Need to check on my desktop - but i’m not seeing a huge difference on my phone.
Is the 3x render time worth it in this instance?
Isn’t it going to be quite hard to compare noise levels on a 279x533 jpg?
2 Likes
Yeh - not seeing a massive difference. This scene is probably not one that PG will help much with. Seems to be a lot of direct lighting - which PG isn’t designed to help with.
For this scene - it’s probably better off spending a bit more time rendering more samples than activating PG.
I think equal time renders - rather than equal sample renders tell far more about the capabilities of PG.
2 Likes
12 Likes
The comparison slider is awesome.
Path guiding really doesn’t seem very helpful in this particular scene. 3x rendertime doesn’t seem worth it.
This particular scene isn’t really a great demo, as the textured walls already can mask a multitude of noise.
7 Likes
The PG version Is slightly brighter, especially below the counter, but the noise is nearly the same
1 Like
Wow! Now, you know, you gotta share the markdown for the slider!
pleeease…
edit: is it this? is that you?
edit2: found it!
2 Likes
PG for glossy surfaces (with RIS) already in test 4.0 daily builds ?
Path guiding doesn’t really help with direct lights - and it looks like this scene is lit almost exclusively with them
Today build
300 samples
PGI OFF/Light Tree OFF PGI ON/Light Tree ON
Add two images here. Change ‘data-direction-horizontal’ to ‘data-direction-vertical’ for a vertical slider.
PGI ON/Light Tree OFF PGI OFF/Light Tree ON
2 Likes
Nice! Would love to see this with the new image comparison sliders (star icon in the edit message toolbar).
2 Likes
you can see a tiny bit more indirect light in the second compared to the first, especially near the top of the left wall
light tree is clearly more important than PGI for this
and PGI + light tree changes the noise profile quite a bit: Some parts are slightly more noisy and some slightly less than in the light tree only image. Overall, activating PGI too is an improvement though, I think
You probably should also compare equal time renders though
1 Like
Yep this. Equal time renders are important because they will inform whether it’s better to activate PGI and light tree (and take the performance hit) - or whether to just render more samples for longer.
The PGI and light tree turned on - the render took almost 8 minutes. What would an 8 minute long render look like with PGI and light tree turned off - it may be as good as, if not better in this instance.
1 Like
Tested the daily build on an old project:
Full GI preset, min light bounces: 8, min samples: 128, light clamping: 0, filter glossy: 0.10, 5min halt condition
RIS (5min, 48 samples) vs RIS with MLS (5min, 43 samples)
The lamp area looks much better without MLS, as well as some highlights on the bottle and window frame and SSS.
GPU (5min, 1022 samples) no MLS and GPU (5min, 980 samples) with MLS results for good measure
Better clarity in the cabin (pretty self-explanatory) compared to both path guiding renders, though the cave area does not nearly have the detail that RIS provides. MLS leads to a bit more illuminance overall.
2 Likes
Just a note:
IME, comparison rendering for same amount of time, reveals much better which result is better.
I think the portals rendering looks better. Did you also try portals + PGI?
Yes. In every interior case rendering using portals perform better (even in Vray, Corona, Indigo or Luxcore (where they got totally removed).
Analogy: It’s like having a dog fetching a stick. In one case it’s sniffing around for it and in other you’re simply showing (leading) the dog to it. Although dog will learn after few exercises, machine stays dumb/pre-programmed. Portals shouldn’t ever get removed, but instead evolved - having a tag on a material could be devised for accelerating workflow (and consequently LP guiding).
Also, bias introduced with such an approach remains within physical nature yet allows an artistic choice.
1 Like