CoolColJ's test pics thread

[Stromberg]

Cant help you with the problem, but the render sure is damn nice

[benn]

So I rerendered an old water test scene with that SSS tabulated spectrum "dirty" water I did ages ago in the lastest version of Indigo, and it sure looks different.
I like how it looks more realistic.
But I don't like how it doesn't accept MLT settings I used before all the time, so even at 3500+ samples, all the caustics and SSS stuff is still multi-coloured

also that front pillar is faceted when it rendered smooth in an old version....

I'll keep rendering it and see where it ends up

heh I thought you'd done that crazy color thing on purpose.

[Pibuz]

Something clearly has to be fixed, isn't it?
Lots of things have changed since the 1.0.4 version, but at least it'd be worth to check the things out.
Benn what do you say?

[suvakas]

I can confirm the "multi-colored" thing. It appears on my sss tests too. Specially when i use small SSS amounts.

Ps. Glad to see you're back CoolColJ.

[CoolColJ]

yeap, there is something wrong - what it looks like after 9500+ samples per pixel

[CoolColJ]

after a bit of testing, it looks like V1.0.9 is the last version that doesn't go multi-coloured when using SSS

Well V1.1.18 has the multi-coloured thing too, but I haven't tried any of the versions between that and V1.0.9

[CoolColJ]

Rendered mostly in v2.43, but switched back to v2.2 for proper tone mapping

Used my Dirty water material I just uploaded to the Material database., but without dispersion and SSS.
Up to 3700+ samples, but still not full converged yet. Took some 13 hours on my p4

Bidirectional MLT. I used 100 bounces and that greatly improved how early the stuff under the water showed up

I'm trying another render with my clear water values, and experimenting with Ono's rippled water shader displacing the current dense water mesh, and some other displacement on the pool wall etc

[everwind]

not even sure this scene could clean with only 100 bounces

If you render less bounces it should clean earlier but finally it looks worst cause it can't render little details, or at least that's what I saw with some of my scenes with water ...

[CoolColJ]

not even sure this scene could clean with only 100 bounces

If you render less bounces it should clean earlier but finally it looks worst cause it can't render little details, or at least that's what I saw with some of my scenes with water ...

not even sure this scene could clean with only 100 bounces

If you render less bounces it should clean earlier but finally it looks worst cause it can't render little details, or at least that's what I saw with some of my scenes with water ...

If there was SSS and dispersion, yeah it might not be enough, but for this, I'm sure it's plenty. You only need 3-4 bounces for the light to move through the water, to the floor and back up to your eyes. Just need a few more for the light to move through other angles etc

I also used a Max Consecutive Rejections value of 1500 vs 1000 and under. I've noticed it helps a lot in getting the caustics and floor to light up earlier. The render I'm doing now has a lot of caustics showing at 1800 samples already.

I subdvivded the water mesh in Indigo a few times. I have no clue what I'm doing here, but it ended up at 2.5 million triangles, a bit of an overkill
It already has ripples on it, but I also applied the shader from the Rippled Water material to displace it. Looks quite realistic re the caustics

The water is the regular tabulated clear water material I compiled before, but without SSS and dispersion. I'll let it render some more and see what develops, using only 100 bounces.

Now I can't seem to get the Indigo subdivsion and UV mapping to behave on the pool wall as you can see....there is supposed to be displacement as well....

[Zom-B]

Finally your famous water renders are back... missed the, its been some time

I have to tell you that this MLT parameter explanation is still a reference for me using SSS and MLT in general!

Looking forward to some interesting tests of you

If your old P4 gets killed in action, I'm sure some people in the IRC would like to contribute some rendertime for your tests,
I will for sure

[CoolColJ]

Finally your famous water renders are back... missed the, its been some time

I have to tell you that this MLT parameter explanation is still a reference for me using SSS and MLT in general!

Looking forward to some interesting tests of you

If your old P4 gets killed in action, I'm sure some people in the IRC would like to contribute some rendertime for your tests,
I will for sure

Haha - yeah I just haven't found the right time to upgrade. I don't want a platform that will not allow me to upgrade to a 6-8 core CPU in the future. Intel has a nasty habit of killing off socket formats in a hurry. In that regard going AMD is better, but AMD have a new platform next year....

I did try one of these scenes on my Dad's low end Quad core system. It uses the series of chips after the Q6600, not sure on the exact chip. But I was getting something like 150,000 samples a second in the scene vs 20,000 on my P4. It was faster, but it didn't blow me away like I expected. I was hoping the samples per pixel rate would jump up in 2s to 5s, but it wasn't that much faster. I hope an i7 920-930 overclocked to 3.8ghz or more will be like that. Off course GPU rendering will be nice and fast

As far as the MLT settings go, I've actually changed my opinions on a few things. Having large mutation probability down to 0.3 or so can help if have really complex light paths.
Man when I get a faster system I'll be able to test things out in more depth!

[CoolColJ]

It looks like 100 bounces isn't enough!
I let the above render continue and it pretty much showed no changes in the last 3 hours of rendering...
The floor didn't light up and smooth out as much as I expected.
Plus the caustic detail isn't that bright and not quite complete. Looks a bit flat.

I'm redoing the render now with some material changes, dispersion in the water, but with 1000 bounces and I can see much more detail and it looks more alive.
I started with Max consecutive rejections of 1500, but it is definitely too high as it reaches a point where the render doesn't do much.

[Pibuz]

..hey Cool: interesting tests!
...but what's going on with your stone wall?

[CoolColJ]

..hey Cool: interesting tests!
...but what's going on with your stone wall?

Indigo's mesh subdivision messes up the UVs

[Zom-B]

Indigo's mesh subdivision messes up the UVs

To be exactly, the smoothing parameter does: http://www.indigorenderer.com/forum/vie ... 387#p90387

My "workarround" for this is to have some good basic subdiv done in your 3D app and then use a viewdependend Indigo Subdiv on it ithout smoothing checked.
Costs RAM & Export, render start time, but thats the price to pay...