I have a project where i use the replicator comp to clone a component, that inhibits a lot of “object” nodes to calculate distances to other SOPs. Middle mouse click on one of the replicated items shows a children CPU Cook time of approx. 1.3ms (AMD Ryzen 3950x, 4,25GHz Stable Clock, 16 Cores/32 Threads). Now if i turn up the number of replicators, the framerate goes down very quickly, since all of the replicas are processed on a single core (Ressource Monitor shows maximum usage on one thread and the rest is bored). I think it would my a big performance leap if the replicator comp would be a device that actively uses multi-threading, so each replica can run on its own core. For now my project is on pause, because of this situation and i hope there’s a technical way to realize a feature like this.
Attached you see 4 replicas (the colored objects) and sphere SOPs whose distances to the replicas get measured (which colors the spheres accordingly). This runs at about 40 fps (but CPU usage in taskmanager is at 4%, because of the single thread that’s being used).
Ideally i could create 64 instances of these objects and have the calculation processed on different threads.
Thank you for your support!