Audioaficionado said:
We just told you why. Intel quad pumps the memory because that's how their engineers designed the Intel platform to run.
Wrong.
Intel quad-pumps the FRONT SIDE BUS. This is totally different than quad-pumping the memory. The front side bus exists between a northbridge and the CPU; the memory bus (which is double-pumped on ANY DDR-based memory platform) exists between the northbridge and the memory.
Any memory bandwidth differences between the two are a result of a) different memory controller designs, b) the speed of the memory in question, and c) the number of channels the memory communicates with the northbridge on.
The front side bus differences are strictly a function of how many times per clock cycle the CPU can transmit to the northbridge and how many cycles per second this transmits. The newest Athlons and Pentium 4s BOTH communicate with their northbridge controllers at 200 MHz; no more, no less (P4C/Athlon 3000/3200.) The main difference lies between the Pentium 4 being able to transmit 4 times per clock cycle for an effective "800MHz FSB." The Athlon is only able to do this 2 times per clock cycle, hence a "400MHz FSB."
Do not confuse the front side bus with the memory bus. They are two independent things. Hence, asynch memory and FSB timings that are possible between a CPU and its memory bus. It doesn't always work efficiently, but it still works.
Now, for the differences in performance that the original poster mentioned. Intel boards are able to scream along at very high effective front side bus speeds because, well, they're designed to! However, most of these transmit cycles are latent; the P4 is figuring out what to send to the northbridge; there's usually not much chance of saturating its own FSB for very long periods of time. Most of the cycles go to waste. The AMD CPUs are able to use more of the bus more efficiently; even if the AMD could quad-pump its FSB, most of these cycles would still be going to waste, anyhow. There's a limit as to how much each CPU can do, and it's usually well below the actual capacity of the memory and Front Side busses.
Now, if you put a 5GHz Athlon in that same place (with the effective 400MHz FSB,) it'd probably have very little trouble turning that previous wide-open bus into a massive bottleneck. The same would go with a 10GHz Pentium 4.
As to the raw clock speeds that the chipsets work at, 250FSB on a Pentium 4 board ("1000FSB," and a reasonable goal) and 230FSB on a good nForce2 board ("460FSB," also within reason) really aren't that much difference. It's raw clock speed that determines a northbridge chip's ceiling; the rare instances where a board can scale up to 300 MHz (as on a few 865/875 boards) is just "one of those chips," usually just an unusually good one out of Intel's batch. Ever notice you never read of the SiS P4 boards getting even close to that speed? Plain and simple, Intel constructs a better northbridge. Nothing against SiS, nVidia, or VIA, but Intel just fabs 'em better...