First, an important preliminary-- there are millions of places in the human genome where any two given people could possible differ, either by a single base change, the addition of an entire chunk of DNA, the inversion of a chunk of DNA, or whatever. Keep that in mind: millions and millions of places (for a database of many of the single base changes, see the HapMap). Now, the intuitive argument: after humans arose in Africa, they dispered themselves throughout the world. By both chance and in response to selection due to their new environments, populations in different parts of the world ended up with different frequencies of those millions of DNA variants. Simple enough. Now, below the fold, I will present the evidence that 1. the patterns of genetic variation form clusters on a world-wide scale, 2. genetic clusters coincide with what is commonly called "race", and 3. genetic variation between clusters is relevant phenotypically.
Jason Rosenhouse hasposted on race recently as well. You can find some of my own opinions on the topic here. Ultimately, I think asking questions about race/population substructure is very interesting because I find human evolutionary genetics very interesting. 2 years ago Armand Leroi could plausibly say we didn't know how skin color was genetically controlled. Today he wouldn't be able to say that.