Collaborative filtering dataset preference?

For collaborative filtering, is it better to have less users with more reviews or more users with less reviews or does it not matter?

For instance:
is 200 users with total 20,000 reviews
2000 users with total 20,000 reviews

The more review per nb_user*nb_item the better (as it means that you have more information on your users).

If you have a lot of user but they only produced one review each then you know very little about them.