Page 3 of 5
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 12:11 pm
by Dade
jgrover110 wrote: Fri Apr 10, 2020 11:52 am
Updated to 20.4.1, still not correct. CPU load reaches 100% in task manager. This must be hardware specific if other people cannot reproduce it.
Does the Ryzen 5 2400g have an integrated GPU ?
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 12:16 pm
by jgrover110
Dade wrote: Fri Apr 10, 2020 12:11 pm
jgrover110 wrote: Fri Apr 10, 2020 11:52 am
Updated to 20.4.1, still not correct. CPU load reaches 100% in task manager. This must be hardware specific if other people cannot reproduce it.
Does the Ryzen 5 2400g have an integrated GPU ?
Yes it does, but I'm sure it's not using that. OpenCL device is labelled as "Ellesmere".
I also tried rendering with Luxcore standalone, same image.
One thing I found while exporting from Blender is that the exported camera FOV does not take into account the sensor fit method, which initially resulted in a much smaller FOV than anticipated
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 12:21 pm
by Dade
jgrover110 wrote: Fri Apr 10, 2020 12:16 pm
I also tried rendering with Luxcore standalone, same image.
Luxcore.PNG
Don't they look quite different (original problem screenshot Vs. stand alone) ?
The shadow in the first one is pitch black, I can get the same result if I set the number of CPU threads to 0. I will check again if I can reproduce this problem in some way.
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 12:39 pm
by jgrover110
Dade wrote: Fri Apr 10, 2020 12:21 pm
Don't they look quite different (original problem screenshot Vs. stand alone) ?
This is just due to the ACES tonemapping.
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 12:42 pm
by CodeHD
jgrover110 wrote: Fri Apr 10, 2020 12:16 pm
One thing I found while exporting from Blender is that the exported camera FOV does not take into account the sensor fit method, which initially resulted in a much smaller FOV than anticipated
This seems to be normal behaviour of LuxcoreUI specifically. Acutally had this question on ym mind for a while, maybe Dade can confirm if this is intended behaviour?
If you render from a terminal using luxcoreconsole.exe the FOV will be as expected.
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 12:49 pm
by jgrover110
For sake of comparison, here is the render in Blendluxcore without ACES, only sRGB.
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 1:16 pm
by Dade
CodeHD wrote: Fri Apr 10, 2020 12:42 pm
jgrover110 wrote: Fri Apr 10, 2020 12:16 pm
One thing I found while exporting from Blender is that the exported camera FOV does not take into account the sensor fit method, which initially resulted in a much smaller FOV than anticipated
This seems to be normal behaviour of LuxcoreUI specifically. Acutally had this question on ym mind for a while, maybe Dade can confirm if this is intended behaviour?
If you render from a terminal using luxcoreconsole.exe the FOV will be as expected.
Yes, luxcoreui resets the screenwindow parameters to use the current window size (so you can resize the window).
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 1:45 pm
by jgrover110
This is a rendering with Caustic Cache disabled. This shows that the Light Tracing is working as there is a caustic from the metal cube, yet the shadows are still incorrect.
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 1:49 pm
by Dade
jgrover110 wrote: Fri Apr 10, 2020 1:45 pm
This is a rendering with Caustic Cache disabled. This shows that the Light Tracing is working as there is a caustic from the metal cube, yet the shadows are still incorrect.
I have some serious trouble to notice any difference (aside from SDS paths)

What do you mean with incorrect shadow exactly ?
Re: HDRi and Light Tracing
Posted: Fri Apr 10, 2020 2:22 pm
by CodeHD
Here is a side-by-side comparison.
Still a tiny bit of color difference of the background plane but I think we can agree the shadow difference goes beyond that.
I could not replicate it (from within Belnder) with some settings variations (cpu/OCL, light tracing on/off, cache on/off)
I will try with standalone also, since it seems you were able to replicate it...