Le 16-11-21 à 07:16, Jim Holsenback a écrit :
> On 11/20/2016 5:35 PM, Alain wrote:
>> Very good. My only complain would be about those area_light artefacts in
>> the left anf right shadowed areas. Try increasing the adaptive value by
>> 1, it should help, but it will obviously take longer to render.
>
> sa-weeeet ... increasing adaptive to 2 fixed it
>
Good :)

Remember:
adaptive 0 = start with only the 4 corners of the area_light.
adaptive 1 = start with 1 subdivision, or a 3*3 array.
adaptive 2 = start with 2 subdivisions, or a 5*5 array.
...

Post a reply to this message

On 11/20/2016 8:20 AM, Jim Holsenback wrote:
> i'm also working on a monochrome version with scattering media.

i reworked the material a bit ... gave blend_mode and blend_gamma a test 
drive. i have a cool purple one but like this jade-ish version the best. 
type 5 scattering media turned out ok too. i don't dare release the 
kraken (area lights) on this one ... wondering if there is a chance of a 
new subsurface option? ignore area lights for finish (subsurface) but 
use with object (shadows)

Post a reply to this message

Attachments:
Download 'jadeknot.png' (326 KB)

Preview of image 'jadeknot.png'

Am 02.12.2016 um 00:35 schrieb Jim Holsenback:
> On 11/20/2016 8:20 AM, Jim Holsenback wrote:
>> i'm also working on a monochrome version with scattering media.
> 
> i reworked the material a bit ... gave blend_mode and blend_gamma a test
> drive. i have a cool purple one but like this jade-ish version the best.
> type 5 scattering media turned out ok too. i don't dare release the
> kraken (area lights) on this one ... wondering if there is a chance of a
> new subsurface option? ignore area lights for finish (subsurface) but
> use with object (shadows)

Unfortunately that's not possible for technical reasons: Shadow handling
is too deeply intertwined with the subsurface light transport algorithm
to be done separately.

Post a reply to this message

On 12/1/2016 7:04 PM, clipka wrote:
> Am 02.12.2016 um 00:35 schrieb Jim Holsenback:
>> On 11/20/2016 8:20 AM, Jim Holsenback wrote:
>>> i'm also working on a monochrome version with scattering media.
>>
>> i reworked the material a bit ... gave blend_mode and blend_gamma a test
>> drive. i have a cool purple one but like this jade-ish version the best.
>> type 5 scattering media turned out ok too. i don't dare release the
>> kraken (area lights) on this one ... wondering if there is a chance of a
>> new subsurface option? ignore area lights for finish (subsurface) but
>> use with object (shadows)
>
> Unfortunately that's not possible for technical reasons: Shadow handling
> is too deeply intertwined with the subsurface light transport algorithm
> to be done separately.
>

we can already opt /not/ to use area lights in scenes (generally 
speaking) and by extension subsurface ... that is good subsurface 
results aren't depended on area lights or is it?

Post a reply to this message

On 11/20/2016 11:16 AM, clipka wrote:
> Have you considered trying UberPOV? It does include some mechanisms that
> might trim down render times considerably when multiple oversampling
> effects (e.g. focal blur, SSS, area lights, anti-aliasing) are combined.

just tried uberpov 3.7.1-x.colour.8871403 and compared it to pov 
3.7.1-alpha.8889454

did a subsurface and scattering comparison (no area lights):

subsurface only: pov 34.315 min  uber 47.6 min  pov 38% faster
scattering only: pov 16.1 min  uber 15.8 min toss up

why the big difference in render time for subsurface!!! not leading me 
to expect any pay off by adding more than one effect or area lights into 
the mix

absolutely no changes to sdl file other than #version 3.7;  -vs- 
#version unofficial patch 3.7;

Post a reply to this message

Am 02.12.2016 um 19:19 schrieb Jim Holsenback:
> On 11/20/2016 11:16 AM, clipka wrote:
>> Have you considered trying UberPOV? It does include some mechanisms that
>> might trim down render times considerably when multiple oversampling
>> effects (e.g. focal blur, SSS, area lights, anti-aliasing) are combined.
> 
> just tried uberpov 3.7.1-x.colour.8871403 and compared it to pov
> 3.7.1-alpha.8889454
> 
> did a subsurface and scattering comparison (no area lights):
> 
> subsurface only: pov 34.315 min  uber 47.6 min  pov 38% faster
> scattering only: pov 16.1 min  uber 15.8 min toss up
> 
> why the big difference in render time for subsurface!!!

Can't say off the top of my head. It probably has to do with subsurface
light transport being inherently based on numerical integration (read:
it needs to shoot rays in lots of directions to do its job), which needs
to be shoehorned in to POV-Ray by brute force, while UberPOV provides a
dedicated framework for such numerical integration stuff. Maybe that
framework adds some unexpected overhead, or maybe it causes some
implicit increase in the number of rays short when default settings are
used.

Or maybe it's some entirely unrelated overhead unexpectedly introduced
with one of the other UberPOV features.

> not leading me
> to expect any pay off by adding more than one effect or area lights into
> the mix

Go ahead and try.

The crucial thing here is that in official POV-Ray, when both subsurface
light transport and area lights are used, the computational effort
doesn't simply add - it multiplies instead. This shouldn't be the case
with UberPOV's special framework.

Post a reply to this message

On 12/1/2016 11:35 PM, Jim Holsenback wrote:
> i reworked the material a bit

Me like. :)

-- 

Regards
     Stephen

Post a reply to this message

On 12/2/2016 3:01 PM, clipka wrote:
> Go ahead and try.

hmmm ... with uber subsurface + area lights render was > 14 hrs. pov is 
/still/ ~ 38 % faster

Post a reply to this message

Jim Holsenback <spa### [at] nothanksnet> wrote:
> On 12/2/2016 3:01 PM, clipka wrote:
> > Go ahead and try.
>
> hmmm ... with uber subsurface + area lights render was > 14 hrs. pov is
> /still/ ~ 38 % faster

That's a curious thing because I found subsurface together with area lights and
radiosity to be faster in most recent UberPOV versus the 3.7.1.1-alpha.

Same scene file was able to get a fraction, about 15%, of the render done while
the "alpha" was still doing the radiosity pretrace. Around 5 minutes to complete
pretrace versus about 25 minutes.

Post a reply to this message

On 12/4/2016 8:53 AM, omniverse wrote:
> Jim Holsenback <spa### [at] nothanksnet> wrote:
>> On 12/2/2016 3:01 PM, clipka wrote:
>>> Go ahead and try.
>>
>> hmmm ... with uber subsurface + area lights render was > 14 hrs. pov is
>> /still/ ~ 38 % faster
>
> That's a curious thing because I found subsurface together with area lights and
> radiosity to be faster in most recent UberPOV versus the 3.7.1.1-alpha

maybe it was edge of that blackhole gravity well that passed through our 
sector last evening

Post a reply to this message