Indigo Renderer 4.0.43 Beta Release

General News and accouncements regarding the Indigo render engine
User avatar
Jay-ko
Posts: 138
Joined: Mon Nov 12, 2012 7:28 pm
Location: Paris

Re: Indigo Renderer 4.0.43 Beta Release

Post by Jay-ko » Fri Aug 19, 2016 8:14 pm

Hi,

It is not critical, but I found a new stuff yesterday. I use the RanchComputing renderfarm.
They use 3.8.23.
So I have to use this version on my workstation to be able to create 3.8.23 pigs files.

With the installation of 4.0.43 in an other location, the pigs files from 3.8.23 create background material and probably material with no "texture_index".

So to correct that, I had to re-intall 3.8.23 (same location of my original 3.8.23) to keep the creation of my pigs compatible.

That prove that 4.0.43 use the same parsing rules files even if you installed it at a different location.

User avatar
Originalplan®
Posts: 757
Joined: Sat Sep 26, 2015 8:58 pm
3D Software: Cinema 4D
Contact:

Re: Indigo Renderer 4.0.43 Beta Release

Post by Originalplan® » Tue Aug 23, 2016 6:05 am

What number??? 3.14??
Attachments
Screen Shot 2016-08-22 at 8.02.26 PM.png

User avatar
OnoSendai
Developer
Posts: 6241
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Indigo Renderer 4.0.43 Beta Release

Post by OnoSendai » Tue Aug 23, 2016 6:26 am

Originalplan® wrote:What number??? 3.14??
The magic number. It means your IGI is corrupt/invalid.

STUDIO AARON
Posts: 24
Joined: Wed Aug 13, 2014 12:39 pm
Contact:

Re: Indigo Renderer 4.0.43 Beta Release

Post by STUDIO AARON » Wed Aug 24, 2016 3:50 am

I have an infinite ram usage issue when I enable OpenCL rendering on a rx480. (skindigo 3.8.33.1)
I was not having this issue with an older video card.
Tested a couple different skp files.
Video card drivers are up to date.

TonyD
Posts: 49
Joined: Sat Aug 20, 2016 2:24 pm

Re: Indigo Renderer 4.0.43 Beta Release

Post by TonyD » Thu Aug 25, 2016 7:24 am

Hello!
Just wanted to say that there is noticeable progress on the OpenCL materials conversion/normals mapping front.

I am seeing black artifacting on hair and lashes, but it looks more reduced with each update release. Sometimes I can add a Normals map which will remove the black boxiness, other times I can't figure it out.
Can bypass it by using other render modes, but OpenCL is SO much faster, I would use it exclusively, very interested to see it advance!
FSB 3 PathTracingOpenCL.38.jpg
Here was on 4.0.38 on OpenCL, noticeable artifacts on hair and lashes
FSB 3 PathTracingOpenCL.jpg
Here is on 4.0.43 on OpenCL, artifacts on hair and lashes getting better...
FSB 3  PathTracing.jpg
Here is on 4.0.43, the Dream for OpenCL!
no artifacts on hair and lashes when just regular Path Tracing...
Sparkle.OpenCL.43.jpg
This is a different one on OpenCL, renders great! (Some blocky shadowing, but that appears without OpenCL on the other render modes also)
Anyway, great work! Looking better each release!
You guys rock!

-Tony D

TonyD
Posts: 49
Joined: Sat Aug 20, 2016 2:24 pm

Re: Indigo Renderer 4.0.43 Beta Release

Post by TonyD » Fri Aug 26, 2016 1:19 am

On multi GPU, I can share my general experience with OpenCL.

I have a rig with 1 Titan X and 6 Titan Z (I use Amfeltec clusters), which equates to 13 GPU.
-My Titan X (12GB VRam) is attached at 16x PCI via motherboard, and
-each of the 12 Z-core (6GB Vram) is attached at 1x PCI via splitters.

In using 2 other Cuda render software, I can achieve a linear increase in render speed (performance % per-GPU drops slightly as more GPU are used), however speed is overall increased.

In case of Indigo OpenCL, here is pattern I find- I have tested a found similar pattern on the 2 test files as well as my own.

