This is a follow-up question regarding these posts:
[Content removed]
[Content removed]
We are using World Partition in 5.6, and we set s.ForceGCAfterLevelStreamedOut to 0 (disabled) to avoid hitches when cells are streamed out.
According to the official documentation, this means Incremental GC may be triggered when a WP cell unloads.
However, wp.runtime.UpdateStreaming.EnableAsyncUpdate, an experimental variable introduced in 5.6, seems to cause a crash like the posts above.
We are using this variable for a performance boost, and we understand that it may cause issues since it is still experimental.
It appears that CL#49409329, mentioned in the following post, may resolve this issue. However, we are unsure if we should stop using one (or both) of these variables, or if it is safe to apply this quick fix.
We are not aware of issues related to using the 2 CVars you mentioned together. The CL you referenced prevents the streaming async update task from running while the garbage collector is running. This could only happen when using the Async streaming update and is unrelated to the first CVar you mentioned. Having s.ForceGCAfterLevelStreamedOut disabled will result in keeping some data loaded for a longer period requiring more RAM. The benefit is that this can avoid reloading assets that could have been GC’ed earlier.
We are aware that are running this configuration with success (2 CVars & the fix) so they can be considered safe. Please share more details (stack, lof, dump…) if you require further assistance.