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
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
(Sarcasm detected) Yea, I'll second that, lol. At best it is like looking obliquely at water, eh, mixed about 10 to 1 ratio with fine carbon ash to the consistency of used charcoal Briquettes. I think a pyroclastic flow may have more mirror like reflectivity then the AoA shader can reproduce, lol.
It looks like environment mapped reflection is the only thing it can do well. So, until there is a way to render a lat-long map in Studio of your scene to drop into that map channel, that is a dead end.
I remember Kadix saying something about having difficulty in spaghetti land with ray-trace reflections, and ended up opting to not use it (some where in the Amazing Skins thread). So it may not be the fault of AoA, rather something in spaghetti land that doesn't work. Not that something being broken in spaghetti land surprises me much.
Aye, he put the function in there, it just doesn't work right..
I'm still working on coffee, tho I needed to start a new log spreadsheet, as the other one was getting rather long. And it dawned on me, I didn't really explain what all the columns were. And I was also trying to figure out a better way to represent the data to be les specific to my processor.
Left to right... 'M' is for the Minutes in the log file, 'S' is for the Seconds. That separation makes it a tad easier to add up for a single unit of time. 'MtS' (Minutes to Seconds) is simply the 'M' column multiplied by 60 to give me that unit in seconds, so it can be added to the 'S' column in the 'Ts' (Total seconds) column. In the former graph, I had simply divided the 'Ts' column by ten for plotting the pixels on the charts in the far right column (tenth).
I'm going to redo the former graph, as it really only applies to an 8-core AMD CPU running at 4.0GHz. So yes, I now have a new set of columns. I took the 'Ts' (Total seconds) column, and multiplied it by the number of calculators (8), and then by the GHz (4) to give me the Total GigaOp's needed to complete the render.
That last column on the right may change, depending how bad, bad is for the render GigaOp's. For the moment, I'm thinking a quarter or so of the Flop's may be good for the number of pixels in the graphs, and that is all that is. The number of pixels on the graph, and that column is not in stone.
So, the base line for the moment, is the total number of GigaOp's my CPU can do in half an hour of cranking away. Well, 30 minutes is 1,800 seconds. and that multiplied by 8 and then 4 is about 57,600 GigaOp's, or 57.6 TeraOp's. Yea, a "TeraOp" is a lot of compute power.
Op's as in Operations in plural, not to be confuse with the more common "Operations Per Second" (Ops). Time to top off the coffee cup, lol.
Oh, the test renders, are Progressive mode (to mimic spot renders while setting up stuff in the scene), and there 1024 x 1024 pixels in size.
So, after cleaning up the spreadsheet a tad (Yes it is saved with a different name), I noticed there already was a test with reflection vs SubD. I will do some tests and I'm not sure if steps of 5% in reflection strength will have the same surprise or not with render time (or GigaOp's). Here is the same old test values with new column sections, That was before I turned off reflection on the test chamber floor.
Oh, and one other note. I'm not trying to make it look like there has not been any improvement at all. Back in Studio 4.7, Face Plant Goddess had a face plant time around 47 minutes (that's no joke). That has been drastically improved to a mere 29 minutes with Studio 4.8 on the same processor. That's a reduction from 90.24 TeraOp's down to about 55.68 TeraOp's just to calculate the Sub Surface Scatter and nothing else. Ironically, once the Precompute delay had finished, it took les then ten minutes to render the test scene in both cases, an additional 20 TeraOp's, lol.
Connecting some dots, sidetrack moment. Before I disconnect from the internet to do some Reflection test renders.
Disclaimer. Beta G3F is only an assumption based on development cycles for everything else in the world. You build something, then you test it, make adjustments, then test it again. Some companies refer to the prototype product as 'Dog food' in a rather funny way. I do not know if there really was a Beta G3F distributed to PAs or not, tho I can assume that it was not made once and not adjusted and refined before the final release. So it is a given that somewhere, there may be a few variations of G3F before the final release version, and I'm guessing that somebody somewhere had to have tried it out for making figures and outfits. That would be part of standard testing practices, everybody dose it regardless of the market or product.
So in any case.
Teething pains for the first few months. I can understand that, as PA's get to grips with the nuances of the new G3F figure rigging and all. Even reshaping outfits as things progress, simply because it before launch of the final version. However, I have difficult believing that a beta version of the G3F shape would still be used as a base shape for making outfits over six months after the release of the final version to the public. That just doesn’t add up for a million reasons.
