How can you even say that? Im sure those extra pipelines, new registers, and the 64-bit backwards compatible with 32-bit abilities was also in the K7 just hiding right?
Two additional packing stages were added to the Integer pipeline, and the GPR's (again, only integer) were extended to 64-bit with addition of 8 additional GPR's and 8 additional SSE/SSE2 registers. The FP pipeline remained largely untouched from the K7 and the Integer execution core remained untouched. The x86 decoders remain the same, the issue rate, scheduler, re-order buffer, cache configuration. Compare this with the transition from the P6 to P7 core. Nothing, absolutely nothing, remained the same. Not the caching structure, not the decoders, not the execution core, not the schedulers and most definitely not the re-order buffers. Again, I was making a relative statement. The K8 is nowhere near as substantial a change from the K7 as the P7 is to the P6 core nor the NV2x to the NV3x core.
The Pentium-M (Banias) had more changes compared to the P6 architecture (which it was supposedly based around) than the K8 does the K7 and that's not stopping people from calling it "just a P3".
As for the comparison between Prescott and the FX. That'd be true of the Athlon64 was released now or 6 months ago and Intel was desperately trying to get Prescott out the door to regain the performance crown (and failed to do so). Nothing about that situation even remotely resembles what's really going on. The Athlon64 will not be released too much prior (perhaps even later) than Prescott and we have absolutely no idea how good Prescott will fair. Not to mention the current P4 line is doing just fine against the competition.
"We are Microsoft, resistance is futile." - Bill Gates, 2015.