AWS Thinkbox Discussion Forums

Brazil R/S restarting with each frame rendered

We've seen some really poor turnaround with short (~30sec) brazil-rendered frames on our render farm.  Brazil spends about 2-3 minutes loading up on a slave before it renders each frame and then purges its memory footprint before ramping up for the next frame in the task.

I'm wondering if anyone knows of some switch that 3dsmax or deadline is passing to brazil that forces it to binge/purge/reload between each frame in a task.

When I render the same short particle effect sequence locally with a 3dsmax workstation, Brazil does not take 2 minutes to binge/purge between each frame and the sequence will render in a 10th of the time.  Brazil loads once and retains its memory footprint, which is the desired behaviour.

This 2-3 minute load/reload between frames doesn't happen with 3dsmax/scanline using deadline and multiple frames per task... just Brazil R/S.

Has anyone encountered this before?  Obviously it's only a problem with short renders, but 3 minutes of overhead per frame is pretty bad and I'd like to get the slaves running optimally.

Privacy | Site terms | Cookie preferences