Out of core rendering

[coilbook]

Hi,
is out of core rendering for textures or meshes?

[paride4331]

Hi coilbook,
Textures, but keep in mind that setting up the System RAM Usage Limit, you can also handle how much system memory OctaneRender uses for out-of-core geometry and textures. Once texture or geometry data is placed on this amount of RAM and the out-of-core feature kicks in.
Regards
Paride

[glimpse]

out of core is for textures, if You want to handle more geometry, look into NVLINK (if You have two of the same GPUs that support it)

[glimpse]

out of core is for textures

OOC geometry is actually supported since v4

have to try that with smaller GPU to see how things are running lately (upgraded GPUs not to run into out of memory situations), however what I noticed in the past year or two with bigger scenes that are heavy in geometry is that they run into issues pretty fast and out of core does not help. But will do some test for that to see what's the current state.

[paride4331]

Hi elsksa,
in order to reduce this issue, you could try setting all geometries to "movable proxy" and reduce geometry groups as much as possible.
Regards
Paride

[coilbook]

Hi coilbook,
Textures, but keep in mind that setting up the System RAM Usage Limit, you can also handle how much system memory OctaneRender uses for out-of-core geometry and textures. Once texture or geometry data is placed on this amount of RAM and the out-of-core feature kicks in.
Regards
Paride

Hi Paride,
There is still a bug with out-of-core memory and slaves. On my master GPUs I can easily go over and have 5-7 gb ram used for out-of-core. Slave GPUs (each PC holds 10 GPUs) crash saying no memory even though each slave has 16 gb for out-of-core assigned to them. All GPUs have same vram size. is it a bug? Thanks


PS Here are the error slaves always get

[Rik]

Wondering if this is something to do with the issue I'm having with Vram.
Never had an issue previously but now I'm having to reduce parallel samples down from 32 to 1 when using models which previously ran fine at 32 parallel samples.

This isn't great as it increases render time. I imagine there would also be some time cost when using OOC too.

It feels like Octane memory usage has become more inefficient than it was. Am sending Paride a file to see what's going on.

Weirdly, changing the display level of a proxy to original now seems to use up Vram and cause a render fail.

[paride4331]

Hi,
Could you please to try setting up the Octane Daemon again, making the same setting for each render node?
In any case consider that just like in a multi-GPU setup, in network render, the amount of RAM available to OctaneRender is not the total amount of VRAM from your GPUs, but it is the amount of VRAM from your smallest GPU.
Let me hear from you.
Regards
Paride