Tweet
*This is using 1.47.03 (I know, we need to upgrade, we're working on it...)
We have a mixed renderfarm, some are old server machines not upgraded, some are new server machines, and then we have the occaisional overnight-only upgraded desktop. These range from dual processor to dual proc quad core. What I am experiencing with our current project (QMC Prim, LC secondary), is that it renders fine on the OLD machines and crashes on the NEW ones.
When watching it crash on the new ones, it appears to be running out of memory. I noticed that all of the ones that crash, are reported as having 4 processors according to backburner. The only machine with more than 2 processors that is not crashing, is the dual proc quad core that is running 64bit OS. (though with 32 bit Max
I've found that by disabling hyperthreading on the new machines it allows them to render the scene fine(and of course the # of procs in BB goes down to 2).
The light cache passes are set to 4, I'm wondering if this is causing the dual proc machines to not load as much memory as the ones that can do all four at once, and is why they can render in static memory mode while the higher-end machines cannot?
I expect the solution to get all machines rendering(without disbaling hyperthreading on them) is going to be to set the scene to render in dynamic memory mode. I'm just curious if my theory on why it renders fine on the older machines is correct.
Thanks for any info you can provide.
We have a mixed renderfarm, some are old server machines not upgraded, some are new server machines, and then we have the occaisional overnight-only upgraded desktop. These range from dual processor to dual proc quad core. What I am experiencing with our current project (QMC Prim, LC secondary), is that it renders fine on the OLD machines and crashes on the NEW ones.
When watching it crash on the new ones, it appears to be running out of memory. I noticed that all of the ones that crash, are reported as having 4 processors according to backburner. The only machine with more than 2 processors that is not crashing, is the dual proc quad core that is running 64bit OS. (though with 32 bit Max
I've found that by disabling hyperthreading on the new machines it allows them to render the scene fine(and of course the # of procs in BB goes down to 2).
The light cache passes are set to 4, I'm wondering if this is causing the dual proc machines to not load as much memory as the ones that can do all four at once, and is why they can render in static memory mode while the higher-end machines cannot?
I expect the solution to get all machines rendering(without disbaling hyperthreading on them) is going to be to set the scene to render in dynamic memory mode. I'm just curious if my theory on why it renders fine on the older machines is correct.
Thanks for any info you can provide.
