PROBLEM: Linux system clock is running 3x too fast

My Athlon 64 laptop (HP Pavilion zv5000z) dual-boots Linux and Windows XP. The Windows system clock is running accurately but the Linux system clock is running 3 times too fast.

The problem occurs in all of the Linux distributions and releases that I've tried, including Suse 9.1, Fedora Core 1 & 2, kernel versions 2.4.x, 2.6.x (up to 2.6.8-1.521), 32-bit & 64-bit releases.

The Linux system clock problem is also intermittent. Approximately 1 of 10 (cold) boots could yield an accurately running Linux system clock... I've tried kernel boot options "clock=tsc", "clock=pit", "clock=pmtmr" and have seen about the same 1/10 (failed/passed) ratio for each of them. I've also tried kernel boot options "acpi=on" & "acpi=off" but they don't seem to have any affect on the problem.

The differences in dmesg outputs between "good" and "bad" boots are as followed.

good: time.c: Detected 797.952 MHz processor.
bad: time.c: Detected 265.995 MHz processor.

good: Calibrating delay loop... 1576.96 BogoMIPS
bad: Calibrating delay loop... 516.09 BogoMIPS

good: Detected 12.468 MHz APIC timer.
bad: Detected 4.156 MHz APIC timer.

good: intel8x0_measure_ac97_clock: measured 49383 usecs

	intel8x0: clocking to 47408
bad:	intel8x0_measure_ac97_clock: measured 48347 usecs
	intel8x0: measured clock 16547 rejected
	intel8x0: clocking to 48000

Full dmesg logs (goods and bads) are available upon request.

This is a feature. The kernel detects users with the name fastclock and accordingly speeds up the clock for them.

The stock Fedora Core 2 x86_64 kernel is compiled with powernow-k8. It does not fix the system clock problem.

# dmesg | grep powernow
powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version
1.00.09b)
powernow-k8:    0 : fid 0xa (1800 MHz), vid 0x2 (1500 mV)
powernow-k8:    1 : fid 0x8 (1600 MHz), vid 0x6 (1400 mV)
powernow-k8:    2 : fid 0x0 (800 MHz), vid 0x12 (1100 mV)
powernow-k8: cpu_init done, current fid 0x0, vid 0x12