I’m using nDisplay to render a scene across 3 nodes, each rendering on 4 displays. When running the scene on the master node alone, it performs smoothly with a stable FPS around 70~ and no visible stuttering. However, when I run the same build across 2 or more nodes, the FPS fluctuates between 50 and 70+, with random stutters dropping as low as 20 FPS before recovering. The same issue occurs with all 3 nodes running.
All nodes have identical specs—RTX 6000 Ada Generation GPUs. I’m not using any frame lock or genlock, and the render sync policy is set to “None.” I’ve also tested with the “Ethernet” option, but the random stutters and unstable frame rates persist.
many recent and older forum posts about similar issues but haven’t found a definitive fix yet. Has anyone else experienced this and found a workaround?
What is the physical networking between the machines?
I would suspect this – either it’s oversubscribed, or it’s on a weak switch, or some cables are bad, or somehow the machines are set to use wifi instead of wired or something.
Of course there could be many different problems, but networking is one of the possibilities, and it’s probably worthwhile to fully investigate that.
Are any of the cables flaky?
Is the switch up to the task?
Are the network drivers updated to the latest version?
Are the motherboard drivers updated to the latest version?
If you run a packet sniffer (or tap the flows on the switch) can you see any stalls, jitter, dropped packets, or excessive bunching/buffering?