For argument sake, lets say I can render 'Scene X' w/Titan X by itself @ 14.000 M samples/sec.
If I deselect Titan X and activate:
2 of Titan Z cores (=1 Titan Z card), I will almost match @12 M samples/sec
3 of Titan Z cores (=1.5 Z card), will get @18 M samples/sec and beat the Titan X
4 of Titan Z cores (=2 Z cards), will give me @20 M samples/sec and is the 'sweet spot', fastest render time
5 of Titan Z cores (=2.5 Z card), will get 18 M samples/sec, only matching 3 Titan Z cores
6 of Titan Z cores (=3 Z cards), will give me 9 M samples/sec, less than the single Titan X
7-12 of Titan Z cores, will give me progressively lower scores, respectively, < 9 M samples/sec
(BTW 11-12 GPU crashes Indigo, sometimes...more unstable.)

(This was just an example, ~20 Ms/sec is by no means top time I saw yet. I have one less complicated scene which gets 90 Ms/sec with the 4Z sweet-spot, whereas X alone got me ~65 Ms/sec.)


As such, there is a 'diminishing return' effect after 4 GPU cores in my case, for OpenCL.

Also find that Titan X with 12GB Vram opens and closes scene on OpenCL faster that the 6GB Vram Titan Z, so the "frame completion time" when rendering a sequence of frames is sometimes faster (this depends too on resolution, desired samples, etc.). For example, I had a situation where a frame would render from start to finish with 4Z cores in 39 seconds, whereas the Titan X did each frame in 29 seconds, even though the 4Zs had the faster "render time". Possibly, the greater # kernels/lower Vram possibly contributed to a slower per-frame .igs open and close. In any event, the Titan X "frame completion time" was faster by 10 seconds. 10 seconds x 8000 frames would be about 24 hour difference in absolute render time.

Just wanted to share and am interested to hear if development may later yield a more linear achievable speed with added GPU, such that more GPU does not at a point negatively congest the flow.
Regards!
-Tony D

User avatar
CTZn
Posts: 7240
Joined: Thu Nov 16, 2006 4:34 pm
Location: Paris, France

Re: Indigo Renderer 4.0.43 Beta Release

Post by CTZn » Sun Aug 28, 2016 11:50 pm

There's an issue with the if statement.

Code: Select all

Error while building scene: Error occurred while parsing function 'clouds': Expected identifier 'else'.
buffer, line 16:
	range = (if(useBox==1) ((pos.z - bottom)/offset) else ((D - bottom) / offset))
	                                                                             ^
Indigo version: Indigo Renderer v4.0.38, Windows 64-bit build.
as in

Code: Select all

 	useBox = paramBox()==1

	bottom = paramBottom()
	top = paramTop()
	offset = top - bottom

	range = (if(useBox==1) ((pos.z - bottom)/offset) else ((D - bottom) / offset))
I'll edit this post after I have tested the if(a,b,c) form.
edit: maybe I've been given an advice before on that issue but I don't remember which...
edit2: this is the blend shader from http://indigorenderer.com/forum/viewtop ... 68#p130468
edit3: when I have seemingly reverted all if statements, I got this confusing message:

Code: Select all

Failed to find function 'l111l1ll11l1l1lllllll1ll11l11llll(int, vec3, float, float, float, float)'.
ShaderDisplaceMatParameter user shader code, line 61:
	l11ll1ll1llll11ll11ll11ll1l111l11 = (l111l1ll11l1l1lllllll1ll11l11llll(0, (ll1111l1l11lllll1111l1l1l1l11l11l * 0.35f), 0.f, 4.f, 8.f, ll11l111ll1ll1l1lll11l1ll111lllll(ll11llll1l11l111llll1l1lll1l1l11l)) * 0.1f)
wich seems to originate from my way of using the multifractal function. Does it still exist ? This is with 4.0.38 sorry about the mess.
Last edited by CTZn on Mon Aug 29, 2016 12:19 am, edited 1 time in total.

User avatar
Oscar J
1st Place Winner
Posts: 2204
Joined: Sat Mar 31, 2012 3:47 am
Location: Gothenburg, Sweden
3D Software: Blender

Re: Indigo Renderer 4.0.43 Beta Release

Post by Oscar J » Mon Aug 29, 2016 12:03 am

TonyD wrote:
Just wanted to share and am interested to hear if development may later yield a more linear achievable speed with added GPU, such that more GPU does not at a point negatively congest the flow.
Regards!
-Tony D
Yeah, looks like they have some work to do in this aspect. It's already supposed to be pretty linear, but It guess it hasn't been tested with that many GPUs.

