OTOY Forums

Finally I've completed my test and probably find the solution, that I'll share here if someone other could be interested.

In the image "test A" I've setted the IES to 100 samples, and rendering the scene at 3500 pixel, pathtracing, 8000 samples.
Caustic blur/path term./coherent ratio= 0,35
with a rendering time of 6h15 min as you can see I've again a lot of noise

In the image "test B" I've setted the IES to 10000 samples, and rendering the scene at 3500 pixel, pathtracing, 8000 samples.
Caustic blur/path term./coherent ratio= 0,35
with a rendering time of 6h13 min (equal than the previous test) I've finally a clean image.
I've noticed that:
- raising the IES samples from 100 to 10000 not affect the render time. Can someone explain me how these samples works?
- the other light in the scene (sampled with a value of 100 and with 0 opacity) have at the end lower luminosity that the prevoious test. Again someone could explain me why?

Again I've made another test giving double resolution and half samples to the image (7000 pixel and 4000 samples) but I've noticed again noise with almost double rendering time (11h 40).

The last test (test D) was made in PMC, same resolution than test C, but with only 1500 samples with a rendering time of about 10h30 having a better results in therm of noises than the previous test, with a lower rendering time.

Summarizingthe better option for my opinion is to go with solution B or D prefering this one also with a rendering time a little bit higher, but with double resolution.

Hi Giolets,

thanks for your final results and your investigations.
I also prefer raising samples in ies lights to 10.000, and in most of my scenes it cleans up a lot faster and is almost clean at 2000 samples. Didn`t test to render your scene with 8.000 or even 10.000 samples, so I thought it will not have an impact... but as you showed, it has.

I cannot explain excactly, but for my understanding raising the samples of IES lights to a crazy high number gives the renderer the information to concentrate really hard on the areas affected by these lights, a lot more than on others... so more of the Ms/sec go to these prefered areas and therefor they get cleaner "more quickly".

For the changes in luminosity: I also experienced this... IES lights with high sample values seem to get brighter than with less sample values. Also it seems to affect the overall Luminosity ratio with other lights.
Cannot explain that excactly, but I think it has to do with the way unbiased rendering of light rays work.... more samples of a lightsource = more light rays are shoot out or at least reach the environment with more intensity?

Increasing the samples of one light doesn`t affect the overall rendertime, it just affects the time it needs to get noisefree.
If you have lots of lightsources sample values just seem to change the ratio of the time it needs to get different lightsources noisefree.

PMC overall produces less fireflies or noise than pathtracing, that can be read in lots of other threads... but it needs more rendertime.

Cya,
ChrisVis