Intel's. You have a ballpark figure when you buy an AMD, but with Intel you just know the 7xx is faster than 5xx. That said, as others have alluded to, AMD's system is not totally reliable either, most noted by the XP 3000+ vs 64 3000+ vs Sempron 3000+ differences. What we really need is a Third Party naming scheme that rates all CPUs based upon a rigid selection of tests. I'd prefer a MHZ based numbering system(like AMD's), but it has to be consistant with no overlapping CPUs with the same Model# with differing Performance.
Though some like to compare Sempron with Celeron(a fair comparison to be sure) and explain away AMD's Model# scheme for the Sempron, it only serves to undermine AMD's Model# scheme in the end. A 3000+ should always be better than a 2800+ and less than a 3200+ of any iteration. Just because Intel was deceptive with MHZ and Celeron, doesn't mean AMD should as well.
I wish Tom, Anand, and other Hardware sites would get together and devise such a Processor Rating scheme. If they stuck to using the scheme eventually AMD and Intel would have to acknowledge it and perhaps even use it for themselves. It is obvious that AMD and Intel are not going to volunteer to subject themselves to such a scheme, it'll have to be thrust upon them by popular demand and/or by Consumers learning to ignore their schemes.