Auto-Follow being broken, I honestly don't thing that. Otherwise it would not work at all on any generation, not just G3F. However I will not say that there is not something drastically different on G3F that renders Auto-follow infective on most G3F figures. The lower base mesh density is a possibility, as some of the repeatedly problematic locations are rather low in mesh density (hip bone ridge, groin, lower but, etc). However not all of the locations that have recurring problems are that drastically low in mesh density (bottom and outer sides of the breasts for one). Besides, a defective Auto-follow program module can not explain why so many G3F outfits do not fit G3f when a shaping dial is held at zero, that just indicates that the base mesh of the outfit dose not even fit G3F without any add on shaping morphs at all.
G3F Shape not making it into a modeling program in tact. It would explain a great many things. The question is partly spawned by reading something in a thread that MJC1016 linked a few pages back. A majority of the shape of G3F is not in the mesh, it is mostly derived by joint bend smoothing morphs. If those morphs did not get transferred to the outfit threw Auto-Follow, then nothing would fit any figure on any generation when a joint is bent, and that is not the case. However it dose have me asking a simple question.
When a PA exports G3F for the mesh shape to make the outfit over in a modeling program, is there something in the shape of the exported obj that no longer matches the shape of G3F in Daz Studio?
It shouldn't matter...if things are being done at Base resolution...but if they aren't then the difference between subdivision algorithms could, possibly, be affecting things.
Blender has recently added OpenSubDiv...I'm not sure which subdivision algorithms it includes, but that could be something to look at.
I don't think SubD is to blame, especially for the Hold The Dial At Zero test that even outfits sold this month fail (or was it last month. 2016-01-31). Also, not all PAs are using Blender, and some of the programs are rather expensive (not that implies much other then the level of anger involved if it doesn't work, lol). Not to say, stuff made for Generation 6, actually works on G2F figures, unlike G3F stuff on G3F.
We are not talking few millimeters of shape mismatch, were talking a few inches inside G3F on the lower but, and a few inches of gap on the outer sides of the hips, with the hip bones sticking threw the mesh on the primary mesh lines on the hip-bone ridge. No way in hell can SubD be responsible for that, It has got to be something else.
Now is the G3F joint bend shaping morph being exported properly to all the modeling programs out there, that is the question?
Unless there is a gross misalignment of the fitting clone...
fitting clone? is that what you call the thing in your modeling program that you make the outfit over, and conform the shape of the outfit mesh to? Or is that the thing you use to set up the rigging jargon in studio after making the outfit?
Set up the rigging in Studio...unless you are a masochist and rig it completely from scratch.
That may be, especialy if it alters the shape of the outfit, not just where stuff is.
I just tried a crude OBJ export
And bring it back into studio, shows it to be the same overall shape (dancing interfearing surface artifacts and all), and even the higher mesh density is there in the number of faces. Hmmm. It was a thought.
Oh, what if the polycount is to high for the modeling program.
Well, the dark spots now don't dance around, however that is not inches off in shape either.
Fisty and Lilflame, plug your ears for two seconds, this dose not apply to you.
Daz, I exported an OBJ of G3F and reimported it and it matches the shape of G3F perfectly. I'll assume that is how a shape reference for making an outfit is done. I have purchased every single underwear set and swimsuit for G3F for the past seven months, Why do none of them work on G3F!!! without resorting to D-former origami.
Play around a bit with the various resolutions...G3F at Base and at High 0 SubD do NOT produce the same results!
And rigging the G3F 'suit' with the transfer utility gives even different results...and flipping the resolutions...even more when fitted. Also there is a bit of difference between full body, none and the various other templates in the transfer utility.
And whether or not Navel and Mouth Realism are on/off matters, too..
Remember several small errors, can and do, often add up to one much larger one...and to me, it's looking like several small ones are piling up. (actually more like compounding than simple addition)
But I can see that if the export base for the item of clothing to be built on happens to be G3F as she loads (hi res, subd 1, navel/mouth on) and the export for the morph being made at navel/mouth off, base res, when all is said and done there being a considerable mismatch between the final shapes.
Sorry MJC1016 for that brief outburst of frustration with Daz.
wow. I just used the G3F straight from the smart tab and didn't bother with any of the adjustment or actor dials when making the G3F OBJ test file. so just the default G3F, and the 'Base' resolution one, only appeared to have millimeter offset differences. An outfit sitting at silk thickness away from the figure and with smoothing on shouldn't have issues at that miniscule offset. However, when actually turning a mesh into a wearable item, that is honestly so far beyond me. About all I grasp from the rest of that, is that something somewhere is 'compounding' into a bigger problem.
