I don't believe StahlFaustus was suggesting that at all, but he'll have to answer for himself.
It's a common mistake that people equate PC and console architecture and try to extrapolate conclusions from that but they're really totally different. The discussion above of the weirdness in the PS3's architecture should give you sufficient hint of that. Consoles also have the advantage of providing more direct low-level access to hardware which is something you can do when 100% of your target audience is on the same platform, so things cut both ways - while console hardware is slower on paper, that lower-level access can make up for it, particularly in cases where API overhead is a bottleneck on PCs. Where API overhead is not a bottleneck (or the bottleneck is sufficiently lower) PC pulls ahead.
Read this: http://forum.beyond3d.com/showthread.php?t=61567 (the second post is particularly instructive).
My own experience in PC space is that AMD are significantly stronger at geometry throughput whereas NVIDIA are much better when it comes to fillrate. I've seen simple screen overlays or render to texture passes pull AMD down to half performance whereas NVIDIA barely even blink an eyelid at it, while start pushing polygon counts to stupid levels and AMD come to the front, being able to take extremely high counts that bring NVIDIA to their knees.
For a more balanced test that stresses both, I can push 1,000,000 fully dynamic particles at 15fps on Kepler but same at just over 20fps on Northern Islands, so even in that case AMD can still pull ahead, even with previous gen hardware vs NVIDIA's current gen. I still prefer NVIDIA for driver quality overall, but there's nothing like some hard figures to wipe away any fanboyism.
The plural of anecdote is not data however and it may be the case that my private benchmarks are sufficiently limited to only be reflective of other conditions prevailing when I've run them.