When the airflow over a surface is laminar, the boundary layer separates earlier creating more drag. When the airflow over a surface is turbulent, the boundary layer separates later so there's less drag. The bottom line is turbulent airflow over a surface creates less drag.
I chewed on this for a while, and can not completely agree with it. Laminar flow is a good thing, but when the surface loses laminar, then there is separation of the flow. That separation is draggy, but you do not improve the total drag by having turbulent airflow over the entire surface. Rather what is desired is to keep the flow laminar for as long as possible, then trip the flow to induce the turbulence to keep the boundary layer as thin as possible.
Not much clearer, but I feel better.
As to the original question, since golf balls spin and are free to rotate during flight, you dimple the entire surface for the reasons already given. However on a vehicle, keep the flow laminear until the flow separates, then take corrective action. Tuft the vehicle and see what the air is doing.
Recently, Damler Benz developed a semi truck that gets 12 mpg @ 50 mph, about 50% better than semi trucks on the roads today. Yet a train still moves the same payload 50% further with the same fuel. A recent article on drag reduction of coal cars showed that having a completely empty car has more drag than one with two bulkheads at the 1/3 and 2/3 point inside of the car.