In this letter, we consider the posterior Cramér-Rao lower bounds (PCRLB) problem for extended target tracking (ETT) from a stack of measurement data that are modeled as random variables in the random finite sets (RFS) framework. We convert the scalars in the traditional PCRLB into vectors based on RFS to derive a theoretical lower bound. In this way, the proposed method can be applied to the multi-target tracking problem and accommodates scenarios with targets of varying. Moreover, we consider solving the data association problem from four parts caused by the conjugate update of the Poisson multi-Bernoulli mixture (PMBM) filter. Simulation results are presented to verify the effectiveness of the derived PCRLB.