View Full Version : Nodes rendering in succession instead of simultaneously?

07-21-2007, 08:08 AM
Just set up a small render farm with 5 CPU's. Got everything working but while watching the network interface from the master drive, the first CPU renders a frame, then the second and so on. It seems like there should be a continued flurry of activity with all CPU's pumpin at once? Does anyone know - Are they actually rendering like they're supposed and just displaying this way or is there something that still needs to be set up [say in a config file] that enables all CPU's to render simultaneously? This is a first attempt with setting up distributed rendering. Thanks for any help.

07-21-2007, 09:51 AM
From what I've observed, the command to load a scene is sent sequentially to each node, as opposed to all of them simultaneously. Probably to cut down on network traffic, or to stop multiple computers trying to read the same files all at once. Then once all computers have loaded the scene, the command to render is sent.
Is that what you mean?
Remember, screamernet is quite old in it's implementation - I don't think it's had a major upgrade in many many years. Third party controllers probably do things another way to speed up the process.

07-21-2007, 10:35 AM
Actually, I think ScreamerNet has been tweaked at somepoint, probably for the reasons Voriax suggests, and I don't think it's as responsive as it once was.

However, if you have a basic setup running, I'd spend the $50 and pickup Butterfly Net Render Personal. That'll run your 5 nodes far more efficiently than LightWave's built in controller.

Depending on the overhead of your scene you could have each node render multiple frames before coming back for more.


07-21-2007, 11:51 AM
Thanks. I just loaded a heavier scene and am going to run tests to see what happens. The first scene was extremely low poly and I think it was just that it rendered so quickly that it just appeared that the frames were being rendered sequentially. With a substantially more detailed scene, I'm hoping this won't be the issue.