CoolColJ's test pics thread

Get feedback from others on your works in progress
User avatar
Polinalkrimizei
Posts: 647
Joined: Sat May 02, 2009 6:59 am

Re: CoolColJ's test pics thread

Post by Polinalkrimizei » Tue Jun 12, 2012 8:25 am

By funny colors I meant this, but it is starting to clear up after 1 hour and 4000 spp finally... looks pretty artsy-fartsy!
Attachments
Pool sss 1h, 4000 ssp.jpg

User avatar
Zom-B
1st Place 100
Posts: 4700
Joined: Tue Jul 04, 2006 4:18 pm
Location: ´'`\_(ò_Ó)_/´'`
Contact:

Re: CoolColJ's test pics thread

Post by Zom-B » Tue Jun 12, 2012 8:49 am

That happens with non uniform SSS and is will disappear over time :)
polygonmanufaktur.de

User avatar
Headroom
Indigo 100
Posts: 1058
Joined: Wed Aug 08, 2007 1:07 pm
Location: Spartanburg, SC, USA
Contact:

Re: CoolColJ's test pics thread

Post by Headroom » Tue Jun 12, 2012 1:48 pm

I'd assume you rendered that with BIDIR MLT. If you use PT-MLT the "noise" will turn monochromatic an will be less annoying (IIRC :wink: ).

There is some reference to it in this thread:
http://www.indigorenderer.com/forum/vie ... ochromatic

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

Re: CoolColJ's test pics thread

Post by CTZn » Tue Jun 12, 2012 11:36 pm

Hey polinalkrimizei, thanks for the PIGS.

I was concerned about the way the water surface was defined as faceted. Depending on the exporter it may be possible to have a faceted object with separated faces, cases in wich the subdivision trick I depicted above might have helped.

But there was no such issue :) I really thought that the faceted water was an asset to your scene, but it's true that the sun reflections are not that elegant then.
obsolete asset

User avatar
Polinalkrimizei
Posts: 647
Joined: Sat May 02, 2009 6:59 am

Re: CoolColJ's test pics thread

Post by Polinalkrimizei » Wed Jun 13, 2012 5:40 am

Ok ok I got it now - I remember your thread Headroom but didn't think about it. I did use BIDIR MLT, as I had the impression it would light the ground of the pool faster than PT with MLT enabled.

And I'm sure the noise will go away one day - but my impression is that the coloured noise will strive to zero while render time strives to infinity! This one is at over 40k spp, I'll stop now and continue my render without sss.

As shown, Indigo's forum is awesome. I'll leave this thread to be CoolColJ's playground again.
Attachments
Pool sss 41600spp.jpg

User avatar
CoolColJ
Posts: 1738
Joined: Mon Jun 25, 2007 1:47 pm

Re: CoolColJ's test pics thread

Post by CoolColJ » Wed Jun 13, 2012 11:32 am

Interesting I never had any issues with the SSS resolving, in these water renders, it was more the dispersion (cauchyB value) which caused the rainbow look. Unless the surfaces have problem finding the sun - something that MLT variants are just not good at unfortunately with bump map and displacement

User avatar
CoolColJ
Posts: 1738
Joined: Mon Jun 25, 2007 1:47 pm

Re: CoolColJ's test pics thread

Post by CoolColJ » Sat Jun 23, 2012 4:37 pm

Putting together an Asus P9X79 3930k 6 core, 16GB system ontop of my old X58a-UD3R i7 930 @ 4 ghz, 12GB system as I type this - then sell off the parts I no longer need 8)
Ram is ridiculously cheap these days!

Hope to hit 4.5 ghz on air - benchmarks and render tests to come..... :wink:
So an extra 30% performance based on architecture and clock speed, + 50% from 2 extra cores = 80% faster? :o

User avatar
CoolColJ
Posts: 1738
Joined: Mon Jun 25, 2007 1:47 pm

Re: CoolColJ's test pics thread

Post by CoolColJ » Mon Jun 25, 2012 3:12 pm

After much messing around, I finally have the system fairly smooth, except for booting off the SSD OS drive- definitely need a SP2 for Windows7...

Anyway the Intel 520 SSD is hella fast and so is the CPU 8)
I wonder how it compares to the i7 970 running at the same clock speed - because I am thinking of getting that for my old x58 system and revert back to that, as it can run my Intensity Shuttle video capture card, which is very finicky about USB3 ports! In any case, the extra 2 cores are worth the extra $$$

Don't have much overclocking knowledge with these Ivy Bridge CPU, so I let the motherboard do most of it for me :)
It can run at 4.5ghz in my setup, but will crash at 4.6ghz. So I run at 4.4 with a bit of headroom for summer.
Doesn't run too hot right now in winter, I have a nice positive pressure case Silverstone FT02 - see temps in desktop pic, at 17 degrees Celsius today. Add another 20 for summer peak and it's still well under thermal limits.


I get around 1.5 million samples/sec with the Pins test scene that comes with Indigo, running at 4.4ghz on the 3930k CPU.
With a simple outdoor scene, it flies, see pics below. The scene needs MLT to render well, as it has lots of reflections/caustics and has physical aperture diffraction.
The first few passes go by so fast, it's like a pure GPU renderer! :shock:
I would say it's pretty close to double the speed of my i7 930 running at 4 ghz....

