There are a few things that I thought I'd mention, since we are having a civil discussion here.
First, processor clockspeed comparisons between two different core designs are really useless. I could explain in gory detail all that is involved in proving this statement, but I'd probably lose 99% of you in the process. (AKDejaVu and I could have some fun debating though.
) In general the
shorter the pipeline is in the CPU, the faster it will be at the same clockspeed as a competing processor with a longer pipeline. Without factoring in ways of "cheating" to shorten the pipeline (branch prediction and hardware prefetch), you need to go through a number of tasks to complete each instruction. Pipelining a CPU breaks the three major tasks in the ALU down into a number of smaller jobs, this allows higher clockspeeds, as less needs to be done per clock. (In case you are interested those 3 main jobs are "fetch", "decode", and "execute". "Fetch" grabs the next instruction from memory, "decode" determines what the instruction is and what memory locations to read and write from/to, and "execute" does just what it says.) Now in the case of the Athlon there are 11 stages in its pipeline. The P4 has 20 stages. What this tells you is that it is much easier to hit higher clockspeeds on the P4 than on the Athlon. It also tells you that at the SAME clockspeed the Athlon has a worst case performance level almost 2x that of the P4. Now those "tricks" I mentioned before like hardware prefetch and branch prediction can allow you to essentially eliminate stages in the pipeline for better performance, but both the Athlon XP and P4 both incorporate these features. (The P4 has somewhat better hardware prefetching abilities, because it has extra bandwidth on it's FSB to allow it to do so, than the Athlon XP. On the other hand, the Athlon XP has a superior branch prediction unit than the P4 does.)
It's also worth touching on the fact that the Athlon XP has a VASTLY superior x87 floating point unit than the P4. The P4 lives and dies by it's SSE2 instruction set to achieve parity with the Athlon XP in floating point intensive tasks. (Most mathematical software and direct x games are very floating point intensive, and in fact you'll find that in benchmarks the Athlon tends to dominate in these types of tasks.)
Now to be fair I should mention that the P4's ALU (arithmetic logic unit) is at least as good as that in the Athlon, but with one noteworthy difference: It operates at 2x core frequency. This will result in the P4 gaining an edge in some applications as the clock speeds continue to ramp up.
Both companies like to resort to clever marketing names when describing their CPU architectures. Intel calls their P4 architecture "netburst", which IMO is deceptive to the general (ie, uneducated in CPU design) public. "Netburst" is TOTALLY UNRELATED to the Internet. AMD calls their CPU architecture "Quantispeed", which doesn't mean anything, but sounds kind of neat.
Because of the general level of ignorance of CPU benchmarks, other than raw clockspeed comparisons, AMD has switched to a PR type rating scheme. If you actually do compare performance between a system identically configured with a P4 1.7 Ghz and an Athlon XP 1700+ (1466 Mhz clock) you will find that AMD has actually underrated their CPUs for the most part. The 1700+ is certainly a match, and generally superior to, the 2 Ghz P4 with Rambus ram. If you put SDRAM in a P4 system you might as well cut it's performance by 30+ percent. When looking at P4 systems it is very important to avoid SDRAM powered (i845) systems like the plague. (Unless you feel good in knowing that your old P3 1 Ghz would actually outperform your brand new P4 1.6 Ghz in most benchmarks
)
I'd write more if others are interested.
