OctaneRender™ Blender 2.92 plugin version 2020.2 - 23.5 [STABLE] released

The 1:23.87 was when I turned Optix on. 2:07.39 was when Optix is off.

I re watched the video. 80 GPUs to render the car in real time. There has to be something with it or they would release it to the public to test out. I also noticed in the room scene they say 60 fps, but when he move you see grain in the shelf area. Can’t call it 60 fps if there is grain. Maybe 3 fps I would say if you don’t want grain.

1 Like

Sounds about right. Did you make sure the diffuse, glossy, scatter depth was the same in Cycles as in Octane? There is a slight difference in the ratio of scene times from mine. That might be it.

No I didn’t, but I did try adaptive sampling, just to see if I could get the noise down a bit. Didn’t make much of a difference with most of the room being indirect light.

The car is from a very old demo from 2015. Today we have new cards and RTX technology.
After the “old” Brigade intro videos, everything else you see is fueled by a single RTX 2080.
I’m not going to further discuss about Brigade, since the video is pretty clear about it.
Jules is very clear about its features.
And when it will be release anyone will be able to judge by him/herself. I personally can’t wait to be able to work on my 3d projects taking advantage of Brigade.

Here a detail comparison performance and image quality for Cycles/E-CyclesRTX/Octane using RTX for all. Samples are 2000 and best tile size with a Nvidia 2x2070. Using the same HDR from Corona original test. See link at HDR link.

Cycles RTX time of 1:55.04 sec:

E-Cycles RTX with noise level/scrambling distance at 1.0: 1:28.52sec

E-Cycles RTX with noise level/scrambling distance at 0: 0:44.00sec

Octane with RTX Path trace: 0:43:09

Image Quality comparison:

Notice the quality of E-Cycles RTX at scrambling distance(noise level) of 0. Has the least GI quality. Even thou it is subtle the edges between ceiling and wall or inside the cabinet have less definition an are slightly blurry. This lack of definition is part of the artifacts caused by using scrambling distance. It has a faster time at 44 seconds than Cycles, but you lose quality.

Cycles and E-Cycles RTX at scrambling distance(noise level) 1 the image quality is identical. They have the second best quality GI notice better edge definition, but the have more noise. Here the advantage of E-Cycles RTX at 1:28 seconds is much less compare to Cycles at 1:55 seconds.

Octane has best quality GI. Notice how clean the GI edge definition is and it has the least amount of GI noise at time of 43 seconds.

Besides GI quality another part of performance that Octane does better than Cycles is shading. Having heavy scenes with lots of shaders usually Octane will be even faster.

4 Likes

I hope that brigade will also have all the features from octane…
Things like the fast texture displacement, round edges or dirt map are super useful.

I agree and as we are both involved in each project (I noticed after posting you are part of the Octane team), it’s better to let user make their mind themselves. Most have both E-Cycles and Octane and are specialist in their fields. I’m still happy to get a real use case file (something you would proudly deliver to a client, like that) were Octane would be faster than E-Cycles if you or someone else has some.

Yes, the fact I work for OTOY is pretty public, and I’m here to help users with Octane for Blender, since a user pointed out to me this thread about it.

Not only I work for Octane, I’m also responsible (among other things) of the development of Octane for Blender; again, I’m here to help users with any issue or questions they may have about it, and for sure I don’t have anything against Cycles or E-Cycles.

This is not about being biased by our own position. Quality and render times are quite a misurable and objective thing. 2+2 it’s 4 for both of us and for any artist here.

You’ve mentioned some videos and threads where users were making comparisons between Octane and E-Cycles, using a very old Octane version than the one currently available (at that time) and basically the wrong settings, leading to “false” conclusions.

And, as you mentioned, users of course have to “make their minds” by themselves (and they do, no one is forcing them to use one render or the other), but they also need their conclusions to be based on fair comparisons.

While I don’t think E-Cycles should even be a topic for this thread, however we’re both here and we can take a look at the way users are comparing Octane, Cycles and E-Cycles; I think it’s very important to help them get the best from both renders using the best possible settings for a fair comparison.

Being Octane free for one GPU, any user can download it and test it.

As someone mentioned here, Octane IPR feedback definitely feels way smoother than Cycles’s, building scenes using Octane is quite an amazing experience.

Is Octane faster than Cycles in any situation? Definitely yes.

Is Octane faster than E-Cycles in any situation when we talk about final render times?

As you say, let the users make their own test. But also help them tweaking the settings of both renders so to have a fair comparison.

If you notice that a user is not using the best settings to get the best render speed in E-Cycles, please don’t hesitate to inform him/her about it. I’ll do the same for Octane.

Also, it would probably be nice to have a dedicated thread about it, since I would like this one to be more focused on questions and clarifications about Octane for Blender.

2 Likes

I totally agree we should help our users and that’s why I will also let you in peace in your thread. But I think it’s fair after your answer to this question, which has a very bold statement, to ask for facts to backup this statement.

