Jon Buller wrote:
> From: Matthias Pfaller <leo@dachau.marco.de>
>=20
> > My machine is running @30mhz >2 years now, so this seems to be pretty
> > "safe". You should put
> >=20
> > 	options 	CPU30MHZ
> >=20
> > into your kernel config file or a couple of timings will be off by 20%.=
..
>=20
> On that thought, has anyone tried a 66.6666MHz crystal?  Or does
> the general populace think that would generate too much heat for
> a fan, or push everything over the edge...

I never tried this. I wouldn't think your CPU will burn out, but I'm
sure you get closer to the margin of the overal system design. You
can try, but I'd advise doing a backup before. It may very well be
the SCSI controller that starts working erratic...

> Also, what timings are affected?  I have noticed that my pc532
> loses about 30 seconds a day.  Would that be due to losing clock
> interrupts, or a crystal that isn't quite 50MHz, or something else?

There is a routine in locore.s that is used to burn microseconds. This
loop should be adjustet. The other place where CPU30MHZ is used is in
dev/scn.c. The scn chip is already quite close to the edge at 30mhz :-)
and needs some time to recover from write accesses to the channel
control register. At 25mhz this is covered by waitstates, but not
at 30mhz. Who knows, what will fail at 33mhz...

> It seems to be only about 10-15 seconds a day if the machine is
> idle, and sometimes up to a 60 secs if I have something like a
> "cd /usr/src; make depend && make" going.

That's the ICU clock. But from what you describe, we probably
are to long at splhigh (or di()) somewhere. Maybe in the SCSI
pseudo dma code.

> My quick and dirty solution to this was to run "nistime" in my
> net-connect script that runs every day to get my mail...

You could also use your RTC. This should run pretty exact.

Matthias
--=20
Matthias Pfaller                            Software Entwicklung
marco Systemanalyse und Entwicklung GmbH    Tel   +49 8131 5161-41
Hans-B=F6ckler-Str. 2, D 85221 Dachau         Fax   +49 8131 5161-66
http://www.marco.de/                        Email leo@dachau.marco.de