
oy vey.
Way back on the old 2GHz dual-core CPU (No Hyper threading or fake cores), I did a brief test starting at 1.0 then going back to 0.9, 0.8, 0.7 etc to 0.1 Shading rate. I noticed a drastic sawtooth patron in the render times, that was a clear indication that the Shading Rate must be matched very carefully with something somewhere, or 3DL just falls on it's face. Thus ever since then, I've been asking some very simple fundamental questions, that apparently no one knows the answer to.
So I need to get my hands on a copy of the code for studio's interface, and look for the stuff that decides how many "sub pixel sample" are in a pixel, and see how the code determines that number from the value in the Shading rate variable. The code should tell me, if it is for the total samples in a single pixel, or just across in each direction X and Y. Likewise, The code will also indicate rather quickly if it is just an arbitrary value that is passed off to the 3DL stuff, and the answer is in that code somewhere. This is not the direction I wanted to go with this, tho for something as critical as proper values under the sinc filter settings to match the Shading Rate setting, it is starting to look like the only way to get the info "I Need".
I'm going to take five before doing any thing, as I have much to consider at this point. I'm sure the Shading rate code is no where near as complex as BSD's memory manager, tho I'm still not sure I want to go threw that again. Not to say the legalities of going threw code that is not my creation. Thing is, I'm not messing with the code, just looking for a simple answer to how it determines the number of samples per pixel. (OK, taking five and thinking about it).
You don't need to tie the various filter samples to a particular shading rate.
It isn't code in DS that you need to look at.
It's easy to see...render to a RIB...what is passed to 3DL is...
PixelFilter "box" 1 1
ShadingRate 1
PixelSamples 4 4
Format 640 800 1
DS doesn't do anything to it, it doesn't determine how many samples, etc. It just passes on the raw numbers.
Shading rate is the overall 'master' speed control. A fractional (decimal) shading rate acts as a multiplier...a shading rate of 0.1 means ten times MORE samples. A shading rate of 1 means the 'base' size multiplied by the samples (and it's basically how many times to oversample this pixel).
So it is total samples per pixel, not number of across the pixel. Thank you. As for the multiplying of shading rates, Yea, I was just reading something like that. I figured I better sift threw the 3delight manual before getting on Daz3d's bad side.
While it hints at a relationship between stuff, it is also not really clear on some stuff. Not to say some controls I'm not seeing in the Studio render settings tab ("Grid Size").
Bucket Size vs Multi-core CPU load.
I've posted else where, I'll just sum it up here. I've had some surface shaders trip over there feet and stop using all the CPU cores if the 'Bucket size' is set to small. On my 8-core CPU (8 real cores, no fake ones), I found a Bucket size of 24 to be a good balance to keep all the CPU cores doing something when working with such shaders, without it taking forever to get threw each bucket. Your computer may have a different optimum setting then mine, so testing is not a bad idea.
The relative shading rates are shader specified ones...like in the SSS shaders.
Just had a Duh! moment here...
Basically, if you want a Shading Rate of 4 for the SSS and have an overall Shading Rate of 0.2, then you need to set the AoA SSS shading rate to 20!
Yep...it speeds it up a little...shaved a few seconds of the precompute and a couple off the actual render for about 4 seconds less over all making sure that the shading rate for the SSS was actually 4 instead of 1 1/5 th.
But it is only a few seconds on a several minute render so is messing around trying to match it, worth the effort?
one and one fifth pixels, lol. yep, calculating values for those fractional pixels will get you every time, lol.
That's a good start on reducing the SSS precompute (Face plant) delay. 4 seconds off of a 3 minute figure is good. Tho for some of the forty seven minute Face plant AoA HD figures, I'll need a lot more then that before I stop dumping Daz Default shaders on them for setting up scenes, lol.
Well, if that is a few seconds every time it has to go threw it for a reflection off of something, not just the figure, it adds up. Tho as you hint at, it's nothing compared to horridly mismatched render settings that has the sinc filter redoing samples multiple times for each pixel.
The thing is, it's not something that 'scales', so it's going to be 4 seconds off a 2 minute render and the same 4 seconds off a 30 minute render. It is just cutting that figure/surface's time a bit. It's not a whole scene setting/savings. So you may pick up a bit more if you went and tweaked everything...but would it be any less than just rendering? If it were more or scaled up in relation to the time of render, then probably.
Re-iterating...
Pixel filter and pixel samples are not connected.
The number you see after the filter name is its filter width.
Pixel samples is a direct quality control. It controls the number of rays the renderer "physically" shoots for a single image element.
Filter width is more like "artistic quality" control. It deals with how the results of those ray hits are averaged together across the image.
Mike, you seem to have mixed up pixel samples and sinc width. You increase pixel samples for quality, but _not_ sinc width.
Sinc will RING.
It's a basic subdivided sphere over a plane lit with a basic distant light, rendered at sinc 16x16.
Now... If everyone would be so kind as to examine the edges of the shadow and the dark surface of the sphere?
See those weird white bands? Which should not be there at all.
This is called "ringing". Sinc will do this to your image at the slightest hint of contrast. These artefacts used to PLAGUE me. Then I forgot about sinc, and these artefacts went away with it.
16x16 is pretty extreme, but they start popping up as early as 6x6, just in less conspicuous places.
Another variable here is memory use. Smaller buckets (think 32) will render a little bit slower, but then the total memory usage will always be lower than if you use, say, 128. It's especially important for memory-intensive tasks like raytracing curves, and if you have 8 GB RAM or lower.
And of course, if you use a 128 bucket and the memory goes into swap, your speed gain may be lost.
I usually eyeball tiling by looking at the UV map layout of the surface (there is a "UV view" dropdown thankfully). No other DS shader but the default one seems to transfer their tiling settings to the OpenGL ones. I wonder if it's possible at all, and if yes, if it's documented somewhere because I sure wouldn't mind adding this to my shaders.
The good thing about UberSurface (as compared to AoA SSS) is that you can use those (ugly) on/off buttons to easily turn off whatever you don't need for test renders. And I have to thank Wowie for pushing me into using those buttons in my stuff, too; they're definitely handy. But ugly =)
It's not AoA's fault, though, because I don't think Shader Mixer has those buttons.
It can use gels ("cookies") and blur them natively in a nice way. Other than that, I'd say it's much like that dzSpotlight in the "DS Default" folder - the one that can use quadratic falloff (or decay, I don't remember the way it's called there), which is a _must_ for "physically based" non-infinite light sources ("infinite" = coming from infinity, "distant light").
It won't be broken because it's a legit new shader, independent of shader mixer.
And using the gaussian filter (which is there in the dropdown) is exactly what many pros would actually do (and do). Though they tend to render at higher resolutions than we "hobbyists" do. I personally have found that Catmull-Rom 2x2 setting, and I've stuck to it, and it gives me the right balance of sharpness and AA.
Not necessarily. The "Indirect lighting" 'environment modes' of UE2 will give you both. It's exactly what they are there for. You can use a colour or a map for your ambient then.
Wowie, I remembed you did a lot of testing of UE modes once... weren't there speed differences that favoured "indirect lighting with directional shadows" over the same with soft shadows?
Very cool! Is the flame an area light?
You know what else you could try adding to this scene, if you have LAMH? Some fur on the cat to glow in the light.