Stegger is right: the Gamescience D5 (see link in Stegger's post and pic below) is fair. There are at least four considerations that make a particular die roll fair: area of the faces of the die, the shape of the faces of the die, the angle of the faces of the die and the way the die is rolled.
We'll assume that any die we discuss has uniform density (which the Gamescience D5 does). Theoretically, you could have a fair die that isn't uniformly dense if you compensate in some way, but for the sake of simplicity let's not go down that road.
The faces of a die need not be all the same area, shape or angle for the die to be fair. However, an equilibrium has to be reached between these factors for each face to have an equal chance of being rolled (at least as random as a conventional D4 or D6). In the case of the Gamescience D5, two of the faces are triangles and three are rectangles. The rectangular faces meet each other at 60 degree angles and meet the triangular faces at 90 degrees. The area of the triangular and rectangular faces are not the same. The areas of the respective faces were determined so that they compensated for the different shape and angle of the faces.
How?
A physics professor in Canada developed a machine that casts and reads dice. He tested various proportions of D5 until he got one that was at least as random as other commonly available polyhedral dice. The Gamescience D5 is based on the professor's findings. It is fair.
With regards to the roll, that's spurious. Any die, including a casino D6, can be rolled unfairly. As long as the Gamescience D5 is rolled in the same way as an honest gamer would roll a conventional D4 or D6, it will give a fair result.
We've had this debate before at the
Dice Maniacs' Club aka The Random Fandom.