Jesus, the article is less than 3 years old. That was the first thing I found when googling "failure rates graphics cards". If you're gonna say "X is better than Y", then maybe it should include the last couple years at least, and not a single series from two particular manufacturers.
If Mercedes S had lower failure rates than BMW X5 in Spain in 2011, does that mean that statistically Mercedes cars have lower failure rates than BMW, and citing a result from 2009 is like comparing cars from the 80's?