Experiencing CUDA errors every 300~ renders using Octane. Trying to learn more and find clues

I understand that you guys can’t really speak for any third-party products, but I wondered, maybe there’s something about how Blender is built that results in my errors, so I thought I’d just ask.

I’ve got a big file full of disabled Collections (containing models with textures) and a script which enables them in different combinations and renders them with:

bpy.ops.render.render(write_still=1)

It renders an image → enables collections → renders an image → enables collections… etc. in one big for loop.

This error appears more or less consistently after 300~ renders.

CUDA error 700 on device 0: an illegal memory access was encountered
 -> failed to create link

Which some Octane forum threads describe as VRAM failure related to too big textures/geometry. And I started wondering, maybe that’s somehow related to my setup, maybe my GPU’s VRAM is being consistently filled up and I need to flush it or something? Why else would the errors appear in consistent intervals? I’m not sure how all that works, I’m just going on a hunch here. I’ll appreciate any insight.

Anyone who renders animations on a regular basis should absolutely own B-Renderon. It has the added benefit that it could possibly allow you to easily overcome issues like these. It renders Blender files without having Blender open, and you have full control over frames to render, which GPUs to use, etc. I implore you to look into it bro.

1 Like

That’s pretty cool, haven’t heard about it.

Unfortunately it’s not a solution to my issue. I’m running a custom script that does many things in there. Thanks though.