All rendered for 6 mins each.

GPU mode over 5 million samples - doesn't tax the GPU too much even then... 46% GPU load
CCJ_Indigo_3930k_4.4ghz_GPU_Desktop.jpg
CCJ_Indigo_3930k_4.4ghz_GPU.jpg
Path tracing 3.6 million samples - fast, but slow :P
CCJ_Indigo_3930k_4.4ghz_PathTracing.jpg
MLT - fairly high sample rate, and overall convergence, but not as accurate as BMLT
1.5+ million samples
CCJ_Indigo_3930k_4.4ghz_MLT.jpg
Bidirectional Path tracing 1.6+ million samples
CCJ_Indigo_3930k_4.4ghz_BiDirectionalPT.jpg
Bidirectional MLT - the slowest smaples wise, but the fastest converger here 813+k samples
CCJ_Indigo_3930k_4.4ghz_BiDirectionalMLT.jpg

User avatar
CoolColJ
Posts: 1738
Joined: Mon Jun 25, 2007 1:47 pm

Re: CoolColJ's test pics thread

Post by CoolColJ » Mon Jun 25, 2012 5:08 pm

Just tried v3.4.2 - and it's even faster across the board!
ie a BMLT render I'm doing now went from 800+k to 1000+k samples per second

edit - looks buggy though, no apeture diffraction happens....

edit2 - it looks like there is a bug in the Indigo renderer, when your tweaking the F-Stop, while the renderer is running. The render updates, and the exposure changes, but the value you enter doesn't reflect the render results of the aperture dispersion. because when you save the scene, quit, and reload Indigo+ scene, it renders differently

User avatar
CoolColJ
Posts: 1738
Joined: Mon Jun 25, 2007 1:47 pm

Re: CoolColJ's test pics thread

Post by CoolColJ » Mon Jun 25, 2012 6:07 pm

CoolColJ wrote:Just tried v3.4.2 - and it's even faster across the board!
ie a BMLT render I'm doing now went from 800+k to 1000+k samples per second

edit - looks buggy though, no apeture diffraction happens....

edit2 - it looks like there is a bug in the Indigo renderer, when your tweaking the F-Stop, while the renderer is running. The render updates, and the exposure changes, but the value you enter doesn't reflect the render results of the aperture dispersion. because when you save the scene, quit, and reload Indigo+ scene, it renders differently

here are the bugs

v3.4.2 no diffraction and what are the red dots?
CCJ_v3.4.2_BMLT_Spheres.jpg
v.3.4 aperture diffraction with F16 on lens - red dots as well
CCJ_v3.4_BMLT_Spheres.jpg
Here I entered 128 into the F-Stop in Indigo, by typing it in, and then typing in 16 again to return it to the above scene, but it renders like it's at F-Stop 128, but with F16 exposure values. And you get this garbage on the walls
If I drag the F-Stop slider back to 16 then it seems to nullify the bug - it seems to happen with a few other values in Indigo as well, dragging the slider vs typing it in.
CCJ_v3.4_BMLT_Spheres_BUG.jpg

edit - it does seem to me Indigo does not has as much aperture diffraction as a real camera with the same settings. I mean having to go up to f128 to get them isn't exactly realistic :lol:

User avatar
CoolColJ
Posts: 1738
Joined: Mon Jun 25, 2007 1:47 pm

Re: CoolColJ's test pics thread

Post by CoolColJ » Mon Jun 25, 2012 6:18 pm

see what I mean?

real camera have way more diffraction at the same fstop as Indigo - me thinks the simulation needs tweaking :wink:

Image

StompinTom
Indigo 100
Posts: 1828
Joined: Mon Sep 04, 2006 3:33 pm

Re: CoolColJ's test pics thread

Post by StompinTom » Mon Jun 25, 2012 8:34 pm

CoolColJ wrote:see what I mean?

real camera have way more diffraction at the same fstop as Indigo - me thinks the simulation needs tweaking :wink:

Image
Could be because it's a shitty lens, though :P Depending on lens / filter coatings, etc., cheaper lenses will flare way more.

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

Re: CoolColJ's test pics thread

Post by OnoSendai » Mon Jun 25, 2012 9:26 pm

Hi CCJ,
Can you please post (or email us) the scene with the red dots?

Non-post pro aperture diffraction is broken in updated bidir, will fix.
Thanks!

User avatar
dcm
Posts: 663
Joined: Sun Jan 03, 2010 1:55 am

Re: CoolColJ's test pics thread

Post by dcm » Fri Jun 29, 2012 8:29 pm

i noticed problem with AD too when using Image

Image
Image

User avatar
Zom-B
1st Place 100
Posts: 4700
Joined: Tue Jul 04, 2006 4:18 pm
Location: ´'`\_(ò_Ó)_/´'`
Contact:

Re: CoolColJ's test pics thread

Post by Zom-B » Fri Jun 29, 2012 9:00 pm

CoolColJ wrote:real camera have way more diffraction at the same fstop as Indigo - me thinks the simulation needs tweaking :wink:
Maybe that (sadly still not released) Paper could be a good starting point:
Polynomial Optics: A Construction Kit for Efficient Ray-Tracing of Lens Systems
polygonmanufaktur.de

Post Reply
695 posts

Who is online

Users browsing this forum: No registered users and 11 guests