I've got a 4600+ Manchester, admittedly not the most OCable CPU on earth.

The oddity is the suddenness of the failure of OC.

At FSB=217, I can run basically stable (one crash of the OCCT torture test, but things like the long wPrime and SuperPi run fine). This is with voltage at automatic, memory timings at DDR400, and LDT at 5x. Temperatures are idle about 40c, load mid-50s.

At FSB=220, nothing can salvage it. Not dropping timings to DDR333, or even turning on "compatibility mode" (which means DDR200 or something timings), or adjusting the voltage to the highest permitted (without voltmodding the board)- 1.40. It rarely makes it to the desktop, and resets promptly upon reaching there. Upping the vMem doesn't help.

The only things I haven't really tried are chipset voltage and voltmodding the board (which is OUT. I have no other 939+PCI-E board, and I'm not gonna blow 60 bucks on a replacement after destroying one on a whim)

At FSB=225 or even 230, it POSTs, but merrily hangs as it tries to load Vista (that little screen with the sliding yellow bar)

Now, I know four sticks of DDR don't OC as well as two, so I'd expect to get some extra leeway by dropping down to DDR333 or below.

The cooling is serviceable if unspectacular (Arctic Cooling Freezer 64 Pro)

I've even tried swapping the Antec SP-500 PSU I had been using for a Delta (Intel badged) DPS-600B (server grade, four 12v rails totaling 43 amperes, no namby-pamby fan-spins-down-for-quietness) with the extra plugs cut off (24+8 doesn't fit onto my board, which is 20+4 and cramped)

Drop the HTT to 4x or even 3x... those chips like to have their HTT at 2000 or less... (FSB x HTT multi x 2 = HTT speed < = 2000) try that, throw some voltage at that chip, if there is a voltage adjustment for your HTT add some there, also throw some RAM voltage at it...

I tried dropping the HT as low as 1X (on my board it's called CPU-NB link and labeled 200/400/600/800/1000) and it didn't get me over the hump; setting memory or chipset voltage to "high" didn't either.

I did notice that the upper chipset component (technically, it's not a northbridge) sink was pretty warm, which bodes poorly; I've heard the ULI 1695 doesn't get that hot (people seem to get 250 and 300FSB out of 1695/1567 boards readily, but this is a 1695/1697 board.

*@#*%. Half the reason I went with ULI was to get a fanless chipset (most of the NF4 boards had stupid little breakable fans)... shoulda gone K8T890.