Is READ UNCOMMITTED equivalent to non-ACID (BASE) transaction management?

Relational databases are often associated to the ACID properties, which "protect" data but may hinder performance, even on a single node. Isolation levels (the I in ACID) shield transactions from various types of interference from other concurrent transaction. The greater the isolation level (think, SERIALIZABLE) the greater the performance penalty, because you are intentionally limiting the number of concurrent transactions on the same data set. BASE databases essentially follow an optimistic approach to conflicts. Conflicts can appear, but they will eventually be solved and all replicas will agree on the same value (eventual consistency). Is the READ UNCOMMITTED isolation level equivalent, in terms of performance, to using a NoSQL/BASE database? It seems that, in both cases, you avoid locks and gain in terms of performance while losing some consistency.