Does anyone know if there's a formula for CPU-based branched path rendering and GPU non-branched path to achieve exactly the same quality? I tried 10 and 40 and whilst the image quality appeared the same, there is a lot more bounced light with the branched path render.

Thank you in advance.

Branched and non-branched path renders use a different approach and they also treat materials differently. Blender - Branched path tracing explained shows the difference between the two and what influence the sample and bounce settings have on each.

Thanks for that link Wim. As a Blender user, this will be doubly useful info for me.

wimvdb posted at 1:24PM Mon, 09 August 2021 - #4424823

Branched and non-branched path renders use a different approach and they also treat materials differently. Blender - Branched path tracing explained shows the difference between the two and what influence the sample and bounce settings have on each.

Thank you Wim. Yes, this does partially illustrate (if not explain) why the lighting is different, but in no way helps me in determining a conversion formula, leaving me simply to eyeball it. Worse still, his approach of test rendering EVERY SINGLE SCENE with multiple settings is ridiculously impractical, AND Poser does not allow you to separate out the channels to see the impact in detail like Blender does. But thank you anyway.

Thank you Wim. Yes, this does partially illustrate (if not explain) why the lighting is different, but in no way helps me in determining a conversion formula, leaving me simply to eyeball it. Worse still, his approach of test rendering EVERY SINGLE SCENE with multiple settings is ridiculously impractical, AND Poser does not allow you to separate out the channels to see the impact in detail like Blender does. But thank you anyway.

It does not give a formula - correct. It does show however which bounce settings are important. It also shows you that a formula is rather difficult to determine since the branch path and non branched path renderers acting differently depending on the nodes used (blend for glossy/diffuse for instance). But the main reason I showed the video link is to explain the difference in light bounces.

I do support you with wanting to have additional render layers for spec, light, shadow, indirect light, etc. Those are very helpful.

wimvdb posted at 6:05PM Mon, 09 August 2021 - #4424876

Thank you Wim. Yes, this does partially illustrate (if not explain) why the lighting is different, but in no way helps me in determining a conversion formula, leaving me simply to eyeball it. Worse still, his approach of test rendering EVERY SINGLE SCENE with multiple settings is ridiculously impractical, AND Poser does not allow you to separate out the channels to see the impact in detail like Blender does. But thank you anyway.

It does not give a formula - correct. It does show however which bounce settings are important. It also shows you that a formula is rather difficult to determine since the branch path and non branched path renderers acting differently depending on the nodes used (blend for glossy/diffuse for instance). But the main reason I showed the video link is to explain the difference in light bounces.

I do support you with wanting to have additional render layers for spec, light, shadow, indirect light, etc. Those are very helpful.

Indeed. Thanks again Wim!

for renders you want to kick diffuse bounces up to 6. This is about the limit that I can tell a difference and should give you maximum light bounces in your scene. It is especially apparent when using the GROUND/Construct in your scene.

ghostship2 posted at 4:35PM Tue, 10 August 2021 - #4424989

for renders you want to kick diffuse bounces up to 6. This is about the limit that I can tell a difference and should give you maximum light bounces in your scene. It is especially apparent when using the GROUND/Construct in your scene.

Thank you Ghost - appreciate you sharing your experience.

Slightly tangential: I prefer Superfly 'progressive' as I can quickly see if scene will 'work'. Also, same results on this CAD Tower's twin GPU cards, its CPU-only, or networked 'Box' CPU-only...

(#1 is fast within now-ageing cards' complexity limits. #2 is reliable but, even as 'background', slows PC. #3 is 'Whoosh'...)

NikKelly posted at 11:43AM Sat, 04 September 2021 - #4426627

Slightly tangential: I prefer Superfly 'progressive' as I can quickly see if scene will 'work'. Oh yes - 100%