Swarm delay?

[njl_aklestad]

We have a main computer with 2 swarm nodes, gbit network and 32 cpu cores each.

So there are 80 cores available for rendering.
However, during the first 1/3 of the time it takes to render the photo only the 16 local cores seems to be 'dancing' on the screen. Then during the last 2/3 of the rendering process all cores are in action.
I understand that the local computer needs some time to prepare packets and distribute them to the swarm nodes over network and then receive them back after processing.

But is there any way to optimize this process to make the nodes kick in quicker?

[Peter.Chaushev]

Hello,
While the render process is distributed, its prepasses (e.g. Light Cache calculation) are not.
This means when the Render button is pressed each machine loads the scene and its assets, independently calculate LC and then join together for the actual render pass.
The reason only the "buckets" corresponding the local machine initially appear is due to the fact this machine has a head start when it comes to loading the scene. Render node machines first need to accept the job, transfer the vrscene file over the network together with all associated asset files, and then start rendering.

The speed at which render node machines join the render process depends mostly on the size of the transferred scene files and LAN speed.

Kind regards,
Peter

[njl_aklestad]

After upgrading to newest vray version, it seems like the swarm nodes start working a lot quicker. So that is very good!