New light solver is fantastic.

[Jpjapers]

Hi Guys

Just wanted to say the new light solver is brilliant.
I work with Archviz as im sure many of you do but I deal with alot of big warehouses (70k sqft+) and its always been a pain to light them since in reality there are hundreds of lights. The new light solver seems to remove a large chunk of time off the renders for these scenes. Is it a probabilistic approach? Is it something that's being further developed for 2.0?

Thanks and good work as always!

Jack

[maru]

The new light solver seems to remove a large chunk of time off the renders for these scenes. Is it a probabilistic approach? Is it something that's being further developed for 2.0?

The new solver can speed up rendering A LOT - check out the example at https://coronarenderer.freshdesk.com/support/solutions/articles/12000039640

But as you can see in the ui it's still marked as "experimental". There are cases where it can misbehave, for example if some "fake" effects are used such as lights which do not cast shadows, and this can result in dark, square artifacts, which are mostly visible on flat solid surfaces like walls.

Yes, we are planning to improve this feature so the word "experimental" would be removed, and it would be on by default in all scenes. Let's hope it will be done for version 2, but it is not 100% certain.


By the way, could you post some examples of on/off results you are getting? :)

[RolandB]

Hi
I just tried this option and for me it's not as different in terms of speed as you ...
On the contrary, without the light solver it's faster :
(I can't publish the render for the moment).

With the light solver : 25 passes for 17'58"
Without (disabled) :  25 passes for 13'37"

The only thing I've noticed is a better quality of rendering.

[NicolasC]

Roland - as far as I understand this feature, it really depends on the complexity of your lighting scenario ... quote from the help article:

Enabling this option may significantly reduce render times in scenes:

With many lights (e.g. 10 or more)
Where lights have complex shapes
With complex lighting scenarios (e.g. if the lights are occluded by some objects, blocking the path from the light source to the camera)

Here with many many lights, it's definitely a game changer !

[TomG]

With the light solver : 25 passes for 17'58"
Without (disabled) :  25 passes for 13'37"

The only thing I've noticed is a better quality of rendering.

Don't forget that sometimes "faster rendering" does NOT mean the same number of passes in less time - sometimes a single pass may take longer, but be of much greater quality (lower Noise Level, and visually much better). In those cases, comparing the same number of passes does not give the comparison, and instead rendering for a specified time or to a specified noise level will show the improvement. In both cases, this means you will need less passes to get to the same result, and so less time = faster rendering (and if you are using passes, now you set a lower number of passes, to get the same or better results in less time)

[RolandB]

With many lights (e.g. 10 or more)
Where lights have complex shapes
With complex lighting scenarios (e.g. if the lights are occluded by some objects, blocking the path from the light source to the camera)

Sure I know when you can get the best result. In my scene there are approximately 20 lights with some different shapes so I think it can be tested with the LS.

Don't forget that sometimes "faster rendering" does NOT mean the same number of passes in less time - sometimes a single pass may take longer, but be of much greater quality (lower Noise Level, and visually much better). In those cases, comparing the same number of passes does not give the comparison, and instead rendering for a specified time or to a specified noise level will show the improvement. In both cases, this means you will need less passes to get to the same result, and so less time = faster rendering (and if you are using passes, now you set a lower number of passes, to get the same or better results in less time)

Completely agree with you : what I've noticed at the end of my answer was that quality is better than the render without light solver; so you're right, in this case, with less passes, I can have the same result -and even better- in less time.
Thanks for your answers guys.

[Jpjapers]

I cant post any renders right now but ill try to figure out how to demonstrate the speedup im seeing in the model next week when im back at work.
Ill post some render times too. Off the top of my head, it pretty much halved render time in my scenes (corona is reporting 8000 lights because of light materials on cylinders + lots of strip lights etc) on 1080p renders from an hour to about 25 mins including denoising of 11 elements. Im not using any fakes the only one i did see a few square artefacts appearing was using Light materials under a mask in a blend node but it was subtle enough that after denoising it was barely noticeable.