A great post over at Awesome Dice examines how randomly RPG dice really roll. They rolled Chessex dice and GameScience dice in two tests: one of 10,000 rolls of a d20 of each, and a follow-up confirmation test of 1,600 rolls each.
Their test showed that none of the dice rolled completely true, but that GameScience rolled more true than Chessex. However, they also stressed that the difference is incredibly small. You would have to roll over 1,000 times before you saw a difference, and even then the difference would be only 5 or 10 off from expected.
I think overall this test shows that our gaming dice all roll pretty good, regardless of the manufacturer. You can read their full write up of the test here.