computing cost of np.random.uniform vs np.random.beta

I have an algorithm that uses np.random.uniform I would like to switch to np.random.beta and/or np.random.gamma to improve the efficiency of an approximation. Since the algorithm is already quite time consuming (~8hours) and memory intensive (5 GB per thread) I wanted to check if there was any information on how much this would cost me before I tried a run.

Since it is called inside a loop, should I assume it is not going to change the memory requirements?

What is the difference in run time?