Its not really late, its when they could. I've said in other threads, essentially their lack of production capacity hurt them in the short term but was always completely unavoidable. When you get to good as they did the demands for their chips left them with massive problems filling orders and as such couldn't shut down to switch to 65nm fast, which delayed their 45nm switch etc, etc. Essentially if they had one more fab this whole time they'd have brought in 65/45nm same time as Intel and been more competitive throughout, they'd also not have to skimp on core size as much as they did on P1 they would have had P2 straight off the bat a year ago and probably have a revised better P2 to go on 45nm 6 months ago. Bulldozer is there nearest real architecture upgrade and very little is known about it. WIll be interesting but its been in the works for a while, it might well be the chip after that which is finally free from limited budgets/timeframes and not designed to be as small and efficient as possible and might be designed more for all out speed and damn the cost. Thats what massive financial backing allows.
The incredibly quick switch to 45nm rather than the years it took to gradually move to 65nm is basically down to short term profits not being important due to that backing, they just said, fab 36 or fab 38(not sure which) offline, heres 10million go install the 45nm as quick as possible, they also paid TCSM i believe a shedload of cash to get their hands on some of their 45nm fab kit. This would never have happened without the Dubai's group money, backing and the financial stability they now have to say screw short term profits. But chips are LONG cycles in the making, its not we made the Ath 64, now lets make the X2, and the X4, X4 was probably in the works before the Ath 64 launched. Likewise Bulldozer has been in the works for a couple years and it will really be that next chip 3 years from now maybe that is just design for pure power. That will be interesting to say the least, also because their 6billion dollar new fab might be up and running by then.
But either way, the AM3 version even with ddr2, once we start getting similar uncore speeds to the i7 series you'll start to see just how important that is to the benchmarking/real world speed. Its a very real limitation on these chips at the second with early yields, though some certainly go higher with good air cooling which no reviews took into account. The Q series doesn't have it, the Phenom's could literally have 10-20% untapped performance waiting to be let out, at that stage they should clearly beat the Q series, and be closer but not close enough I fear to the i7. Still vastly better value.