Paul, end users *should* look at actual performance before making their decisions. If you want to get idealistic, then AMD should not have had to introduce a TPI model scheme in the first place, much less one that was designed to fool consumers into thinking they were the same as Pentium 4 megahertz - the very myth that AMD aimed to dispel. The Opteron naming scheme does not correlate with actual performance, but it does offer relative performance to end users who are not capable of figuring out performance on their own.
You are wrong that people will suddenly decide to learn for themselves how to determine performance. The reality is that end users will never learn, because computer performance takes far more knowledge in micro-architecture and implementation details for the average consumer to understand. They need a simple metric, that offers them relative performance across multiple families. Megahertz was a good relative metric among members of the same family, but things have proliferated far too much, and there are too many different CPU lines for anyone to keep track.
That's why the new model numbering scheme is required. You can't boil down frequency, cache size, bus speed, and micro-architectural features in any other simple way. Guys like us will know the difference between a 525 and a 730, but Joe Consumer will look at the last two digits and correlate from there.
Market Data 
Markets 
AI
