OctaneRender 2021.1 LightWave EOL Tribute build

[promity]

Here is the archive. Simply change material in the Gold Spheres surface to Diffuse with Emitter.

https://disk.yandex.ru/d/jMrt4J3z-FCsuQ

[BorderLine]

Octane and Lightwave are frozen immediately too !

[smount1]

2021.1 LightWave EOL Tribute build works fine for me. thnx

[KVNOODLES]

H there,

I wanna just report a bug with Octane 2021.

The denoiser seems to be broken when you're using an RTX card (2080ti) and a TitanX Maxwell in the same computer.
Precisely,the denoiser starts and stay blocked...
I've tried in the plugin option to desactivate the TitanX, but the denoiser stay blocked too.

However Octane 2020.2.5.1 works perfectly with my config and denoiser achieves his task perfectly.

Probably an issue related to the standalone...

[dough-cgi]

Anything new about the AI Light Problem, I encountered? (my post from Monday)

Could it be a drivers or system issue. Here it seems work fine even with 30x30x30 instances. What GPUs and NVIDIA drivers do you have?

Thanks,
-Juanjo

Totally missed your reply, sorry. I have 472.39 Studio Drivers, 3080 TI.
I am going to update to WIN11 after current projects are done and look into that Light AI problem again then. Everything else seems so work ok.

Thanks
Roman

[BorisGoreta]

Two 3090s in the node throw out of memory errors.
If only one of them is active there is no out of memory error.
Node has 32GB of RAM and it doesn't waste all of it when two are active.

[juanjgon]

Two 3090s in the node throw out of memory errors.
If only one of them is active there is no out of memory error.
Node has 32GB of RAM and it doesn't waste all of it when two are active.

Hmm, 32GB of RAM for two 24GB GPUs is not enough for sure unless you have a huge fixed system paging file of at least 64GB. Be aware that the Nvidia drivers try to reserve all the GPUs memory also in the CPU memory for memory mapping even if it is not used. You can try to configure a fixed system paging file of at least 64GB to check if it helps.

Thanks,
-Juanjo

[BorisGoreta]

It doesn't matter how many 3090s are in the system, RAM usage is the same for 1 or 2 3090s, I've checked.
And they consumed only 18GB of RAM, started rendering and after 20 seconds, out of memory.

(this happens only in the node, in master computer 3090s render fine)

[juanjgon]

What errors do you get in the logs?

Thanks,
-Juanjo

[BorisGoreta]

It's a regular out of memory -> failed to allocate device array error.

Funny thing is that it renders for 20 seconds and then throws out this error.