Column-oriented databases are faster because computers are built for arrays and not rows. It is a lot easier to pass through a simple column in memory or read a simple column from the disk. The pass through the memory benefits from all the caching functionality on the chip, and the pass through the disk benefits from the arm not having to bounce around. Essentially you move at the maximum speed allowed by the equipment. Row-based storage defeats all that.

Indexing is also improved with a column store. If you have a row-based storage mechanism, particularly one with variably sized entries, you're going to need a more complicated indexing scheme.