Progressive sampler - set maximum passes

[Jon3D]

Hi,
It would be nice to have an option in the Progressive image sampler Rollout to set the maximum Passes for rendering the final Image.

I often use a small region of an important area of my image and testrender it with the progressive sampler. After for example 250 passes these Area is noise free, so I know I have to render the whole final image at least 250 passes to get the same noise quality as is my testrenderings.

The "Render time" option is nice but in the most cases its difficult to estimate what time to choose if you testrender with small regions. A option " render up to x passes " would be very helpful in my opinion.

These options could also be combined for example: "render up to 200 passes but not longer than 200 min "

regards
Jon

[vlado]

But you can already do that? There's the max. subdivs value (which directly translates to a number of passes), the timeout value and the noise threshold. Rendering will stop as soon as any of these are reached.

Best regards,
Vlado

[Jon3D]

my fault! I didnt´t know that. Just tried it out and it works exactly like desired.
Thanks!!!

[Morbid Angel]

Also must set global subdivs multipler to 0. So for me this only works if I use the following settings:

min 1
max 5 (this can be any value for max subdivs)

noise 0.005
time 0

global subdivs 0

[vlado]

Also must set global subdivs multipler to 0.

This is only relevant if you have a relatively low number for max. subdivs (lower than the subdivs for materials, lights etc). It's exactly the same logic as the bucket sampler.

Best regards,
Vlado

[Jon3D]

When testing it with more complex scenes It really works only if I change the Global subdivs mult. to zero, like Morbid Angel said. Thanks.
If I only change the max subdivs parameter to a lower value than 100 the passes itself take longer to calculate and the image renders a lot longer.
Is that because the lights for example will render with more shadow rays because they are divided by less image samples when I only lower the max AA subdivs?
sorry for the trouble
kind regards
Jon

[vlado]

Is that because the lights for example will render with more shadow rays because they are divided by less image samples when I only lower the max AA subdivs?

Yep. V-Ray 3.1 actually shows you for each light how it will be sampled with the current render settings.

Best regards,
Vlado

[Mnietek2707]

What if irradiance map is used? Lowering Global subdivs mult. to zero leads to splotches...

[vlado]

What if irradiance map is used? Lowering Global subdivs mult. to zero leads to splotches...

You will have to find out the subdivs parameter that slows things down and make sure that is has a more reasonable value. Then you would leave the global subdivs multiplier to 1.0.

Best regards,
Vlado

[Subburb]

Hello ! i'm very interested in this thread, as i'm in troubles setting the progressive sampler.

i found as well that lowering max subdivs can lead to longer rendertimes.
another thing : changing the ray bundle size changes the rendertime. Is there some simple "rules" i can take in account to adjust ?

my goal is to reproduce in SR the same quality/rendertime than in bucket mode.
All is computed a this stage.
materials are parsed, and, for example, the biggest subdivs found in materials is 17. global multiplier=5, so total = sqrt(17*17*5) wich represents 38 subdivs if i'm not wrong. noise thr for AA and DMC are 0.01.
Does i need to put this value (38 ) as max subdivs in progressive (global mult is still 5.0), or 17 ? or i'm totally wrong ?
is it preferable to keep ray bundle size at 128 ?

in resume, is it possible to "match" a bucket quality/time with progressive when all parameters are known ?
Thanks in advance, and congrats for this SP1, it rocks !

[vlado]

Both the adaptive sampler and the progressive sampler work in identical way concerning subdivs values. There should be no need to adjust anything to get them to match.

Best regards,
Vlado

[Subburb]

Thanks Vlado, that's what i figured.
but, if i compare results : 1-5 in adaptive sampler with noise thr = 0.01 and progressive with same values does not gives same results...at all !
bucket render is twice longer with far less grain. whatever the global mult.
ray bundle size does play a role in this ?

note :
my adaptive settings are 1-5 color thr 0.01
my DMC settings are adapt amount 0.9, noise thr 0.012, min samples 48, global mult 5.0

when switching to progressive, i need to divide the color thr (0.01) by 5 to get approx same results..
Is there a difference between clr thr (adaptive)and noise thr(progressive) ?

Thank you, still playing with it