Divider 1:1.20 needs more CPU voltage than 1:1 ???

[zfooz]

I don't understand exactly why I can run my CPU at FSB 460 with 1:1 divider
and it's about 1.26v to do that, but at 1:1.20 divider for the ram it's 1.31v to run it..
How could there possibly be that much difference between simply switching dividers? It must require more current on the cpu to do 1:1.20 for some reason.
Has anyone else noticed this? Maybe doing 1:1 allows for higher clocking per voltage?
I tried increasing the MCH a LOT and the DDR V and VTT with 1:1.20 without increasing cpu volts to no avail, but increasing cpu 3 notches works at 2.40 better..

E8400, Abit IX38 board

This board seems to need a whole lot of MCH voltage over 400FSB for some reason..
1.55V right now. VTT doesn't seem to matter what I set it to for this CPU.

Thanks!

 

[lopri]

It's because the CPU/MCH/Memory and FSB are all tightly woven togeter. And Intel chipsets are horribly inefficient handling dividers. (Not that it's needed. 1:1/1:2 will cover the most performance anyway)

 

[zfooz]

So it's normal to need more CPU voltage even if I crank the NB and VTT for 1:1.20 ?
Man that's weird. I don't remember having to do that on my AMD, or my old p4.
I remember having 1200 FSB on my p4 northwood.. That was the day :>
This E8400 is the most finicky cpu I've dealt with.. It can run 3.6ghz forever at 1.15v and then it needs almost 1.32v to run at 3900mhz.. Shesh
And the board needs 1.5v NB where default is 1.25 just for 100mhz fsb more.
Is it just me or is everyone experiencing this with the new boards? The last machine I seriously overclocked was about 4-5 years ago.

 

[lopri]

From my experience, yes. It is normal that changing one affects other(s). By how much will depend on chipsets and boards/BIOS so it's common to see even the same CPU needs different voltages per different boards. Divide and conquer doesn't work with current Intel overclocking.