What FSB were you running?

I found the QX's to clock the best if you leave the FSB close to 333 and tweak the multiplier with a little FSB changing. 380x10 (1.65 mch) was not prime stable even for 15 minutes of the small fft test even at 1.4v, though its worth noting that higher vcore did help stability slightly as i started with 1.3625v, while 333x12 passed a full pass of small fft prime (8k-64k-about 3.25 hours) with 1.3725v (1.368v shown in cpu-z), with the vdroop pencil mod. (200 mhz higher at lower vcore, take note of that).

to compare, 400x10 was a complete no POST at 1.4v. Oddly enough with both a 1.02G P5W (using a X6800 dual) and the 1.04G with qx9650, I couldn't reach 400 fsb, but I didn't try 1.75v mch. The 1.02G wouldnt even do 370x10 on the X6800 unless i pushed the vcore to 1.65v (!). But 340x11 posted at 1.55v.

Someone mentioned a long long time ago that they 'felt' that the board, when set to 370x10, was actually initially posting (before bios init finished) at the default multiplier (11), which meant the cpu was momentarily at 370x11 (4.05 ghz) then setting it to at the end of the BIOS init, which explained why 370x10 posted at 1.65v but not at 1.55v, while 340x11 posted at 1.55v. Possibly the CPU has to power up and accept signals at the FSB and its stock multiplier, before it can be told to lower it by the BIOS signal (?)

I never bothered to test something like this, but it would have been easy enough to set it to 370x11 to see if it still needed 1.65v for POST. I no longer have the x6800 or the 1.02 board though....