OctaneRender® 2022.1.1 Tweet

[john_otoy]

When denoiser is enabled it doubles render time twice compared to regular render.

The denoiser will increase render times, but it should not be doubling them unless you have a relatively fast render in the first place.

Do you have a scene we can look at? Please PM me.

[J.C]

When denoiser is enabled it doubles render time twice compared to regular render.

The denoiser will increase render times, but it should not be doubling them unless you have a relatively fast render in the first place.

Do you have a scene we can look at? Please PM me.

Here is the scene. Without denoiser it renders 6 sec. With denoiser it renders 12 sec. This is pretty strange as for example with Cycles the denoiser does not slow down render times.
https://www.dropbox.com/s/a8zryxf0tnhje ... .orbx?dl=0

[nuno1980]

Here is the scene. Without denoiser it renders 6 sec. With denoiser it renders 12 sec. This is pretty strange as for example with Cycles the denoiser does not slow down render times.
https://www.dropbox.com/s/a8zryxf0tnhje ... .orbx?dl=0

+1!! I like Cycles for not slow at denoiser enabled. But unfortunately, Cycles still lacks "dispersion coefficient" and "nested dielectrics" options for transparent/specular material.

[john_otoy]

When denoiser is enabled it doubles render time twice compared to regular render.

The denoiser will increase render times, but it should not be doubling them unless you have a relatively fast render in the first place.

Do you have a scene we can look at? Please PM me.

Here is the scene. Without denoiser it renders 6 sec. With denoiser it renders 12 sec. This is pretty strange as for example with Cycles the denoiser does not slow down render times.
https://www.dropbox.com/s/a8zryxf0tnhje ... .orbx?dl=0

Thanks for the scene. I ran some tests on a 4090 and my results aren't quite 2x slower, but we'll look into it.

2022.1.1 / 4090 / 200 max samples
00:05 no denoiser
00:08 with denoiser (1.6x slower)

I also increased the samples to get a more accurate result.

2022.1.1 / 4090 / 2000 max samples
00:45 no denoiser
01:06 with denoiser (1.47x slower)

[J.C]

On my PC with 4090 I get:
5 sec – 200 samples, no denoiser
12 sec – 200 samples and denoiser
This is 2 times slower

60 sec – 2000 samples, no denoiser
1:30 sec – 2000 samples and denoiser
This is 1.5 times slower

In both cases it still takes too much time and when denoiser is on scene editing also gets slower and less responsive.

[Obizzz]

On my PC with 4090 I get:
5 sec – 200 samples, no denoiser
12 sec – 200 samples and denoiser
This is 2 times slower

60 sec – 2000 samples, no denoiser
1:30 sec – 2000 samples and denoiser
This is 1.5 times slower

In both cases it still takes too much time and when denoiser is on scene editing also gets slower and less responsive.

A similar pattern here with 4 and 8 seconds for 200 samples and 30 seconds vs 60 seconds for the 2000 sample render.

I tried 4000 samples too and got 1 minute and 1:57 so it's pretty consistent at doubling the render time on this scene which pretty much defeats the purpose of using a denoiser in the first place.

The denoiser used in Octane is lacking in both speed an quality compared to more modern offerings such as the Open Image Denoiser Blender uses for Cycles.

Viewport responsiveness is awful in Octane once you turn on the denoiser too.

Edit: Using a 4090 and a 3090 via network rendering. I let the scene load properly and just switched denoiser on and off so scene load time is not included.

[J.C]

The denoiser used in Octane is lacking in both speed an quality compared to more modern offerings such as the Open Image Denoiser Blender uses for Cycles.

Viewport responsiveness is awful in Octane once you turn on the denoiser too.

Edit: Using a 4090 and a 3090 via network rendering. I let the scene load properly and just switched denoiser on and off so scene load time is not included.

Couldn't agree more. Guys from Otoy, could you please fix this?

[PolderAnimation]

I did a test on my current scene:

low resolution 4000 samples

14 sec with Spectral AI Denoise
30 sec without Spectral AI Denoise

full hd 4000 samples

78 sec with Spectral AI Denoise
69 sec without Spectral AI Denoise

On a 4090 with octane 2022.1.1

[coilbook]

On my PC with 4090 I get:
5 sec – 200 samples, no denoiser
12 sec – 200 samples and denoiser
This is 2 times slower

60 sec – 2000 samples, no denoiser
1:30 sec – 2000 samples and denoiser
This is 1.5 times slower

In both cases it still takes too much time and when denoiser is on scene editing also gets slower and less responsive.

That's why we don't use denoiser anymore. It's not worth using. I hope they can fix it where denoiser requires 0 extra time to render.

[PolderAnimation]

If I render my groom (curves) and I disable shadow viibilty, my render becomes 35 times slower. (I render with path tracing kernel and alpha shadows on)