Mark Otto "derping around with measuring CSS performance". Nothing looks particularly concerning to me, to the point where I would change how I do things. The most surprising numbers were in
background-color, but Paul Irish debunks that a bit.
There must be something in the water as Ben Frain has dug in a little bit on CSS performance as well. I like Ben's soundbite:
With CSS, architecture is outside the braces; performance is inside