<<
<< That's what I was thinking, but given a high number of transistors, couldn't you achieve similar performance (at a lower clock frequency) and still have scalability? >>
Not entirely, though the P4 was designed for more than JUST scalability, as it is geared more torwards high bandwith mulitmedia, which is why the marketing is based on better digital music and video. But In the long run, its cheaper to create an architecture that will last through many clock speed increases, because just inceasing the clock and shrinking the die is the cheapest way to improve performance, redesigning the architecture is by far the biggest investment. Notice the P3 architecture, which started with the Pentium Pro somewhere around 100mhz, went all the way up to 1ghz P3 (with minor architecural changes and die shrinks) That architecture couldnt handle higher speeds. Intel is hoping the P4 architecture will last just as long. AMD fans put too much emphasis on IPC (instructions per clock) IMO, notice the Athlon is getting mods to its design much more often, eventually they are going to have to follow the same route of the P4, with longer pipelines and more transistors, to achieve better OVERALL performance. >>
I've never thought long pipe lines were a fine idea because of the engineering costs of creating such an advanced BPU. IMO, the route of PowerPC CPUs is the best, where performance is increased by die shrinks and additional transistors, not by lengthening pipe lines. However, I do believe that AMD will follow Intel's route and use longer pipe lines in the future because of marketing concerns.