fpg, Re: Hammer runs at faster clockspeeds than Athlon, a simpler design but 130nm nevertheless. That's enough proof for me that SOI may provide some non-negligible benefit.
As per AMD's assertion, SOI provides approximately *one* speed grade of performance over bulk silicon. So it shouldn't be much of a surprise to see that Athlon 64 can clock at 2.4GHz while Athlon XP only hits 2.2GHz. Man, what an epiphany.
Re: Do you know for a fact that Prescott or Dothan wouldn't improve with SOI?
I'd expect a similar improvement in Dothan or Prescott if Intel implemented SOI, but there are tradeoffs involved. And even if you could justify these tradeoffs using today's data, then I would still argue that 200MHz more in performance to Dothan and 300-400MHz to Prescott would not make the competitive world of difference that people here are trying to claim.
Think about it: Prescott at 3.8GHz still wouldn't outperform Athlon 64 right now, so what difference does it make? If Intel is 5% behind or 15% behind, it only means something to the benchmarking cheerleader geeks who make a big deal over 10 frames per second on their favorite video game. Long term, this will be damaging, and people will eventually learn that AMD has a faster part, but for >50% of Intel's sales which are now coming from developing countries, in which the people have no clue about these subjects, it makes no difference.
In the end, it comes down to manufacturing volume, and SOI doesn't help yields, it hurts them. That is to say, you'll gain a bin of performance, but you'll lose some of your lower bins to defects, and that will affect the bottom line more than anything.
