I tested the free one, and limited E-Cycles RTX to one GPU to be fair. My users compare to the paid version like here, here, here or here for a quick search.
So you don’t have any scene you can share to backup your claims? If so, I’ll make one.
1 Like
I know those videos and I’ve been tweaking that same scene properly when they were published.
I also commented those, but it looks like my comments are not there anymore.
And I also commented some of the post of the author, mentioning the render time of my tweaked scene and the fact the original renders where different and presenting a different level of noise.
I just said the tests were not accurate at all (he was using Octane 4 to compare, Octane settings where not good, renders where different).
I’ve also answered the posts in the forums you highlighted.
Please don’t see this as a “war” between renders. You’ve made and you’re still doing an awesome job with E-Cycles.
Testing renders is not easy at all and requires a deep knowledge of both the renders considered for the comparison.
And, of course, only the latest version should be used for both.
I’m absolutely available to get your Cycles scene and “tweak” it in Octane.
Or even better, you can provide the Octane scene directly so I can verify if the render settings are the best possible to get a fast render. That way I could render the original scene and my tweaked scene and compare the results without even having to work in E-Cycles (I’m not as expert in E-Cycles as I am in Octane for sure).
Thanks!
I’ll share a scene for everyone to try rendering. It’s not great, but no textures and I did it quick. I got it all set up for cycles. 35 sec using RTX 2070 super with Optix on and no eCycles because I don’t have eCycles. Make sure to not use any denoise. We want to see the grain with lower samples so we know the quality is the same.
Edit: I updated the file with one below that is easier to reproduce a Octane scene for.
bliblubli Mind testing the scene I made with eCycles. I don’t have eCycles to test it.
or @dobe if you could test the scene I posted above with eCycles that would be great.
You’re using an RGB Curve node in Cycles, which has no conresponding node in Octane.
A texture is needed. Moreover, in Octane you can use Sphere and Area Lights (or any emissive mesh), not point or distant lights.
This is one of the reasons why comparing two different renders offering different shading system can be difficult.
1 Like
I meant to delete that point light which actually was a sphere light because it had area to it, so it could be replaced with a sphere light in Octane with a diameter of 1.28m. The RGB curve I’m just using for a plane color, but I replaced that too with a straight color now. 36 sec. RTX 2070 super Optix, no eCycles.
grass field no tex.blend (794.6 KB)
Ideally it would nice to have an image with no visible noise, since comparing different levels of noise could be tricky.
By the way I’m pretty sure I can render a noise-free image in about the same time Cycles would still shows the noise in the render.
I’ll play with this scene! Thanks!
Not easy to match the render, especially the ligthing.
This is Octane, with RTX off(17.88 seconds):
Octane with RTX on (6 seconds):
Your Cycle scene (18.82 seconds):
The Octane renders are showing way less noise than Cycles (they’re close to be noise free).
AI Denoiser was not used.
By the way, I’m not sure this is the best scene we can use to compare the render engines.
I’ve used a Daylight Environment in Octane (which means environment+sun), while in Cycles there’s just a Gradient Texture.
That’s my modified scene:
grass field no tex_octane.blend (1.4 MB)
Well… From my experience octane is always faster and especially when it comes to complex+lighting and shading. It’s always also easier to achieve realistic looking materials (but this is maybe just my experience)
Anyways, if you guys want a neat testscene, Rawalanche made one here:
Use an HDRI
basic replicable shaders
no denoising
no GI-AO tricks (they tend to introduce very bad bias in more complex scenes)
fixed bounces
I was tempted to do this comparison myself but A: I have a lot of work and B: For my purposes I already know that octane does better and faster job.
Also keep in mind that clamping values might have different impact on performance and looks of each renderer and other things like scrambling (which is present in both renderers, just in octane is named coherent sampling) also works differently and might work in some cases better than others.
These comparisons might give some rough idea, but also remember it’s not always just render speed. Like I mentioned for me it is way faster and easier to get materials to look the way I want in octane, and in cycles I would always spend more time tweaking and tweaking and also octane has custom lut loader, better glare effects etc… Even if renderspeed of cycles and octane was identical I’d still would work faster in octane overall…
Artist time is always more valuable than rendertime, so if you have an important Paid work you might still wanna push things on renderfarm and pick renderer that makes YOU work faster.
I think LuxCoreRender is more similar to octane than cycles, different feature sets and the speed is not there but when it comes to providing realistic looking output with little effort, it is “faster” than with cycles… Again, unless renderer is extremely slow, or almost realtime, it’s just a personal preference IMHO…
2 Likes
Doing the original scene in Cycles with the Octane one provided I got
37 sec in Cycles no eCycles
15 sec in Octane RTX on
To make the comparison more fair with the same lighting I replaced the scenes with an area light and tweeked the colors till they looked as similar as possible.
Octane 51 sec, rtx on
Cycles 5 sec, rtx on, SSS off
I thought this to be a little unfair to Octane though as it seemed to be doing some sort of SSS calculation. At least I’m going to say it is to make the times more fair. I turned SSS on in Cycles
Cylces 57 sec, RTX on, SSS on, no Ecycles
All the depths were the same for each scene. It also seems the way Cycles and Octane handle the same color is very different so I had to manually change the grass color and fiddle with it till it got to be as close as possible. With eCycles probably cutting the time in half Cycles is the clear winner.
grass field no tex area3.blend (795.3 KB)
grass field no tex_octane3.blend (1.4 MB)
This is the fast render no SSS Cycles scene.
grass field no tex area2.blend (796.0 KB)
Another interesting note is how much RAM each takes. In Cycles 6MB is used when rendering. In Octane it is around 2300MB. It seems Octane plugin is not using the instances right yet.
Octane scene was calculated in 15 or 51 seconds?
I think the best practice would be to make a screen capture of Blender’s Render View, so render times can be read there.
About instancing, it should work perfectly in Octane.
All the times above are correct. The Octane scene you provided to me was 15 sec before the lighting and mats were changed to area light. It was 51 after I changed the mat and the lighting to area. The grass just looked bad after I changed to area light so I changed the mat to look more grass like. Open the new scenes to see.
It is interesting that they are so neck and neck that it can be hard to see a clear winner in all cases. Octane with RTX has sped up way more than when I last tried it. It would interesting to try another scene without SSS which is super hard to match.
If instancing is working IDK why it took so much RAM.
I’m not getting the reason why you added an area light in Octane, by the way. There are no light in the Cycles scene.
Actually I made a mistake on the second picture labeling it as Octane when it should have said Cycles. I fixed it now. Everything is labeled right.
I changed the light in both scenes to only have one area light of the same size and intensity so the lighting could be exactly the same. I just downloaded and opened both new Cycles scenes and they do have an area light.
I noticed you’ve used PMC Kernel in Octane. Why? You should use Path Tracing, which is way faster, still unbiased. PMC should basically never be used.
1 Like
The Path Tracing was taking longer to get rid of the grain. PMC was faster. Try it yourself.
1 Like
switch to pathtracing and set caustic blur to 1 on octane.
1 Like
Good tip. It makes the caustics less accurate, but cuts the render time to 30 sec. This should be close to the ecycles time.
Well, first of all this is a terrible scene to compare anything…
Another problem is totally black environment for an open scene
then you have a specular material with saturated green color set as translucency on a bunch of singlesided objects, which is the source of strong specular caustics and slow rendering.
Everything is just absolutely wrong about this scene setup 
1 Like