3delight tests

Plugins linking to this thread: (hide)

Blob StrokesAuthor: Anto Matkovic
Here's a small ICE compound used to generate these 'blob-strokes'. I had something like this before, but I really wasn't so excited by the speed of generation. Now with 3delight, it's definitely feasible, updates are performed immediately. The compound creates particles on curve list, evenly distributed, count is relative to particular sub-curve length (still, you'll need to take care of curve re-parametrization). It's also able to perform 'write on' on one-after-another subcurve (by sub-curve, or normalized by sub-curve length). In short, you can 'draw' 3d strokes with it, by animating the slider value. There are a few additional parameters. [..] contd. under si-community link

local backup: Emit Blobs From CurveList.xsicompound
3DelightAuthor: dna research
v4.0.41 updated Feb 21st 2015. 3Delight is a fast, high quality, RenderMan®-compliant renderer designed to produce photo-realistic images in demanding production environments. The renderer was introduced to the public in the year 2000 after being used for more than a year as the sole renderer in a sister production company. It is now widely used and earning a reputation as a benchmark in rendering technology. [..] (cont'd on product page)

DNA research recently released the 4.0.41 upgrade along with making the free license now support up to 8-core CPUs. Highlights from the changelog: Volumetric Smoke and Shards Any shader as light source projection Exposure, Gamma Controls Mesh Light support ESC to stop render. Last year's big 4.0 update introduced features such as: support for deep EXR files Native MARI textures support Improved sampling of environment maps Added the physically plausible 3Delight Material "Continuous rendering" is now enabled by default Up to x2 acceleration in hair/fur rendering Memory usage is down 30% when using the path tracer Skin shader for 3Delight for Softimage Motion Map property is now supported

A list of movie project rendered with the 3Delight renderer. As mentioned the product is free of charge for the first 8-core license. Additional quad-core licenses are available for 400$ each and unlimited multi-core licenses for 900$ each. Yearly support and updates 190$/290$ resp.


Comments on work in progress
Falam

Re: 3delight tests

Post by Falam » 18 Mar 2013, 03:20

Maybe I've been misinformed, maybe I'll consider buying a license :)

Matt which license did you buy a Quad License ?

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia
Contact:

Re: 3delight tests

Post by Mathaeus » 10 Aug 2013, 21:22

Just an assembly of things I already had. It's very old school render, about 50 spot lights with DSM, third of that only for making a window light. Point-based color bleeding on top. In short no any ray tracing, at all. Everything took about 2.5 hours to render, on quad-core machine.
Tried new path tracing, too. Hopefully all that mass fitted in available 8 gigs of memory, but didn't have a patience to wait for final. That grainy preview really doesn't tells that much when it comes to fine details, like hair.
What's tested here.... probably Sinc filter (available for ages). It is a very sharp filter, but it also gives a slight bloom effect when it runs over bright pixels (even without DOF or like), because of large area considered for oversampling.

Image

User avatar
gustavoeb
Moderator
Posts: 587
Joined: 21 Jul 2010, 00:33
Skype: gustavoboehs

Re: 3delight tests

Post by gustavoeb » 10 Aug 2013, 23:08

very impressive, strand master! :D
Gustavo Eggert Boehs
Blog: http://www.gustavoeb.com.br/

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia
Contact:

Re: 3delight tests

Post by Mathaeus » 21 Aug 2013, 01:29

gustavoeb wrote:very impressive, strand master! :D
Thank you kindly. Just to clarify, rubber ducks were created using the rare technique called 'box modeling'. Strange some method...
Now seriously, this is one should be more artistic, including inevitable DOF-ed object in front of camera. This time I put complete GI on everything, including hair. In first pic hair was only visible to other objects. Removed a lot of lights, now there's about 30. Half for a row on floor, to simulate the primary bounce. Render time jumped to about 3.5 hours on QuadCore, slowly walking around all that hairs. But it looks more consistent, imho.

Image

Kzin
Posts: 432
Joined: 09 Jun 2009, 11:36

Re: 3delight tests

Post by Kzin » 21 Aug 2013, 21:02

looks great sofar.
could you test the newer path trace stuff on that, how long is it rendering?

Moderator edit: removed double post - HB

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia
Contact:

Re: 3delight tests

Post by Mathaeus » 22 Aug 2013, 00:32

thank you. This is rendered with 3.1.39, current is 3.1.49, probably it's good idea to wait a bit. Maybe there's someone around, having a stronger machine, than my home QuadCore.

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia
Contact:

Re: 3delight tests

Post by Mathaeus » 17 Oct 2013, 01:30

This is probably last in REYES row, I hope next one will be solely path trace. Just wanted to try it with very dark hair.
There's about 40k hairs for main hairdo, about 250 segments for each hair, that's about 11mio segments for everything - rest of hairdos has relative small weight.
Created especial curls compound to take the more precise shape from less segments. It does turn by segments, one turn always has the same number od segments, about 6-8. Render time is around 2.5 hours on QuadCore machine, using 3dfSI 3.1.39.

