st62k fsb/multiplier error

[contrex]

I have recently had an ST62K built for me by a firm in Plymouth, Devon
UK. (But that's another story.) It uses a 3.0 GHz Prescott and I find
that from time to time when starting up, the bios incorrectly sets the
multiplier & FSB to 15 x 166 MHz instead of 15 x 200 MHz, giving me a
2.49 GHz cpu clock. The RAM is always correctly recognised as 400 MHz
dual channel.

The error occurs when I explicitly set the multiplier & fsb to 15 x 200
and also if I set it to "auto".

Sometimes 2 reboots are needed before the correct settings are seen.

I have flashed the bios to what I believe is the latest revision
(suffix 'X'). (It had 'P' when I got it.)

I have seen this error mentioned before; has anybody any idea if it is
fixable, or will I just have to live with it?

Also, what are good cpu temps? I am getting 54-56 on idle, and about 62
on load. If I set the fan to 'ultra-low' in the bios, it hardly ever
ramps up)

[Jeff_Lincoln]

I can't help you, but I have the same problem, I have a 3.0 Northwood cpu, but it comes up as a 2.59 like yours, maybe 3 boots out of 10.

[Moeric]

I solved the problem by switching the ICE FAN connector from FAN1 to FAN3 on the mobo, near the PCI slot. I know it is strange, but it solves the FSB problem.

For more info search 'st62k' on http://forums.sudhian.com/.

[Moeric]

I forgot to tell about your temps : your temps are a bit too high! Try lowering the CPU core voltage down to 1.200V in the bios, should be perfectly stable if your CPU is a Prescott.

[contrex]

I tried the fan header switch from FAN1 to FAN3, this cured the problem for a while, but the next day I was back to wrong speed boots, so I switched it back. This gave me another run of correct boots, but as always, leaving the Shuttle off overnight was sufficient time to bring back the problem next morning.

I discovered that if I overclocked in the BIOS to a crazy extent so that Windows XP froze during loading and I had to press the reset button, the result would be a correct multiplier on the reboot. (But I had to clear the CMOS afterwards). This would last all day over any number of reboots.

But perhaps I should have just pressed the reset button anyway. Yesterday i read a description of this problem on Sudhian (I think) where the guy said he watched the bios boot screen and if he saw the wrong cpu speed, he pressed the reset button at that moment.

I just tried this, and it got me a correct cpu speed. I was ready and waiting and I pressed reset at the exact moment when the cpu speed appeared on screen. I noticed that there was only a very tiny pause indeed before the correct speed appeared, not the second ot two needed for a complete restart.

So following a cold boot, a wrong multiplier is corrected by a warm reset?

I forgot to say, I undervolted down to 1.175 v, all seems stable, idle temp is down to 38 C and load to 52 approx.