8k better 12k render

[ricorio]

Hi,

I'm trying to render a image with 8000 x 8000 px.
There is not a lot of geometry (120000 triangles) in the scene and I got 32gb of RAM.
My primary bounce is irr map. Secondary brute force. I also set the dynamic memory limit to 30000 mb.
In the beginning the RAM fills up to 30 gb then the swapping starts. After about an hour the RAM usage is down to about 10 gb and the cpu load is between 12 - 50 %.
Rendering in standalone.

Is there anything I'm doing wrong or a way to improve the settings to keep the RAM filled up and the cpu load at max?

Thanks for your help.

rico

[Neilg]

Are you using displacement? Any plugins? Are you rendering to the VFB? Going directly to disk will help in this case.

You could also precalc the IR on a lower res image, save it, then send it in 10/20 strips through backburner so it doesn't have as much to do for each.

[Morbid Angel]

render to vray image format without the framebuffer.

[ricorio]

Thanks for your help guys!
There is no displacement, I'm using the VraySimbiont SHD for some xray effect.
Now I render in standalone to vrayimg with display ="0" (no VFB) but the same thing starts to happen.
The RAM fills up with a 100% cpu load then it starts to empty the RAM and the CPU load is mostly about 12%... sometimes 50 %.
The render eats up the RAM very fast in the area file textures are used ... should I convert them to some other format ? Anyway there are only 2-3 Textures involved.
Is there a way to tune my BSP settings so the render doesn't need so much RAM?

cubiclegangster: I would like to do the precalc of the IR in low res. But how do I sent it in 10/20 stripes though backburner? (deadline)?

Thanks again!

Rico

[ricorio]

Well ... the 14 k image is rendering now with not even reaching 4 gig in RAM.
Stupid me ... the scene itself didn't show a lot of polys because everything was in smooth preview.
So I converted all the preview subds to smooth polys and exported each as a proxies.
I think now I could probably render a 50k image with not reaching the swapping of RAM!!!

cheers

rico

[vlado]

You could also have limited the subdivs for displacement and subdivision surfaces to something like 4 for 8. But proxies will probably render faster anyways...

Best regards,
Vlado

[grantwarwick]

was going to say...

Render 10-15k images all the time and vray optimised ram very efficiently. Never in my entire career have I had it hang as you described :/