The biggest blocker in all of this ecosystem’s history, going back to 2019 is memory. Each era of content was determined by how much memory we had and what optimization tools we were given to better optimize it. This still applies, as UEFN has unlocked a multitude of new tools and allows for the scope to be increased, and memory recently also received a patch that made actor and device cost less.
That is all great, but then 2 updates later the bugs started again, massive increases in memory when world streaming was used, and now UEFN doesn’t have the optimization tools that we had in Creative to know what causes the peaks or try to debug. Also, you can’t see memory in UEFN right now, because the bar only shows you the memory cluster you are currently standing on, not the highest one. So if you want to plan your features and scope, you can’t really. Devices take overall memory because of their fancy meshes, overriding world streaming because they include functionality. This messes up with the scope of games, because any attempt to make a complex game will quickly hit the memory limit due to the combination of actor # and devices needed to appeal to a complex world. Verse partly solves that, but it’s not competent to handle multiple device behaviors on its own yet. We are also left with a memory calculation tool that takes 20 mins to complete on complex maps, whereas in Creative it was instant, which is a huge waste of dev time.
We need all of Creative’s optimization tools, and devices to not count over memory (ditch their meshes if you have to, we need the functionality to work not their looks) and honestly we need to stop having memory be the blocker. Even in Creative, it killed one of my largest projects and all it took was one update that makes memory calculate differently, could have been a bug too who knows. But it’s another huge risk for our development to have thousands of assets tell you if you can publish or not, because even if one calculation breaks out of those thousands, you are locked from continue working or publishing. It should be a decision of a benchmarking tool, not an arbitrary number.