Regarding helping users, this discussion started with @dobe coming with an issue of your engine being 3.5 to 5x slower than E-Cycles in his use case with an RTX 2070 and he ask for tips to speed-up your engine. You answer with kind of “we are the fastest in my experience” just facing a reality with words. Then kind of “with 2x 2080Ti it’s fast” and no single help on how to use your program. I’m more of a coder, but I think it’s worth questioning if that was a good answer to his issue.

Anyway, last post here, you can continue with any statements in your own thread and I’ll not react anymore I promise. Anyone providing a real use case were Octane is faster or higher quality than E-Cycles is welcome to contact me :slight_smile:

I stated what my experience was after tweaking the settings and verifing the render time/quality of the scene mentioned in one of the links you provided (this one).

No, I “faced a reality with words” that are consequence of “reality”, as I stated above.

I’ve been answering any user question, here, on the Octane forums (this is the specific thread with the same grass renders posted here) and on Twitch.

I also asked the user to share his scene or just post his settings.

This is not “my own thread”, but a community thread about Octane for Blender. And of course you’re very welcome here as anybody else is.

I have no time (and really don’t care about, since I don’t even consider the Octane vs E-Cycles topic really “central” when considering Octane) to build “comparative” scenes to “measure” the render capabilities of the two. For sure I’ll be available to “tweak” the settings and the material of any submitted Octane scene to help improve quality and render times, which is the most important thing to me.

Sorry for the late reply.

I didn’t want to start a render comparison or war here.

I have been working with octane for a long time because I like the quality out of the box more than that from cycles -> personal opinion!

I only created the test because suddenly I had unusually high rendering times with Octane.

Unfortunately, I can´t make the file public here because it contains partially paid assets. If someone is interested, you can write to me personally.

Note: I also work with e-cycles from time to time. I really like the speed! BUT when it comes to complex scenes I reach the limits with cycles. It doesn’t look as “crispy” as in octane. Filmic always makes the rendering in cycles a bit bland. -> for my taste. This is the reason why I use Octane. Speed ​​is secondary.

1 Like

@linograndi Not sure this is a good place to ask ,but im using the free tier version of octane (latest one available), and im now constantly getting this error when I try to render my scene:
CUDA error 700 on device 0: an illegal memory access was encountered
-> failed to wait for event
device 0: path tracing kernel failed

googling the issue the answer seems to be about the classic TdrDelay, which I have already fixed in my regeditor. I also have the latest studio driver for my rtx 2070S.

I have no issue rendering this scene in cycles.

It seems to specifically crash in Octane when adding one more area light to my scene. Before I do that, it seems fine to move around with the camera and preview the render. Its not a heavy scene at all, just a room with a few props, most of it is untextured at the moment as i was trying to remake my old cycles scene in octane.

Edit: OK i switched off RTX, and now i dont get the crash. RTX works fine in cycles however.

I’m curious: when using Octane then you don’t use Filmic? What do you use?
I’m asking because I love Cycles but I too find Filmic sometimes not quite what satisfying. From my little (very little) understanding, ACES is the next big thing and from what I’ve seen around the results are much better than with Filmic. That said, due to my profound ignorance in this regard, I don’t even know if I’m comparing apple to orange here :rofl:

I’ve created a new thread to take the heat of comparisons off of this thread. Please post there for anymore comparisons.

@linograndi I didn’t know you worked at Octane. I think it’s great you being here answering peoples questions. It is generous that Octane works with 1 video card forever on Blender. Thank you for the Octane plugin. It’s a huge amount of work to make a plugin.
I would like to ask if the rotation of hdris is working or maybe I hooked things together wrong?
It only rotates in what should be the Z direction which seems like the X direction. The top two slots under Rotation don’t seem to do anything.

@eklein The difference in the render time between the Cycles w RTX and Octane w RTX seems way off of what it should be. I copied the hdri and got Cycles w RTX to be less than 2x the time of Octane RTX. I was thinking maybe it is the two cards, but both engines should see a linear speedup with 2 cards. If you want to post your stuff on the new thread we could talk about it more there and try to find the problem.

i think you need to hook up a spherical projection to the hdri, to get that to work properly (might be wrong) but thats how i do it.

1 Like

Yes so Full Transform hooked to Spherical Projection hooked to Projection seems to work. TY

I have this error only if i enable both rtx and ai lights. If I keep ai lights off it doesn’t happen even when rtx is on.(i contacted their tech support and they said they would investigate further)

Anyway, I also noticed that with rtx the wireframe pass doesn’t work and dirt map was giving me some weird issues too. I hope they fix all the issues so that the rtx can be more reliable.

1 Like

This is currently the best possible setup for full HDRI transformation control:

Thank you! At your service.

1 Like

I had the same issue with AI Lights and RTX, which seems solved in the current internal build.

I can’t reproduce this issue right now. Would you mind sharing a scene showing the problem?
I really want to be sure most issues will be fixed in the next build. Thanks!