Ex:

X4 (4 Cores) 3000mhz (Per CPU Core) 1800mhz (CPU-NB) 1800 (Hyper Transport) 200Mhz (HTT) 1066Mhz (Ram)

Sandra Mem Bandwidth : 7000Mb
SuperPI :20 seconds

Or you can gain another 500mb or shave 1-2 seconds without the raise of CPU Clock Speed by;

-Raise CPU/NB up to 2400 using stock volts (1.1v)
-Raise Mem Speed to lets say 1200Mhz
-Raise HTT around 225

All three combo increases performance which can be seen in benchmarking tools...

Yap...because it is linked all together via multplier...

Lets say 200HTT 1800 CPU/NB 1800 Hyper Transport;

when u raise 10HTT the CPU/NB and Hyper Transport also raise accordingly but is multiplied... 1HTT will be times 10 if u use 10 multiplier...if u use 12 then (HTT value will) multiply with 12....

Some reads the CPU/NB Frequency by multiplier and some by the speed...but both is actually the same thing...

When u put CPU/NB at 12, lets say u run 200HTT ur CPU/NB frequency will be operating at 2400mhz...if u use 10x multi, it will result to 2000mhz frequency speed...