Arnold Optix denoise issue

[Anonymous]

Hi there,

I started experimenting with the Optix denoiser and came to the following issue.
I want my beauty pass to be denoised.
So I add the RGBA pass and check on the denoise box.
From that point on the optix works perfectly together with IPR.

BUT

When I want to render out a still image. I first get the following error:

// Error: [aov] more than one output statement was associated with single layer driver: defaultArnoldDisplayDriver@renderview_display //

the solution for this is to simple rename the aov but the moment I rename the aov,

it loses connection with the data from arnold for some reason because I don't get the beauty pass anymore.
It renders out blank images. Even in the IPR i dont get the correct result.

Is this a bug or am I doing something wrong?

tech info

(running maya2017 - and Geforce 1060 with latest driver)

[Anonymous]

To clarify this a bit extra:

the optix tutorial says:

3. If you wish to have the same AOV saved to EXR in both original and denoised mode, to avoid name duplication you can add the suffix "_denoise" to the AOV name and the Optix denoiser will detect the original AOV.

But the denoiser doesn't detect the original aov.

If I rename the RGBA aov to fe. RGBA_denoise.

I get in my renderfolder the following:

-beauty (which looks ok but noise)

-RGBA_denoise (which is completely empty)

-RGBA_denoise_denoise (where there is little information in but unusable)

[Slava_91]

Hi all! So as not to create a new branch, I will add here a question too.

Where does the optix denoiser take the information at the black point?

https://i.imgur.com/uaCkMhz.jpg

https://i.imgur.com/2RIy43E.jpg

[Slava_91]

you don't need create duplicate aov with denoise suffix, this automatically create duplicate to all aov with denoise attribute in driver

[mikinik]

@Slava Sych, what gpu do you have? I have the same problem in 3dsmax on kepler card. + I'm also concerned about the white spots on geometry and ugly stroke along the contour of geometry.

[Anonymous]

Yes,that's just a bug.
This workaround is to add RGBA AOV with any other AOV as a dummy.
I confirmed this bug was fixed on MtoA3.0.0.1.

[madsd]

You need AA passes, so lower individual samples and up the adaptive samples.

[mikinik]

@Mads Drøschler, it seems to me that there is something wrong. This is normal?

[madsd]

yeah, I've seen it all as well to some degree.
I can only recommend you to try the stand alone denoiser noice.exe - it is far superior to OptiX denoiser in my initial tests.

[Anonymous · Accepted solution]

Big thx for the 3.0.0.1 update this fixes my problem!

The speed of your support can only be adored!
Love it!