Only scale preculling data by 50% over 32MB and fix it overwriting data (plan to put in a new allocation system later, for now it's maybe 100MB wasted ram)