If the data is less (less than like million points). Then just do a scatter plot and visually count the number of clusters. This is also a good place to decide wether to use KMeans or DBSCAN. If the data is in circle-ish clusters then k-means should work fine, it not DBSCAN would be better.
I am not sure what to do when data is very large. I guess something specific about where the data came from and what we need can be used.