We've got a modest renderfarm here with 15 blades in the rack with between 16 and 24 gigs of ram, 8-16 cores, and we're seeing some absolutely random rendertimes with our huge scenes. Some matte passes and beauty passes can take only 10-30 minutes to render, but some batches on our farm will randomly decide to take 1-10 hours (not exaggerating).. If I kill that render and start it again (sometimes on the same machine), it may sometimes work fine and render fast again. Sometimes in a multiple-frame batch, 2 of 3 frames for example might render fast, then somewhere in the middle of the last frame, it will slow down horribly (with completely wrong render estimates too).
What sort of issues would cause this sort of randomness? The systems are kept in quite good shape and despite our huge scenes have plenty of RAM and HD space to take advantage of. Our dynamic memory limit for these scenes is set anywhere between 4 and 10G typically. (Is there a good way to calculate the best setting for a scene?)
Thanks for any answers, we've been thrashing through this for quite awhile now with no idea how to get consistent rendertime and not hang.
What sort of issues would cause this sort of randomness? The systems are kept in quite good shape and despite our huge scenes have plenty of RAM and HD space to take advantage of. Our dynamic memory limit for these scenes is set anywhere between 4 and 10G typically. (Is there a good way to calculate the best setting for a scene?)
Thanks for any answers, we've been thrashing through this for quite awhile now with no idea how to get consistent rendertime and not hang.
Comment