If the scene is rendering properly on your master computer but on the nodes you get errors it might be 1 of these 3 things:
1. There is not enough GPU RAM in the nodes GPUs.
Yes you have the same 6GB cards as in the workstation but on the workstation you use dedicated display card and render on the rest of the GPUs which means those GPUs have all 6GB at their disposal while in the nodes you don't have a dedicated GPU and those node GPUs have less GPU ram available. The solution is to reset the node and unplug display cable so when it boots it will have all 6GB for rendering
2. There is not enough system RAM in the nodes
You should monitor RAM usage while the node is rendering, my Octane render nodes eat up to 4.2 GB system RAM so 8GB in the system is minimum for 6GB cards.
3. PCI connections in the nodes are sensitive. If you don't plug the GPUs directly to the motherboard but with PCI cables these can be the problem too, I have tried 6 different PCI cable varieties and they perform differently , some make nodes give errors from time to time or crash them and some are rock solid so these components are crucial for system stability
kavorka wrote:Every once in a while our daemons crash. Just happened last night. on 2 of our computers, the daemons say "Slave crashed or was stopped via CTRL-C. They were running all night and nobody was on the computers.
One computer had no problems.
All computers are on win 7 x64. The computer that made it has a single 780 ti. One of the comps that didnt had 2 580s and the other has a 690 and a 780ti.
This has happened before as well.
Is it possible to add time stamps to the daemons so we can see when they crash? I dont knoe if it happened at the same time or not.