I'll definitely agree that a deviant Bapco score alone means very little except to the really hardcore. Even on a somewhat geek-oriented site like fatwallet.com, I've seen that people had very little knowledge about what Centrino was or how fast it is relative to P4 or Athlon. One particular benchmark really doesn't matter all that much, particularly to the consumers who are buying these products. They go more by brand names and marketing. That's why Ruiz's new strategy is much better than old Sanders'.
However, AMD's problem is not "designing and manufacturing" CPUs. The designs are great, the manufacturing, in general, has gone well, too. I'll even say amazingly well, given the relative (to Intel) dearth of resources. It's more a matter of quantity and pricing power. OEMs will only design around a product for which they can be certain of having adequate and steady supplies (along with good marketing to boost demand). The marketing side looks very promising--benchmarks and HPC wins are really just marketing to the geek segment, after all. Now, if AMD ever manages to get their hands on enough manufacturing capacity to supply more than half the market or so, then they'll be unstoppable.
Given the lack of Intel response to AMD64, it looks like they are going with the Itanium-desktop strategy. The ONLY way this is going to work is there is no competitive x86 chip available in sufficient quantities. And VIA and Transmeta are not competitive (and not 64bit).
So the future success of AMD depends on
1) Intel sticking to Itanium strategy
2) AMD investing in decent marketing
3) AMD securing more capacity
4) AMD getting .09 right
Number 3 is what troubles me the most, especially after the fiasco with the foundry partner. The whole game now is about scale, and AMD needs to scale up or get rolled over. To address another argument that I've seen here, whatever AMD will end up paying for IBM to make its chips will be meaningless compared to the value that the increased production and IBM backing will create. I don't know whether or not it will be enough, however.
