Hello everyone, I have already written about this issue, but I want to rewrite everything in a more logical way!
When migrating the project from Unreal Engine 5.3.2 to 5.6.1, we encountered a critical bug in nDisplay on the following configuration:
1 cluster, 1 host, 1 local node
4 viewports arranged in a row
Each viewport has a resolution of 1920x1080, with a total width of 7680x1080
Rendering on a single GPU, without Multi-GPU/Multi-Process
Projection occurs through Mesh, each viewport is its own Mesh with a UV map
Problem:
Frozen frames randomly “jump” from one viewport to another, with the “jumping” frame being displayed behind the active viewport, creating distortions in the projection on the simulator.
The error is unpredictable.
Logs at the maximum LogDisplayCluster VeryVerbose level do not show any errors or warnings.
Resetting the configuration and recreating nDisplay Config does not help.
Using Render Sync Policy and other standard synchronization methods does not affect the problem.
Only one GPU and one local node are used, Multi-GPU and Multi-Process are disabled.
Measures taken to resolve the issue:
Recreating nDisplay configurations from scratch.
Updating video drivers and Windows SDK.
Request:
Is it possible to get some clarification or official guidance on the cause and nature of this issue?
Are there plans to fix bugs related to viewport desynchronization in nDisplay when using Mesh Projection on a single GPU in upcoming patches?
Are there any internal methods of the nDisplay plugin or UE API that would allow this desynchronization to be safely eliminated or avoided?
Will improving logging for these cases be considered to facilitate diagnostics?
I am ready to provide detailed logs (I didn’t find anything in them, though), configurations, and project files to reproduce the problem.
If you want, I can help supplement the request with your data.