Oh, if the Mouth realism and the Navel dials are changing the shape of the breasts and hips, then that is another big issue on G3F, lol. I can comprehend if the Navel thing causes and outfit made for it off, becoming a rather pointy issue, or a cannon ball hole. I guess whatever is used, at the end of the day, the outfit needs to fit G3F as she is loaded from the smart tab without any adjustments or extra dials. In other words, an outfit made to fit Victoria 7 in a modeling program, just is not going to work on G3F in Studio, lol.
Dog Food. The first batch of Reflection strength tests on the Daz Default shader. Aside from the 0% strength setting, there was not a significant amount of difference vs reflection strength. I'll run some further tests on Omni and AoA after some rest. (Face hits keyboard in exhaustion, lol). Zzzzz.
Mouth Realism is an HD morph...so it activates whatever process it is that turns HD into 'readable by other things' and does some other strange things when going through the various resolutions. And Navel is a morph that will double up things (and it slightly changes the abdomen shape).
I think I know _why_ it is happening, but can't see a way to fix it.
I applied the AoA Subsurface onto a sphere and exported it to RIB, and I found a sad thing.
The way 3Delight works, it listens to the commands DS sends when it describes the scene. And it turns out that DS tells 3Delight to always use the "shader" hitmode when lights calculate shadows from the AoA surfaces - so 3DL always runs the shader, even when the opacity is zero.
It's the command that the Shader Mixer plugin sends, and I don't know how to tell it not to issue it.
Are you sure it would not break many other things, and not be a 'fix' in the end result, lol.
Well, thinking of Spaghetti land in the simplest of terms. Each module has an input, and an output. And most of all, it is probably a protection against a Divide by zero, or keeping other effects working properly.
I'm guessing that there is not an option for a 'switch' built into a dial for the modules in spaghetti land. So something like a combined volume knob with built in power switch is not there. Besides, such a switch would need a 'output' to allow a bypass to be made for some things when the module is in it's off state.
One example I can think of for opacity, is that distortion trick I used for that SORH scene. It was a simple multi-face plane placed at the location of the staff, and a few d-formers put on it to distort the plane. That's how I made that gravity well distortion effect at the ends of the staff.
The shader needs to remain active, simply because it is distorting the view of the stuff behind it. Similar can apply for transparent surfaces that are reflective and such.
P.S. it was a bit of work to get the d-formers lined up with the view from the camera, tho the results were worth it.
http://www.daz3d.com/gallery/images/41199
The hitmode is set by a separate command - a rendertime script. It's a different thing from a shader, and it's DS-specific.
I know my way around "normal" rendertime scripts, for RSL shaders, but the ShaderMixer plugin apparently generates its own scripts on the fly, same as it creates the shader code.
I did some tests, and it looks like that whenever you plug any brick into the opacity slot, it invokes the "shader" hitmode. So just un-sticking whatever spaghetti is hanging there from the opacity handle, it could help...
OK, so the 'hitmode' thing dose the is it worth rendering or not for each surface, and the shader parts do not.
Also I worded the time thing slightly wrong with the opacity. I'm not sure exactly what the mechanism is, tho apparently there is more math/steps involved with mid 'intensity' values of opacity then the extreme 'intensity' values. 95% opacity and 0% opacity consume more then 100% opacity (Daz and Omni), however values between that takes even more. It appears that when the opacity reaches 0% opacity, something happens in the Daz and Omni shader to take less math/steps then at 5% to 90% opacity.
The AoA being the exception and dose not 'back off' at 0% opacity. None of the shaders actually turn off, as the times are all more then the same environment and camera angle with no cylinder at all. It is completely unexpected results for opacity render times, as everything other then 100% opacity and 0% opacity, should be calculating the same stuff (front face of the cylinder, distant face of the cylinder, and the floor behind). So, whatever is going on inside the shader code, apparently Opacity is an extremely complex beast.
Reflection, appears to be rather simple. My assumption that some values (25% 50%, 75%) are faster then others, is completely wrong. The only value that is measurably faster, is 0% intensity. And all the shader behave about the same across the board. Tha AoA dose take longer then the other two, however the behavior between values is still the same 'envelope' for the shape of the times.
