The reason for the difference is that it used to be that the memory bus clock speed was determined by the cpu base clock (hypertransport clock) speed, which is 200mhz. But with socket AM2 (and socket F) cpu's, the memory bus is based on the internal cpu clock speed, using a divider of 1/2 of the cpu multiplier.
So, in the example of that article, the FX-62 and 5600+ have a clock speed of 2.8ghz, with a multiplier of 14. In this case, half of the multiplier is 7, so the memory bus speed is 2800/7=400mhz*2(b/c it is ddr)=800mhz, and DDR2-800 ram.
However, to get the 3.0ghz for the 6000+, the cpu multiplier is 15, and half of the multiplier would be 7.5. However, the amd64 memory bus doesn't work with non whole number dividers, so it rounds up to 8. This gives a memory bus speed of 3000/8=375mhz*2=750mhz (not sure how they got 746, I think it's a mistake). This gives a slightly lower available memory bandwidth.
Thus, you see that your DDR2-800 memory is not in fact being quite fully utilized by the 6000+. The higher clock speed of the 6000+ generally more than makes up for this, and the 6000+ does outperform the 5600+ and FX-62 in nearly all applications.
So I hope this long-winded explanation helps a little, and I would still get the 6000+. Just think, now you can overclock the bus speed a little without even stressing the ram!
