Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

General project and community related discussions and offtopic threads.
daros
Posts: 223
Joined: Thu Dec 12, 2019 3:25 pm
Location: inside human language
Contact:

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by daros » Sun Jul 11, 2021 11:17 pm

Ops, it is getting worst and worst for Maxwell.

Maxwell 60 seconds
View 01 400x400max.jpg


Luxcore 60 seconds
View 01 400x400lux.jpg

test set-up
test3.jpg
Attached the scene file
Attachments
light bounce stress test3.zip
(103.05 KiB) Downloaded 49 times

daros
Posts: 223
Joined: Thu Dec 12, 2019 3:25 pm
Location: inside human language
Contact:

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by daros » Sun Jul 11, 2021 11:39 pm

That's why i'm saying the giga benchmark battle tests are totally misleading. They simply do not squeeze the render engines showing their real capabilities.

User avatar
Dade
Developer
Posts: 5594
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by Dade » Sun Jul 11, 2021 11:46 pm

I'm a bit surprised of Maxwell: it is pretty much supposed to be the best BiDir out there, may be with Indigo Render. Actually, they haven't very much competition, as far as I know, they are the only commercial available BiDir to say the true.
Sharlybg wrote:
Sun Jul 11, 2021 7:13 pm
Cycles
cycles 30.jpg
"Pitch black" ... or to use another movie tittle: "The Mist". Cycles (like Blender internal before) is really intended for toons animation, I have no idea why people use it for ArchiViz or Photorealism.
Support LuxCoreRender project with salts and bounties

daros
Posts: 223
Joined: Thu Dec 12, 2019 3:25 pm
Location: inside human language
Contact:

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by daros » Mon Jul 12, 2021 12:00 am

And i can show about other 4-5 tests in which luxcore beats Maxwell. I think luxcore is by far the best bidir around.

User avatar
DionXein
Posts: 79
Joined: Mon Jun 01, 2020 10:22 am

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by DionXein » Mon Jul 12, 2021 5:01 am

Dade wrote:
Sun Jul 11, 2021 11:46 pm
Cycles (like Blender internal before) is really intended for toons animation, I have no idea why people use it for ArchiViz or Photorealism.
I think because of high flexibility of setting up scene. If you took closer look to people like Gleb Alexandrov, Ian Hubert or "Default Cube", you can notice how ingeniously they use the full set of Cycles features to create really interesting scenes, compositions and virtual spaces. Gleb provides interesting thoughts about setting lights and env in scene, Ian shows quick and dirty, but pleasant way to build 3d environment in scene and animate it, Default Cube gives us understanding of non-destructive node based material creation.
And what about ArchiViz - I guess it's quicker to send scene to render, cuz you have no special "nerd orientated" settings to tweak - everything is simple.

daros
Posts: 223
Joined: Thu Dec 12, 2019 3:25 pm
Location: inside human language
Contact:

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by daros » Mon Jul 12, 2021 6:29 am

Most of the light in this scene is indirect light.
You can see how maxwell is faster in removing the overall noise. But there is a big cost in acccuracy compared to Luxcore.

Maxwell 2 minutes
View 01 1200x600max_1.jpg

Luxcore 2 minutes
View 02 1200x600lux_1.jpg


Maxwell almost completely ignores reflecting surfaces illuminated by indirect light. Similarly, Maxwell is unable to reflect caustics in a mirror.
Suddenly a small tweak on the code in one of the most reputable rendering engines gives a completely wrong result.

If you are working in a project with these lighting conditions, which are quite common, you can go crazy if the rendering engine doesn't behave correctly and gives you dull images. What would Corona do with a scene like this? What about Cycles?

User avatar
Sharlybg
Donor
Posts: 2953
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by Sharlybg » Mon Jul 12, 2021 9:24 am

Dade wrote:
Sat Jul 10, 2021 11:04 am
Sharlybg wrote:
Sat Jul 10, 2021 10:37 am
Sorry I remenber i have to add thoses lines yself :mrgreen:
There are more settings:

Code: Select all

# Higher is the max. depth and more effect RR settings will have
path.maxdepth = 32
# At what depth, RR will start to operate: lower is faster. Range [0, path.maxdepth]
path.russianroulette.depth = 3
# At what value, RR will start to operate: lower is faster. Range [0.0, 1.0]
path.russianroulette.cap = 0.5
Please Dade can i have all the settings for the super fast RR ? the one that produce this image :
Image
Support LuxCoreRender project with salts and bounties

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

marcatore
Donor
Posts: 459
Joined: Wed Jan 10, 2018 8:04 am

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by marcatore » Mon Jul 12, 2021 12:01 pm

Dade wrote:
Sat Jul 10, 2021 11:00 am
This is a rendering with intentionally fast RR settings (6.12M samples/sec):

good.jpg

As you see I can manipolate the samples/sec however I want. But check the "rays per samples": the first one traces ~36 rays per samples while the last one only ~4 per samples. Clearly the last one is perceived 6 time "faster" ... but it isn't, it has just more noise and potentially fireflies.
I trust on you and your words but I can't match what you wrote with what my eyes is watching.
I mean that you wrote it has more noise and potentially fireflies but in your example it seems the reverse.

At least the noise is many many times lower and I really like the noise free reflection.

About the fireflies it seems lower as number but probably they could arise later when an higher sampling level will be reached...but as now we can clamp, it could be interesting to see if it's possible to avoid them just with clamping.

User avatar
Dade
Developer
Posts: 5594
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by Dade » Mon Jul 12, 2021 12:07 pm

Sharlybg wrote:
Mon Jul 12, 2021 9:24 am
Dade wrote:
Sat Jul 10, 2021 11:04 am
Sharlybg wrote:
Sat Jul 10, 2021 10:37 am
Sorry I remenber i have to add thoses lines yself :mrgreen:
There are more settings:

Code: Select all

# Higher is the max. depth and more effect RR settings will have
path.maxdepth = 32
# At what depth, RR will start to operate: lower is faster. Range [0, path.maxdepth]
path.russianroulette.depth = 3
# At what value, RR will start to operate: lower is faster. Range [0.0, 1.0]
path.russianroulette.cap = 0.5
Please Dade can i have all the settings for the super fast RR ? the one that produce this image :
Image
This (faster):

Code: Select all

path.maxdepth = 32
path.russianroulette.depth = 1
path.russianroulette.cap = 0.0
Vs. (slower):

Code: Select all

path.maxdepth = 32
path.russianroulette.depth = 32
path.russianroulette.cap = 1.0
Support LuxCoreRender project with salts and bounties

User avatar
Dade
Developer
Posts: 5594
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Giga Benchmark battle : Luxcore Redshift CyclesX ECycles Vray Corona

Post by Dade » Mon Jul 12, 2021 12:11 pm

marcatore wrote:
Mon Jul 12, 2021 12:01 pm
I trust on you and your words but I can't match what you wrote with what my eyes is watching.
Are you mislead by the number of samples per pixel ? I took the screen shots at random points to show the difference in samples/sec and not in noise.

First was taken with 31 samples per pixel while the last was taken with 207 samples per pixel (clearly the last has less noise even with fast RR settings).
marcatore wrote:
Mon Jul 12, 2021 12:01 pm
About the fireflies it seems lower as number but probably they could arise later when an higher sampling level will be reached...but as now we can clamp, it could be interesting to see if it's possible to avoid them just with clamping.
Yes, like I wrote before, now with clamping and denoiser it may be worth having more aggressive RR settings.
Support LuxCoreRender project with salts and bounties

Post Reply