So, over all, reflection is not all that bad, unlike opacity. Now, granted Opacity is rendering both sides of the cylinder, both reflection and Opacity are also including the environment. Time to plot the reflection numbers so I can compare the charts.
I know it is not funny for many, I just laugh because it is not very often I get spam from someone impersonating me, I had no idea I was so popular. My email address is not zarcondeegrissom@Earthling, nor have I ever done business with EarthLink.
There was a brief time, that I had a MSN and hotmail and AOL accounts just for the instant messengers. That was over fifteen years ago, and they all elapsed because I never used them for email stuff. I also never got much time to rest because of having so many IMs going with people IM-ing me all the time, and that is why I'm not all that big with 'social media'.
LOL I have an AIM account that I log into every day, it used to exist to talk to my husband in our first year of relationship when I still lived 1/2 way across the USA. Now it continues to exist to talk shop with Fred. I have a Skype account that I only ever log into if I have to hunt down a PA rep urgently quick or AIM isn't working that day and I need to ask Fred something. (I think like 3 people in the whole world know that AIM account)
Last month sometime...
Husband: "Why don't you get (my one wonderful beta tester's name) on AIM so you don't have to send one sentance long emails back and forth?"
Me: "Because then I wouldn't get anything done for him to test.
That is the same reason I tell my siblings why I don't get a facebook account, I'd never have any time to get any thing done, lol.
For MW_HNL.
Yellowstone and Iceland, as for Hawaii? There has got to be something somewhere?
http://news.berkeley.edu/2015/09/02/ct-scan-of-earth-links-deep-mantle-plumes-with-volcanic-hotspots/
Well, not very detailed, tho it is kind of something, lol. I guess there to busy keeping track of what the volcanoes are doing, and understandably so. Kilauea has only been erupting nonstop since 1983.
I admit, I took yesterday off to clean up the house a tad, then play with some shaping dials.
I cropped out all dials that are automatically linked to the included figures (I didn't fuss with them), leaving just the dials I did fuss with. Some V7, Gia7, Fwsa Paloma, and Fwsa Uma. Then a few more dials (Growing up for G3F)in the mix to fine tune some features a tad. Granted I did use some of the growing up dials, she is not a child by a long shot, nor dose she have a gas giant planet class bra size, lol. Well, there is a tad more I want to do to the face, however that was the brief bit of fun I had before plotting out a new graph.
Reflection compared to Opacity for the different shaders.
Oh, ahhhh, the claw. lol. Well, the shape of the reflection graph is a tad more like what I expected to see with opacity, however there apparently quite different in many ways. Also, I suspect that IF (stressing the "if"), reflection on the AoA shader is ever fixed to be more like a mirror rather then used charcoal, the times will change drastically for that. Omni is about on par with the Daz default, averaging around fifteen seconds longer for Omni then the Daz default to do Reflection of any strength (That's a hell of a lot better then I expected it to do). One other note, with the Opacity tests, and the Reflection tests, the Face Plant times of all the shaders was less then a second without any SSS in the mix.
I'm still trying to figure out a definitive way to clock Face-plant time of the shaders, without adding load on the CPU running the test renders.
I'm sure my current method of measuring that is not going to give many a lot of confidence. It requires me not blinking the instant the render starts to do something and noting the exact time in the progress bar. Some of them I missed it, and had to cancel the render and do it again, making for a very long set of tests. From all the tests I've done so far with SSS, the only thing I know, is that the AoA shader is consistently five times slower then the Omni shader when percomputing the Subsurface stuff.
P.S. ignore the 0.7mm used graphite stubs over on the left, I was making electrical "Conductive Paper" for doing some tests not related to anything Daz.
(EDIT)
Sitting here looking at those sticky notes, I do see a patron. The Onmi shader goes from 10 or 11 seconds to around 30 seconds Face-plant time between Base resolution and high resolution on the SubD settings. A three fold increase in Face plant time. AoA goes from about 105 seconds to 161 seconds, a jump far less then three times grater. If going from "high resolution" to HD has a similar jump in Face plant time, that may begin to explain Some of the intolerable precompute delay of HD figures, tho not completely. It is a huge difference between half an hour of precompute delay and a mere two minutes and change (or 30 seconds for the Omni shader).
At this point, I'm sure something incredibly bad is going on with the HD mechanism in Daz Studio, I just don't know what it is. I have already ruled out it being just a single thing, like polygon-count, SubD, or multiple faces on a single object. I'm guessing it is some combination of stuff that dose not play well together. Also, I've yet to look at the effects of maps on the total render times and SSS precompute delay.