Image

User avatar
McNistor
Posts: 605
Joined: 06 Aug 2009, 17:26

Re: 3delight tests

Post by McNistor » 24 Oct 2013, 20:16

Looking good Mathaeus!
Can you please shed some light on whether the path-tracer is as fast as the reyes when it comes to displacement, dof, mo-blur and other things which a reyes based render shines at?
The society that separates its scholars from its warriors will have its thinking done by cowards and its fighting done by fools.
-Thucydides

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia
Contact:

Re: 3delight tests

Post by Mathaeus » 25 Oct 2013, 01:07

Well these setups for lights and point-based color bleeding, wants a lot of lights, about 50 here, as used color bleeding allows only one bounce. Not fast to do, not for everything, completely arbitrary in this case. While path tracing equivalent needs only one or two area lights, then it's path tracer to do rest, realistically. Setup time should be counted too.

Btw, finally there's ray traced subsurface scattering in 3delight, should be possible to get the fast preview of all mentioned, in short time. Six months old 3dl version I've used here, has only point-based subsurface scattering, with unpleasant pre-computing delay. But this is past,
hope I'll put something with all that new stuff.

User avatar
McNistor
Posts: 605
Joined: 06 Aug 2009, 17:26

Re: 3delight tests

Post by McNistor » 25 Oct 2013, 12:12

I don't understand what you're saying to be honest.
Let's consider the following case:
what engine would be more useful (path-trace or reyes) for an animated character with displacement maps, SSS and mo-blur activated?

My question could easily be nonsensical because I barely tried 3delight a long time ago, but if you understand at least in part what I'm addressing here I'd really appreciate you giving another try in answering it. :)

edit: My guess is that you read shed some light literally. To shed light on is an expression which means to clarify something.
The society that separates its scholars from its warriors will have its thinking done by cowards and its fighting done by fools.
-Thucydides

Kzin
Posts: 432
Joined: 09 Jun 2009, 11:36

Re: 3delight tests

Post by Kzin » 25 Oct 2013, 13:11

path tracing is slower, needs more ram and you have the noise/quality tradeoff.
it cant be that fast then reyes because like the name says, it renders only what the cam sees in the bucket which is currently rendering. it does not care about the things outside of the actual bucket. pt is completely different, takes alot more renderpower and solves slowly. but thanks to all the crazy amount of processing power which the big studios have they switch to pt or bidirectionel pt. it saves setup time, thats right, it renders more realistic out of the box with modern implemented shading models but the downside is the extrem long rendertimes and memory usage compared to reyes.

if you need fast renders with alot of dispmapping, mb and perhaps dof, then reyes could be a better choise when you dont need gi. but with point based color bleeding, which is really fast but not accurate as path tracing, you can also render with gi and, most important, noisefree.
noise is the big problem here for a path tracer, especially with alot of indirect light. bidirectionel path tracing can help alot in such cases. of course, all of this does not cover caustics, for this exists better methods.

User avatar
McNistor
Posts: 605
Joined: 06 Aug 2009, 17:26

Re: 3delight tests

Post by McNistor » 25 Oct 2013, 13:54

Thanks kzin for your input.
However, are you talking about 3delight or raytracing vs reyes in general? If that's the case, I already know about the trade-offs when choosing one or the other and I was more interested to know if 3delight changes or brings something new regarding all these well-known facts.
The society that separates its scholars from its warriors will have its thinking done by cowards and its fighting done by fools.
-Thucydides

User avatar
Mathaeus
Posts: 1778
Joined: 08 Jun 2009, 21:11
Location: Zagreb, Croatia
Contact:

Re: 3delight tests

Post by Mathaeus » 25 Oct 2013, 16:17

McNistor wrote: If that's the case, I already know about the trade-offs when choosing one or the other and I was more interested to know if 3delight changes or brings something new regarding all these well-known facts.
I don't know, honestly. Tried some path tracers, but didn't used any of them for something serious. For now, simply there is no reason to do that.

Kzin
Posts: 432
Joined: 09 Jun 2009, 11:36

Re: 3delight tests

Post by Kzin » 26 Oct 2013, 16:54

McNistor wrote: more interested to know if 3delight changes or brings something new regarding all these well-known facts.
no, it does not. it has the same problems like arnold for example with some things arnold can handle better actually, like the clamp thing.

User avatar
McNistor
Posts: 605
Joined: 06 Aug 2009, 17:26

Re: 3delight tests

Post by McNistor » 26 Oct 2013, 16:55

Ah, I see.
What clamp thing,
is it the equivalent of color clipping in Mentalray?
The society that separates its scholars from its warriors will have its thinking done by cowards and its fighting done by fools.
-Thucydides

Kzin
Posts: 432
Joined: 09 Jun 2009, 11:36

Re: 3delight tests

Post by Kzin » 26 Oct 2013, 17:04

clamping values is needed for really high energy pixels which are hard to solve when only a small amount of pixels in the scene have this energy and reflecting on surfaces for example. you would need a hell lot of sample to sample such small areas and solve them. its easier and result and better solved images when you clamp suxch values to a max value.

in vray its the color mapping stuff, in arnold the clamp option and in mr together with mila shader the clamp string option.

Post Reply

Who is online

Users browsing this forum: No registered users and 32 guests