OctaneRender™ 2019.1 XB2

[Notiusweb]

I made a Ginormous discovery -
When I just use only my main Titan Xp on the Mobo, I get NO issues with any of things I have been yapping about related to crashing-
1) Saving EXR, 16 bit PNG
2) GPUs going orange-warning when changing AI-sampling modes
3) not being able to use Render passes with up-sampling

BUT - as soon as I enter in either my 1 Titan X Pascal, which is also on Mobo, or my 6 external Titan X Pascals connected on PCI 1x USB 3.0 risers, I get the issues...

In fact, with just 1 Titan X Pascal, either on Mobo or external, things fail. For example, with render passes when using up-sampling GPU gets the red warning that it failed.

Almost as if XB2 can't work right with Titan X Pascal.
Now, XB1 had no Titan X Pascal issues, so this is very interesting, weird, and most importantly...

FUN

Thanks, this is helpful. I am using a Titan X Pascal exclusively for rendering so it is likely something else. We will figure it out.

In V3 you guys made Tonemapping checkbox so we could just use 1 and avoid all the GPUs from handling it (some of us were getting crashing otherwise when Post Processing was enabled...
maybe Up-sampling could work smoother if run only off 1 select GPU?...

I am using NVidia 425.31, which now isn't the most recent....let me know if you guys think another driver is safer.

[Notiusweb]

Hello, thank you for the awesome innovations, am always happy to give feedback.

1) Render Pass Crash, with Up-Sampling
Okay, here is a test scene where I can reproduce XB2 issues while in Up-Sampling mode

XB2_Cube_RenderPasses.orbx

(22.29 MiB) Downloaded 196 times

Main pass works

Then GPUs go Orange when selecting the Pass tabs

If I try to go back to Main, it gets weird and just crashes, so I have to stop the scene and restart the render

(2) Render Pass Fringe
While I am at it with the layers, it seems that the pass layers do not ever completely mask out the Main. For some reason there is always a fringe:

For example, at 1080p I render these:
Main Pass

Render Layer Mask

Which yields these

But what happens is that when layered in Premiere Pro, or Photoshop, we get a fringe outline

Now, there are some workarounds for this in the 3rd party applications themselves, but in this case of layer-compositing, what I am concerned with here is that Octane's own produced images are yielding the non-functional result.
I am not focusing on how to use 3rd party apps to make it work. I am looking to see that Octane itself produces the right images to begin with, so that workarounds will not be required.

*Also, Pre-multiplied Alpha and Disable Partial Alpha do not do anything for compositing a main pass with a render pass, they only can help when using Octane's alpha images over a background.
Not only that, but camera-imager does not affect the pass layers at all, so you couldn't somehow tweak the color-space of the white vs black on the pass even if you wanted to...

Can we see if the layer passes can better match and mask the main pass, the passes are awesome but then are problems to work with.

THX!

[enricocerica]

I noticed something weird and would like to know if someone else noticed the same behavior.
Increasing the amount of Parallel samples uses more memory but doesn't increase the amount of samples/s.
I'm using DL but noticed the same with pathtracing.
Any thought about that ?

[Notiusweb]

I noticed something weird and would like to know if someone else noticed the same behavior.
Increasing the amount of Parallel samples uses more memory but doesn't increase the amount of samples/s.
I'm using DL but noticed the same with pathtracing.
Any thought about that ?

I play with these a lot (Tiles and Parallel), and I notice working/normal functionality in XB2.
I will note when I use I adjust both of the sliders more so in concert with each other.
I use these to either get more acceleration (for frames) vs higher top speed (for single higher fidelity renders at higher sample targets).

But, sounds like you see no impact other than a higher mem usage, right?

[wallace]

A weird layer related bug. Both meshes have the same material, which uses a material layer. One renders purple while the other is black

NOTE: If I set this up using a layered material, it works OK. I'm using a base material with specular layer plugged into the material layer pin and this seems to be causing the bug.

I can confirm thin film Layer is broken for legacy Glossy Material in this release, but works fine for other materials.

I have fixed both of these bugs, and they (the fixes) will be in the next release.

[funk]

Universal material bug: Adding a coating to a highly specular color base, makes the clearcoat look like a mirror instead of a clearcoat

top = coating 0, 0, 0
middle = coating 0.5, 0.5, 0.5
bottom = coating 1, 1, 1

Load the orbx, select the sphere > universal material and change coating color to see the effect

[funk]

Universal bug: There is a blue "tint" in this material when albedo is set to 0. If albedo is set to 0.1 it goes away.

It seems like albedo of 0 is creating some transparency/transmission?

uni_trans_bug.jpg (13.55 KiB) Viewed 3048 times

Load the orbx, select the sphere > universal material and change albedo from 0,0,0 to 0.1,0.1,0.1 to see the problem vanish

[funk]

Anisotropic rotation is broken on the Glossy Material. It doesnt do anything at all

[coilbook]

Hi otoy,
sometimes large scenes cause BSOD on windows 10 with the following bugcheck. The computer has rebooted from a bugcheck. The bugcheck was: 0x00000133 (0x0000000000000001, 0x0000000000001e00, 0xfffff8004524c378, 0x0000000000000000).

I think it is related to octane, nvidia and large scenes with particles like phoenix fire/smoke. Sometimes slaves have this error, sometimes the master. I can send a dump file if you think it can help you.

[divasoft]

Hi guys. Can you give an approximate release date with RTX support? I replaced almost of my 1080ti with 2080ti and now I'm working at much lower speed than before.
RTX support is very needed as fast as possible.
Thanks.