You know. In a way, I can almost understand why so few G3F items actually fit G3F figures. Especially when the PA making the fit morphs needs to wait over half an hour before a spot render begins to show anything at all, lol. Oh that thing (70's Gym Uniform for G3F), that was a 2015 Christmas Giveaway item, and has already been yanked from the store. Also, I honestly don't know what Mina's face plant time is, I never tried it, I saw the AoA and instinctively drooped in a Daz Default alt shader using her maps without a second thought.
Oh, and I do use the AoA and Omni SSS shaders for final renders that I post at DA. I kind of just reserve them for renders I can let run while I'm sleeping, so I'm not sitting here watching my computer do nothing when I'm awake.
Not really drastically, fresnel() should be a fairly efficient shadeop. What you could do to truly level the playing field is to set the DS default one to use refraction index of 1.3 (no refraction strength, just the index value), it will do Fresnel attenuation then, just like AoA. There´s not real equivalent in UberSurface because it uses a custom Fresnel curve, but I'd say that if you merely enable Fresnel with reflection in UberSurface, then it will eat the same amount of resources regardless of settings.
That's not a reason.. you've seen my work in progress renders, I use a G3 without any texture maps, just a brown diffuse color, and the clothes/jewelry are plain white. All basic dz-default surface, my 900x900 test renders take like 20 seconds at most, even with full HD. And you don't have to render anything at all to fit clothes, the only thing you have to render before making matfiles is testing that HD morphs don't mess things up, and that's really a last step.
That was only in render, it looked fine in the view field, lol. Understood tho, for the most part things are the same both places.
I looked around a bit, and I'm not seeing much for improving 3DL performance at the hardware level. (there in 'Time', lower is better)
http://www.tomshardware.com/charts/cpu-charts-2015/-22-3DS-Max-2013,3713.html
http://www.tomshardware.com/charts/cpu-charts-2015/-23-Blender,3714.html
So, it looks like I need to scrape up a few thousand dollars if I want a faster CPU then the one I have that cost around a hundred dollars, lol.
Also, I'm curious about something. Rendering. For years I thought that was the process of calculating light bouncing around a 3D virtual environment to produce an image. Recently when I started digging into massive multi-socket computers, I find quite a few video bloggers calling splicing recorded videos together, "Rendering". Some of the stuff shown looks more like "video editing" or "splicing", not "Rendering". Is this the new fad, like 'Engineers", it is such a 'cool' thing that every one wants to call it that, even when what is going on is not what there calling it?
Nope, it's okay - to create a video with all the FX etc has long been called "rendering", too. Even in 2D image processing, some packages use the word "render".
It seems to be a word that is older than CG - it has existed for a few centuries: http://www.merriam-webster.com/dictionary/render
It's also used in to cooking for simmering something.
Being rendered speechless ...
Yea, I do remember something from a very old 'Battlefield' documentary series, where the narrator would use impartial phrases like "They were reduced" or "Rendered ineffective".
Sorry about my absence the past few days, I was dealing with bills and other stuff. I did take a few minutes to look at a idea from a few months back.
(they said what about my niece. lol.)
Still needs a lot of work, and quite a few things I just don't have the ambition to make at this time (Ankh, Anklets, khopesh/polearm). I'm also not sure what I'm going to do with the bracelets as well. I wanted it to be friendly with tillable shaders, however the layout of the UV mapping doesn't allow that without adding geometry for the beads in the band. The above renders was with a simple glass-like shader setting with no maps at all, and it came out better then I expected it to.
I've also been looking at a PCB and trying to figure out the best type of Shim/heat-spreader for the chip layout on the thing. I found a basic STP file with the chip heights and locations, however it appears to be missing some SMT components that may be in my way of making a shim.
So that is on hold till the PCB arrives for me to make some measurements. The STP file is of little use to me, so I'm going to need to make a new one from scratch. It looks from photos that the heights of some things are accurate, however the rest looks way off to be using for mapping out shims that need to not touch other things.
As you can probably guess from the marked up photo, I'm thinking about a plastic shim (blue), to hold copper blocks in place. So yes, this is something I want to see in 3D, especially if I can render it in Studio from different angles, that is the idea at least. Chips A (CPU), B (Power), and C (North/south bridge) are all different heights, and to add some fun, the caps around B are taller then the chip.