I had a similar problem with a PowerCenter 132, and when I
upgraded it with a 300Mhz PPC CPU, the characteristics of the
time drift changed and I had to re-calculate.

I put the "fix" into the boot command in the open firmware so
that I don't have to re-enter it when I reboot:

" /PowerPC,604" select-dev A7D8C0 encode-int " timebase-frequency" property boot

(obvious, huh? ;-> )

-dgl-


At 2:51 PM -0500 8/3/01, Adam Lebsack wrote:
>I just started using netbsd-macppc once again and my biggest problem 
>was the time drift on my PowerCenter Pro 180 (overclocked to 210). 
>The RTC was running at about 94% of the actual speed of time, so  I 
>was losing several minutes/hour and ntpd could not keep up.
>
>Anyway, I figured out an accurate fix if anyone is interested.  use 
>the date command to set the clock in sync with an accurate watch or 
>clock.  Be aware of the time when you set it.
>
>To get a good reading, take a sample of at least a day.  One day 
>later, take the seconds passed on your computer and divide it by the 
>seconds passed on your clock.  This gave me a value of about 0.94 
>(that's rounded off, use as many decimal places as you can).
>
>Next, boot into OF:
>
>>  dev /PowerPC,604
>>  .properties
>
>look at the timebase-frequency value (it's in hex, convert it to 
>decimal).  I had a value of about 15800000 Hz. For your new value, 
>use this simple algebra:
>
>	new_timebase_frequency = timebase_frequency * drift_percentage
>
>My new value was something like 14.9 MHz.  encode it into hex, 
>0xE4891A, and then enter this in OF:
>
>>  dev /PowerPC,604
>>  E4891A encode-int " timebase-frequency" property
>>  .properties
>
>The last command is to verify.  Now, a reboot will clobber this new 
>value, so just boot directly:
>
>>  boot
>
>Now, if you got a nice and accurate reading, your clock should be 
>pretty sane.  You may spend a lot of time getting it "just right", 
>but don't bother.  If it's close enough, just run ntpd and that will 
>keep the clock as accurate as you can get it.
>
>I posted this because I felt there weren't any decent explanations of 
>this on the web.  Now, it may work with other processors, but I have 
>not tested it.
>
>Cheers,
>Adam
>
>--