fryke: I read the article, don't worry. The SPEC benchmark is right where c't is biased. The SPEC benchmark still favors processors with higher clock cycles with regards to floating point operations, and it DOES NOT TAKE THE VELOCITY ENGINE into account. That's exactly where c't messed up, and they still haven't made any correction to the article.
Using the SPEC benchmark is certainly not unbiased. The only way you can reliably take performance tests is by taking real-life situations like 3D rendering, scrolling in PDF files, Photoshop, raw processing applications like SETI@home or Blast, etc. You cannot reliably use one little thing to bash or promote a processor.