How to cut render times

[jirzi]

Hi,
I'm rendering product, which has top side from glossy transparent plastic. Below it, there is a circular plate with grooves, also from same material. Below plate, there is another plate with eight blue LED diodes, which illuminate it from inside. LEDs are small Vray spherical light with NoDecay enabled. I tried with decayed lights, but did not illuminate interior enough without very high multipliers, but lights then turn white.
Scene is illuminated with three Vray plane lights. I had to increase subdivs to 80 to eliminate noise.
Transparent material is standard Vray material with glossy refractions (0.92), 20 subdivs and interpolation on. Max depth is 20.
It is part of greater scene, which consist also from parts, made of regular plastic, but all scene is rendered fast except transparent stand on picture below.
Render times are ridiculous, for 960*1280 picture more than 40 hours. Can anyone help me to cut render times without sacrificing quality.
Below is picture of problematic stand and render setting. Picture was NOT made with setting below, it si a low quality preview just to see, how it looks, because higher quality picture is still rendering. Second picture are render settings for picture without noise, which is currently still rendering. Quality is good so far, but renders times are not.
I'm quite new with Vray, so any suggestions are very welcome. Thanks.

[gilicom]

try checking the store with irr map on them vray lights.

[trixian]

First thing I'd do is turn off "store direct light" in the LC roll out (seldom worth it as the quality is poo). Then i would lower the LC samples to 500, and turn on adaptive tracing. I'd also pre-filter it, and enable the "use light cache for glossy rays" option.
Secondly, figure out if you can render without glossy refractions on the plastic lid. Would it be possible for you to render out a refraction mask and blur what is seen through it in post?
Another thing....if using glossies, you'd generally be better off with Adaptive QMC AA instead of Adaptive Subdivision.
Oh, and I can hardly imagine where you would be able to see the 20 reflected ray-bounces inside that plastic. 10 should be plenty. Just make sure you have a good environment for the reflections, and\or play with the exit colour to avoid blackness in the refractions.

[jirzi]

Big thanks to both. I will try this at once.
BTW, how do you render refraction mask?

[trixian]

Big thanks to both. I will try this at once.
BTW, how do you render refraction mask?

Several ways. Temporarily setting self illuminating, pitch black and refractive materials to your pieces, and render it out with no lights etc. All you want is a black image, with anything that is supposed to be a refraction (of the bottom half of the thing you are making) visible as pure white. Then use that as a layer mask\alpha channel in photoshop\"insert favorite comp. suite".
There are more elegant ways though. play around with the render elements.....

[jirzi]

Thanks for explanation, I will try that.

[Paul Oblomov]

Hm..... You didn't tell PC specs!
Well, my settings would smth like
adaptive qmc (default)
medium irrmap
lc 800 (with store glossy to LC on)
no interpolation at all (on glossy)
15 refraction depth
rqmc sampler - 0.01
NO AA filter
subdivs on vraylight... mmm turn off GI and test just lighting noiseness

Turn min samples to default. 20 - too much. Overlaping - maybe density based ? Try to turn off gi caustics.
And you will get more noise on glossy and shadows (as ur using adaptive qmc AA). Well.... slightly noticable, than adaptve subdivision.
P.S. And maybe it's a good idea to test render settings using region render ? Start with glossy and GI off. Then turn on glossy, then turn on gi.

[priad]

80 subdiv is crazy high! It must be an other thing to tweak to exit noise.
In your case you try to achive an effect which call all the power of Vray with an adaptative solution (irmap) Why do not use directly a kind of brute force and only tweak for the noise?
May be you will win time and result.
try QMC/LC
IMO

[priad]

try also to turn down the adaptive in the qmc sampler rollout.
Sometime less adaptive things win time.

[jirzi]

Thanks, guys, I'm working on your suggestions, will post some results later.