AMD and Intel are pushing as hard as they can, but the bench marks attest that there is only about a 25% difference in performance between an AMD Athlon 2800+ and the new 64-bit 939 Athlon 64s. To me, performance matters. I go to bed at night wondering if my Sony Vegas project will be rendered before I get up, or wonder how late I must stay up to start a new render while Windows Media Encoder converts the last render to wmv for CD delivery. But I can not see paying 3 times more for a processor that will reduce these times by 25%.
It occurs to me that the problem is that the compiler vendors and software developers (Main Concept, Microsoft, among others) have let us down. Why is it that the SE3 and 64-bit extensions offered by Intel and AMD are not doubling performance?
I will buy a new computer when the encoding (and/or gaming) performance of a new one is roughly twice that of my current computer. I'll bet there are many others who gage their buying cycles similarly. Therefore, this is an issue for the hardware businesses as well.
I noticed Intel sells a high-performance math library through "Component Paradise". I strongly recommend that both Intel and AMD also sell (or offer for free) a media library that leverages SE3 and 64-bit extensions into a doubling of performance.
Why is it that the SE3 and 64-bit extensions offered by Intel and AMD are not doubling performance?
Because SSE3 is just a few instructions that do vector operations a tiny bit faster (need less data moving around), and 64-bit is only useful for code that needs such a large range, or when using more than 4 GB memory.
So, they don't double performance at all. Not even the best programmer and compiler can help achieve more than a few percent speedup using them, on the average application.
To really nearly double performance, you'll have to wait for dual-core processors and software that makes optimal use of it. In two years maybe... that's Moore's law anyway.