Hi,
We are using Unreal 5.7.0 and want to use ndisplay to launch a packaged application across our cluster in our VR immersive room.
We have one master node and 6 slave nodes. Each slave launches 2 applications (1 viewport per user eye).
We managed to have the application launch but the performances are very low. We have a blanc map with nothing in it and are barely hitting 20fps per eye.
We have been using unity applications so far that can reach 55fps per eye.
Looking at the log files, it seems every node is having difficulties talking to itself, which may explain the poor performances?
Example for the master node (131.254.154.50), we have lots of logs like this
[2026.01.21-14.39.34:013][ 89]LogUdpMessaging: Display: Unicast socket bound to ‘131.254.154.50:56504’.
[2026.01.21-14.39.34:013][ 89]LogUdpMessaging: Display: Added local interface ‘131.254.154.50’ to multicast group ‘230.0.0.1:6666’
[2026.01.21-14.39.34:014][ 89]LogUdpMessaging: Display: Using asynchronous task graph for message deserialization.
[2026.01.21-14.39.34:014][ 89]LogUdpMessaging: Error: Sender FUdpMessageProcessor.Sender: SendTo failed (destination: 127.0.0.1:6666) (SE_NO_ERROR)
[2026.01.21-14.39.35:024][112]LogUdpMessaging: Warning: UDP messaging encountered an error. Auto repair routine started for reinitialization
[2026.01.21-14.39.36:320][130]LogUdpMessaging: Display: Unicast socket bound to ‘131.254.154.50:60800’.
[2026.01.21-14.39.36:320][130]LogUdpMessaging: Display: Added local interface ‘131.254.154.50’ to multicast group ‘230.0.0.1:6666’
[2026.01.21-14.39.36:321][130]LogUdpMessaging: Display: Using asynchronous task graph for message deserialization.
[2026.01.21-14.39.36:321][130]LogUdpMessaging: Error: Sender FUdpMessageProcessor.Sender: SendTo failed (destination: 127.0.0.1:6666) (SE_NO_ERROR)
[2026.01.21-14.39.36:816][142]LogUdpMessaging: Warning: FUdpMessageProcessor::AcknowledgeReceipt failed to send 131.254.154.50:51736.
[2026.01.21-14.39.36:816][142]LogUdpMessaging: Warning: FUdpMessageProcessor::AcknowledgeReceipt failed to send 131.254.154.50:51736.
[2026.01.21-14.39.36:816][142]LogUdpMessaging: Warning: FUdpMessageProcessor::AcknowledgeReceipt failed to send 131.254.154.50:51736.
[2026.01.21-14.39.36:818][142]LogUdpMessaging: Error: UpdateSegmenters, Socket error detected when communicating with 131.254.154.50:51736, Banning communication to that endpoint.
[2026.01.21-14.39.36:820][142]LogUdpMessaging: Warning: UDP messaging encountered an error. Auto repair routine started for reinitialization
[2026.01.21-14.39.37:925][162]LogUdpMessaging: Display: Unicast socket bound to ‘131.254.154.50:65041’.
[2026.01.21-14.39.37:925][162]LogUdpMessaging: Display: Added local interface ‘131.254.154.50’ to multicast group ‘230.0.0.1:6666’
[2026.01.21-14.39.37:926][162]LogUdpMessaging: Display: Using asynchronous task graph for message deserialization.
[2026.01.21-14.39.37:926][162]LogUdpMessaging: Error: Sender FUdpMessageProcessor.Sender: SendTo failed (destination: 127.0.0.1:6666) (SE_NO_ERROR)
We are using network sync in the ndsiplay configuration because our master node is not equipped with a nvidia sync card. Setting it to None gives barely better performances (25fps max) but all nodes still seem to display the same fps and the same logs appear.
On another note, we could not find a way to exclude that node from the sync (as this is a control node, we do not need it to be perfeclty synchronized with the rest of the cluster). Any hint about how to do that would be appreciated too !