It'll depend on the allocator. Generally speaking, the better an allocator is at minimizing fragmentation, the slower it's going to be with new allocations.

That's why bump allocators are wicked fast, but make the trade-off the article mentions.