ERPT chain length

[Meelis]

Hi

What's good value for ERPT chain length when basesampler is metropolis with maxconsecrejects 16384?

[jeanphi]

Hi,

Don't use metropolis as a base sampler for ERPT. I'd be inclined to say don't use ERPT at all, but whatever... ERPT results have never been consistent with other samplers and it's not even converging faster than metropolis.

Jeanphi

[Meelis]

Hi

Thx Jeanphi.

Yes i noticed after 15 hours of rendering, there was allmost no coverage of SSS medium, but atleast no white noise pixels
Even lowdiscrepancy apeared much faster, but with some noise that will convert to caustics:

Code: Select all

Sampler "lowdiscrepancy"
   "integer pixelsamples" [32]
   "string pixelsampler" ["vegas"]

[J the Ninja]

Btw, if you are using a very recent version of luxblend, you might notice metropolis is no longer an option for base sampler.

[Meelis]

Btw, if you are using a very recent version of luxblend, you might notice metropolis is no longer an option for base sampler.

Just looked in Blender 2.62.2 r45133 and no MLT for ERPT.
I exported my last scene with older Blender 2.61.0 r43210 because new blender wasn't able to export with ERROR:

'Mesh' object has no attribute 'faces'

[Meelis]

I don't know what this chainlength does but the random basesampler is good for ERPT like advised
and it seems ERPT likes a bit higher pixelsamples.

Code: Select all

Sampler "erpt"
   "integer pixelsamples" [128]
   "string pixelsampler" ["vegas"]
   "integer chainlength" [1024]
   "string basesampler" ["random"]

And ERPT seems to be able to avoid MLT overburning problem on large scale SSS volumes while usually the SSS is no problem for the MLT
(you get this bright pixels that mute out step by step and over and over till the render is considered finished and not much noticable changes happen anymore)

But sometimes you find yourself large scale SSS volume overburning starting after 50 or 100 hours (depending on scene ofc) of rendering and from there every next update the overburning gets stronger and stronger.

ERPT 1 hour


MLT 1 hour


Random 1 hour


MLT Render in starting stage


MLT Render After Some Point In Futture (Overburning) ~50-100 hours


ERPT 16 hours

[CoolColJ]

After some experimentation, I finally found some good settings for ERPT

ERPT has always puzzled me, but today I got a good feel for it.

Tried these -

Bidrectional
32 depth all round
Random Sampler
Hilbert pixel sampler
2 samples per pixel

I tried chain length 8 and 16 so far for 15 mins each - the lower number means caustics don't develop as quickly, but it seems to render bigger chunks of the image with Hilbert on a semi-tiling basis when using a 3 sec screen update.
So it covers the whole render and develops quite quickly. Reaches higher samples per pixel much quicker.
First pass is sorta darker, and gets brighter every pass in a bucket rendering fashion.

If you use Vegas or Low discrepency pixel samplers it looks like a mess!
Tile and linear is not bad, it goes down in a scanline fashion, but is slower than Hilbert

I'll try some higher chain numbers and report back. When I used to text edit my scene files I have gone as high as 1 million

I'll post some pics of my tests in my render tests thread soon.

[Meelis]

CoolColJ BTW do you get that glow problem

[jeanphi]

Hi,

ERPT integrator is completely buggy, it doesn't give results consistent with the other integrators, and doesn't even converge faster than metropolis.

Jeanphi

[CoolColJ]

yeah it kinda acts funny, and it doesn't seem to be able to render refracted caustics