Comparison Between Lux and Cycles

Post your tests, experiments and unfinished renderings here.
User avatar
rrubberr
Posts: 23
Joined: Sun Mar 25, 2018 5:21 am
Location: California

Comparison Between Lux and Cycles

Post by rrubberr » Thu Mar 07, 2019 1:29 am

I was interested today after downloading the latest beta of Blender 2.8, and thought I might try comparing it to Luxcore. I modeled a very simple 'Cornell Box,' just to test the performance in a slightly difficult lighting scenario, and the results were a bit interesting, especially when it comes to the quality of denoising.

Both examples were 128 samples, with the same size and type of light-source, the same hex-color values, and the same sobol sampler. The Cycles image was rendered with "Full Global Illumination" (128 light bounces), while the Lux scene used my default setting of 10/8/8/10 bounces.
Luxcore.png
Cycles.png
The top or first is Lux, the bottom or second is Cycles. The same scene can be viewed without the denoiser below.
LuxcoreNoDN.png
CyclesNoDN.png
Again, the top or first is Lux, the bottom or second is Cycles. The light appears brighter in Cycles because it does not have automatic camera exposure.

A couple takeaways are that Cycles has much more color noise, and OIDN is a FAR superior denoiser to pretty much all the competition.

What do you guys think?

User avatar
Sharlybg
Donor
Donor
Posts: 1239
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: Comparison Between Lux and Cycles

Post by Sharlybg » Thu Mar 07, 2019 8:36 am

Don't think it is fair to compare luxcore to cycles with 128 path deph for cycles and just 10 max for lux. if you want good comparison try to simulate exactly same setting for each renderer : Cameera/light / engine.
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA

User avatar
Dade
Developer
Developer
Posts: 2383
Joined: Mon Dec 04, 2017 8:36 pm

Re: Comparison Between Lux and Cycles

Post by Dade » Thu Mar 07, 2019 10:50 am

rrubberr wrote:
Thu Mar 07, 2019 1:29 am
A couple takeaways are that Cycles has much more color noise
Try to use PhotonGI in the last v2.2alpha0, the result will embracing for Cycles :mrgreen:

But I agree with Sharlybg, can't you set the max. depth to 10 in Cycles ?
rrubberr wrote:
Thu Mar 07, 2019 1:29 am
OIDN is a FAR superior denoiser to pretty much all the competition.
Blender foundation isn't exactly new to NIV (Not Invented Here) syndrome. Oidn is damn good and cost nothing to integrate.
Support LuxCoreRender project with salts and bounties

User avatar
B.Y.O.B.
Developer
Developer
Posts: 2505
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Comparison Between Lux and Cycles

Post by B.Y.O.B. » Thu Mar 07, 2019 11:21 am

Dade wrote:
Thu Mar 07, 2019 10:50 am
Blender foundation isn't exactly new to NIV (Not Invented Here) syndrome.
When the Cycles denoiser was created, there was no OIDN (or BCD, or Optix denoiser).
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Developer
Posts: 2383
Joined: Mon Dec 04, 2017 8:36 pm

Re: Comparison Between Lux and Cycles

Post by Dade » Thu Mar 07, 2019 11:54 am

B.Y.O.B. wrote:
Thu Mar 07, 2019 11:21 am
Dade wrote:
Thu Mar 07, 2019 10:50 am
Blender foundation isn't exactly new to NIV (Not Invented Here) syndrome.
When the Cycles denoiser was created, there was no OIDN (or BCD, or Optix denoiser).
When we integrated BCD, OIDN wasn't available too and so :?: Didn't we integrated OIDN :?:

Try to read the comments to the patch with OIDN support for Blender...
Support LuxCoreRender project with salts and bounties

wasd
Posts: 168
Joined: Tue Apr 24, 2018 7:20 pm

Re: Comparison Between Lux and Cycles

Post by wasd » Thu Mar 07, 2019 12:27 pm

Cycles has no bidir, nor Metropolis. It's just of of league.
p.s. cycles has bump. luxcore has NO bump
CPU Bidir + Metropolis | Log Power | Core i5-4570

User avatar
Sharlybg
Donor
Donor
Posts: 1239
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: Comparison Between Lux and Cycles

Post by Sharlybg » Thu Mar 07, 2019 1:35 pm

cycles has bump. luxcore has NO bump
Bump ?
Or displacement ?
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA

S0rda
Posts: 69
Joined: Sun May 27, 2018 8:34 am

Re: Comparison Between Lux and Cycles

Post by S0rda » Thu Mar 07, 2019 5:58 pm

I also noticed that in cycles the noise is suppressed much worse, and that, in fact, if the fireflies appear on the image at the beginning of the render, even 5k passes may not have any effect. While luxrender handles such moments much better. Especially when there are some reflections, causing metal glare or various refractions (caustics) of glass.

wasd
Posts: 168
Joined: Tue Apr 24, 2018 7:20 pm

Re: Comparison Between Lux and Cycles

Post by wasd » Fri Mar 08, 2019 12:26 am

Sharlybg wrote:
Thu Mar 07, 2019 1:35 pm
cycles has bump. luxcore has NO bump
Bump ?
Or displacement ?
Bump. I cannot use height map reliably, it's something which pretty realiably (every single time) gives me ugly unpredictable result.
CPU Bidir + Metropolis | Log Power | Core i5-4570

User avatar
Sharlybg
Donor
Donor
Posts: 1239
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: Comparison Between Lux and Cycles

Post by Sharlybg » Fri Mar 08, 2019 9:17 am

Bump. I cannot use height map reliably, it's something which pretty realiably (every single time) gives me ugly unpredictable result.
Don't have this issue. but most of the time unpredictable result come frome bad/disoder in object relative size. to solve it check your normal in edit mode. if they are not correct just Add a new mesh (cucbe/plane or whatever you want) and first select your problematic object and select the new added mesh then Ctrl +J. in edit mode delete new mesh verticles. this will reset your object relative size to make bump work better.
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA

Post Reply