This is not true. It is fast as a general purpose hash table, but claiming it's the fastest across all datasets and workloads is silly.