Good generational GC doesn't need to deallocate objects one by one, instead of that it can deallocate objects in bulk (if there is no finalizers). It can alloca

Good generational GC doesn't need to deallocate objects one by one, instead of that it can deallocate objects in bulk (if there is no finalizers). It can allocate memory faster because it doesn't need to use memory allocator to create every object, it just increment pointer that points to the beginning of the free space. It's true that modern GC has greater throughput than simple allocation/deallocation schemes. Problem is predictability (as been mentioned).