Petz, Re: " I remember it being disclosed back in Spring of 1999 that the IPC of the 1.4 GHz Willamette would be about the same as the 1.0 GHz PIII. Indeed, I exchanged email with Robert Coldwell, one of the designers of Willamette in which I tried to pin him down on the performance of Willamette. I wasn't very successful, but he left me with the distinct impression that P4 was not going to be a real barn burner."
Willamette might have been a real barn burner, if it had been released on the original schedule. In actuality, the Willamette design was delayed for a year, and the Coppermine core was put on the roadmap to take its place. In the original roadmap, instead of a Coppermine 600MHz core, which was to follow the 550MHz Katmai, there was going to be a 1.2GHz Willamette. Lower IPC or not, that would have been a significant upgrade!
Unfortunately, Willamette shipped a year late, and only gained 300MHz in the mean time. Given the historical rate for processor improvements (2x every 18 months, or so), the 25% gained from the small frequency boost wasn't enough to give it a sizable lead against the previous generation 1GHz Pentium III.
Willamette wasn't a story of computer architecture gone wrong; rather, it was a story of architects biting off more than the designers could chew, and inviting delays as a result. Willamette could have been an amazing processor, one year earlier, but it was late. The same happened to Merced, but to a larger degree. Increasing that core frequency 33% hardly made up for the fact that it was 2 years late to market.
I'd say that to a lesser extent, AMD is having a schedule problem with their .13u processors. They've arrived 6 months too late, and with no bonus to frequency in the mean time. That sort of makes them lame competition to the Intel equivalent, but if they had launched on time (December, 2001), they would have done extremely well.
You should make an effort in the future not to confuse the abilities of a processor design team with the difficulties of getting a complex product like a microprocessor out to market. It's not easy - ever - but sometimes unexpected things can make a complicated process even that much more difficult.
wbmw |
