Consistent estimator of theta for a normal with mean twice the sd

Let $X_1,…,X_n$ be a random sample from a $N(2\theta,\theta^2)$ population, $\theta>0$. A consistent estimator for $\theta$ is