In message <20060919014047.GO29625@bcd.geek.com.au>, Daniel Carosone writes:
>My guess: you're running current, on a laptop with some form of
>speedstep, and for a time you had a kernel that used the TSC
>timecounter without paying attention to the fact that the cpu clock
>rate changes the tsc clock rate on your cpu.  There was a relatively
>short time-window in current when this would happen on some machines.

>You might even find that 35:60 is pretty close to the ratio of the cpu
>idle:max frequencies.

That seems quite possible.  I am running current, on a laptop with speedstep,
and I seem to be using TSC.  However:

1.  There were other ranges; I might lose a second per minute for an hour or
two.  It was fine shortly after a reboot.  It didn't seem to correlate at all
to system load.

Also:
>> I replaced the laptop's CMOS battery, and this no longer happens.

>I suspect this is a coincidence, and a kernel update was what actually
>fixed the problem (by using the acpi or some other stable timecounter
>instead).

Nope!  I left the kernel alone when I swapped the battery.  Wanted no
ambiguity.

-s