Why do ppl always try to overclock by increasing the FSB instead of increasing the multiplier? Is it because most of the CPUs are multiplier locked ?. I’m asking this because I recently bought an Athlon XP 2600+ and an Asus A7V8X-X, which allows me to change the multiplier and the FSB without any problems. But most Sites/forums are talking only about FSB overclocking. What I thought was that its better to do multiplier overclocking since it wont overclock the PCI and AGP busses. Could some one help me to understand this ?

It's kind of an old vs new way of overclocking. With older boards the PCI/AGP frequency was a factor, because the PCI/AGP busses were sensitive to drastic bus speed changes. With new boards(your board) the PCI/AGP is locked and independent of multiplier or FSB adjustments.

We like to OC the FSB because that not only increases the processor speed, but the whole system's bus speed. Simply increasing the multiplier only increases the speed of the processor. dark.

Ok, thanks for the reply. If you can lock the PCI/AGP then as you said its better to oc the FSB since it might also increase the mem bandwidth. But as I have read in several sites, the KT400 (not the A/Pro) doesn’t allow to lock the PCI/AGP. In my board the PCI increased proportionally to FSB. I’ve seen 2 other boards Gigabyte GA-7V and MSI KT4V both based on KT400 wont allow locking the PCI bus either. So for me I guess the better option would be multiplier oc ?

With a kt400 you can go to 190fsb and still have the pci at a safe 38mhz. 39mhz-40mhz is when hard disk corruption becomes a risk.

If you increase your cpu speed via the multiplier and get a 10% benchmark score improvement, you would have gotten a 14% improvement had you increased the cpu speed via the fsb instead.

Another thing is that multiplier oc'ing only works with AMD cpu's and not with Intel because since the MMX cpus the multiplier is locked internally so noone can change it.

As example for FSB oc'ing:
I overclocked my Pentium 2 from 100 to 112 MHz and that brought about 10% in 3DMark results.
But the next step was 133 MHz and that didn't work because my two ISA card will not work at 44MHz while the rest worked!
Now I'm running a 733MHz Celeron (11x66MHz) and it works well at 1232 MHz (11x112MHz)!

As long as you can change the FSB, but if you can't get farther there you can still try another multiplier with your Athlon!

Johnoh's explanation in response 3 can be seen when running some benchmarks. He reminded me today of a slick little utility from Tom's Hardware:

http://www.tomshardware.com/cpu/20030521/images/thg_clock.exe

Now, aida32 has a plugin tab. If I click on that and disk benchmark, I can click on any of the read tabs and when I start the test the cpu clock speed increases.

If I do the math the 1.4:1 ratio johnoh speaks of holds true.

Currently run a Duron1300 at 13x102=1326Mhz
Start aida32 quick linear read benchmark
Reported clock frequency of 1337Mhz
1326-1300=26
26x1.4=36.4
1300+36.4=Damn close to 1337Mhz

Doesn't show up on all benchmarks, but it's interesting to see it happen in real time.

Skip