DI is fun :) I ended up @4645 with 1.75v. Raising up to 1.82 did not help so I have to get better cooling. :) Unfortunatly it did that bench with cpu-z 1.37 the same hours that v1.38 was released so I can not confirm it. :( It booted and was benchable at atleast 4550. anything higher than that was starting to be unstable. I have a few screenshots and photos posted from that session though:
http://www.techsweden.org/forum/viewtopic.php?t=171459
About overvoltage/undervoltage. I can not tell atm, because I have modded my board. But with BIOS 0005 it seems more correct than with the other two BIOS I have tried (witch ever that was... hrm). If I recall right one was overvolting and the other undervolting. Can it be so har to fix? :D
With BIOS 0005, when raising the vcore from 1.635 to 1.65 the voltage droops like a stone. I think it falls with about 0.125v... I have not found any more strange things about the voltage yet. I have heard that over 1.75v often kills the P5WDG2 WS Pro but I have been up to 1.82v so far without problems.
I does not seem to drop under load. I have done the vdrop mod but I have not touched the potentiometer. Even before the vdrop mod i think it was pretty stable.
The 'biggest' issue for me at the moment is that I can not get clockgen to work with my motherboard. It only works 'halfway' (great english!). AIBooster have some life of it owns so sometimes it drops voltages even when untouched.
About 3DMark06: Again you have higher clock than I have at my current screenshots. But even if we had the same cpu speed I think that it is the fact that you have the B3 and/or the fact that you can get higher FSB on the P5B that makes yours perform so much better. Max FSB for me is 318MHz "stable".
You get 550 cpu points more than me with a 180MHz higher clock.