After reading a lot about these 3 renderer i decied to test them with an indoor 3D scene. So i take a benchmark file from Fstorm and convert it to Luxcore. both Fstorm and Luxcore are GPU base (Luxcore opencl/opensource… Fstorm Cuda/shareware).
-The luxcore file was rendered under pathocl engine (but i think upcoming biaspathocl splited kernel will do miracles)
Luxcore ====>> 2X R9 390 =====>> 36 mn (i’ve double the rendering time to compensate titanX and 980Ti strong normally 2X Fury X will be more fair for 18 mn sprint)
Corona =====>> i7 970 ====>> 15mn but it is more biased result and the convertion miss the yellow lighting effect and indirect light quality here is a bit down. speed with visible sacrifice i think.
Some of the materials in each of them are quite different from one another, which makes me wonder if you were careful in really trying to make sure they match up well.
Also, there is a chance the Fstorm comparison may become useless due to Otoy taking them to them to court over the alleged theft of source code (meaning that the product may no longer exist if they win).
Some of the materials in each of them are quite different from one another, which makes me wonder if you were careful in really trying to make sure they match up well.
Also, there is a chance the Fstorm comparison may become useless due to Otoy taking them to them to court over the alleged theft of source code (meaning that the product may no longer exist if they win).
I’am sure octane is wrong again ! in adition i’ve choose Fstorm because it is to me a more stronger challenger.
I am a just a bit curious when you do a test like this do you do your own tone mapping. Interesting tests.
Fstorm render is from fstorm user and this the same for corona. in luxcore i use a simple linear tone mapper.
i decied to be close as i can. you can also download the blend file to do a better correspondance this why i post it here !
you would also download this luxcore opencl 64bits version from august :
Battle, drum battle? Oh no, render engine battle.
I spend some time to adapt the file to Octane 3.0, just for fun.
It is really hard to get the materials and lightning the same for another engine so please be kind.
Some materials are even not set at all, Octane doesn´t have multimaterials so was a lot of work anyway.
Render time was 13 minutes on my GTX 670 + GTX 760, linear tonemapping, Octane glare.
They will not win this case because they have no case. Their “proof” is laughable and will not stand before a judge.
Otoy’s managment is nuts and borderline toxic.
They alienated their own devs and when they left, they tried to break their business or tried to keep them from compete with them.
Now that the brainchild of Octane is doing his own thing i really hope Octane crashes and burns, while Fstorm rises like a phoenix.
I don’t see quite why the big deal is made about everything being exactly the same. I understand the reasons, but wouldn’t it be more important to get the desired result from each and then compare the resources used?
This need 18 minutes for 4000 samples, first was 13 minutes for 5000 samples.
Cheers, mib
your attached picture link isn’t working. you can render with lower depth path ( i use only 4 to 6 for luxcore version) but try to do less biased (blurred) reflection. try to add more reflection on your dark red sahder. but i think octane is very faster for GPu like 760 and 670.
i think my Luxcore version need a little bit of gamma like yours. i wonder how luxcore perform for your rig ?
@sharlybg
your scene is not well prepared for LuxCore (always crashed using OCL)
went thru, made all materials nodal… so here’s first attempt on LuX… 10min compiling (?), 10min render (Q-M5000 // 21min on FStorm) next step optimization and corrections
B.Y.O.B
(Node Preview and LuxCore Addon Developer)
14
In the OP, the LuxCore render has the best AA out of the three in my opinion (look at the left door for example).
And what exactly does “always look odd” about the materials? IMHO that’s an artist issue, Lux provides physically based material models and the users have to use them in the correct way for the job (e.g. apply bump/normal/specmaps to make them look realistic).
Could you provide more specific examples so we can either improve this testscene or the renderer?
next step optimization and corrections