Fivethirtyeight.com did a study on the non-black vote per county, assuming (for the study) equal turnout and 96% of blacks in every county voting for Obama. There were a few counties in the Deep South where Obama was projected to have gotten less than 0% of the non-black vote (and several less than 5%). Obviously, below zero is impossible, but those counties would have majority-black precincts for McCain.
Almost certainly not with residential segregation being what it is. Turnout matters.
Oh, they exist. Because racially close precincts exist, and turnout matters in them too.
Of course, in urban areas, 50-55% Black translates into overwhelmingly Democratic. And even in rural areas, it frequently translates as solidly (say 60-70%) Democratic, though elsewhere racial and political shares match. Where it gets most interesting is probably the outer edge of suburbia.
From a cursory glance at East Central Georgia, where the DRA has presidential results:
Precinct 805, Richmond County. 50.2% Black, 55.8% McCain (Whites and Blacks are tied in VAP at 46.5%)
White Plains precinct, Greene County. 54.4% Black, 50.2% McCain. (Black majority even in VAP).
Two more McCain precincts with less than 50% White in Richmond and Burke Counties.
Oh wow - Mullis precinct, Dodge County. 49.9% White, 71.2% McCain. Rural place; Black and Democratic vote shares mostly match very well in the surrounding precincts - and they're a lot whiter.
There's a matching oddity in the whitest precinct of heavily Black Hancock County, Devereux 1A: 51.9% Black, 97.2% Obama. Whoever said Black Belt Whites all vote Republican? Or maybe both are simply results of fraud...