User avatar
arc en ciel
Posts: 138
Joined: Tue Apr 12, 2016 2:33 am

Re: Indigo Renderer 4.0.43 Beta Release

Post by arc en ciel » Mon Aug 29, 2016 1:12 am

CTZn wrote:There's an issue with the if statement.
this statement works

Code: Select all

		if(	belowClouds,
			cloudless,
			clouds
		)
this is probably what you call the if(a,b,c) form, isn't it ?
CTZn wrote:I got this confusing message:

Code: Select all

Failed to find function 'l111l1ll11l1l1lllllll1ll11l11llll(int, vec3, float, float, float, float)'.
ShaderDisplaceMatParameter user shader code, line 61:
	l11ll1ll1llll11ll11ll11ll1l111l11 = (l111l1ll11l1l1lllllll1ll11l11llll(0, (ll1111l1l11lllll1111l1l1l1l11l11l * 0.35f), 0.f, 4.f, 8.f, ll11l111ll1ll1l1lll11l1ll111lllll(ll11llll1l11l111llll1l1lll1l1l11l)) * 0.1f)
wich seems to originate from my way of using the multifractal function. Does it still exist ? This is with 4.0.38 sorry about the mess.
that happen in GPU mode only, right ?

User avatar
CTZn
Posts: 7240
Joined: Thu Nov 16, 2006 4:34 pm
Location: Paris, France

Re: Indigo Renderer 4.0.43 Beta Release

Post by CTZn » Mon Aug 29, 2016 8:08 am

Yes that's what I call the if(a,b,c) form, and it does work.

No, I have not updated my CG board yet. It manages GPU Acceleration but not OpenCL rendering, that was all on CPU's. I have kind of fixed it by reverting to fbm's instead of multifractal...

Merci !

User avatar
OnoSendai
Developer
Posts: 6241
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Indigo Renderer 4.0.43 Beta Release

Post by OnoSendai » Wed Aug 31, 2016 4:07 am

CTZn wrote:Yes that's what I call the if(a,b,c) form, and it does work.

No, I have not updated my CG board yet. It manages GPU Acceleration but not OpenCL rendering, that was all on CPU's. I have kind of fixed it by reverting to fbm's instead of multifractal...

Merci !

Try
if a then b else c

User avatar
OnoSendai
Developer
Posts: 6241
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Indigo Renderer 4.0.43 Beta Release

Post by OnoSendai » Wed Aug 31, 2016 4:35 am

Originalplan® wrote:Material upload still does not work. Same problem as reported before.
Should be fixed for 4.0.45.

Iamme
Posts: 5
Joined: Wed Aug 31, 2016 4:45 am

Re: Indigo Renderer 4.0.43 Beta Release

Post by Iamme » Wed Aug 31, 2016 7:28 am

Hi all,

This is my first post on this amazing forum! :) And I need some help!

I am using SKindigo to create a short animation sequence; 25 fps with a 10 sec. scene transition between 2 scenes. So far, I noticed the following issues (I also included images to better illustrate the problem):

- Indigo defaults to path tracing before rendering a new scene in the render queue.
- Clamp contribution is turned off before rendering a new scene in the queue.
- Open CL is turned off before rendering a new scene in the render queue

Is there a way to override these behaviors? So that user defined render settings on one scene does not change and is implemented across all scenes in the render queue?

Comments and/or suggestion would be greatly appreciated!

:)
Attachments
Render Queue_edited.jpg
Animation Issue_1 edited.jpg
Animation Issue_2 edited.jpg

User avatar
OnoSendai
Developer
Posts: 6241
Joined: Sat May 20, 2006 6:16 pm
Location: Wellington, NZ
Contact:

Re: Indigo Renderer 4.0.43 Beta Release

Post by OnoSendai » Wed Aug 31, 2016 7:54 am

Lamme: check out the 4.0.44 release, with queue settings overrides: http://www.indigorenderer.com/forum/vie ... =1&t=14118

User avatar
CTZn
Posts: 7240
Joined: Thu Nov 16, 2006 4:34 pm
Location: Paris, France

Re: Indigo Renderer 4.0.43 Beta Release

Post by CTZn » Wed Aug 31, 2016 8:58 am

OnoSendai wrote:
CTZn wrote:Yes that's what I call the if(a,b,c) form, and it does work.

No, I have not updated my CG board yet. It manages GPU Acceleration but not OpenCL rendering, that was all on CPU's. I have kind of fixed it by reverting to fbm's instead of multifractal...

Merci !

Try
if a then b else c
Ahw yesss I forgot about it, that's what I've been told once already. Thanks again !

Post Reply
35 posts

Who is online

Users browsing this forum: No registered users and 4 guests