Re: Problems with the new "smp-calibration" routines

Mogens Valentin (monz@danbbs.dk)
Sun, 22 Feb 1998 15:53:22 +0100


Christian Rost wrote:

> I have an Asus P65UP5 motherboard with a C-P55T2D pentium-CPU-card
> and two P166MMX Intel CPU's. The CPU's are overclocked, they run at 200MHz
> For the last 6 month (using 2.0.x, 2.1.5x) I have no problems, the machine
> runs very stable ! But I notice something strange: One CPU seems to be slower

<snipped>

> Running the CPU's at "normal" speed (166Mhz), everything works fine, but I
> don't want to do without the additional 34Mhz ;)

As others have already pointed out: You almost got away with it :-
But did you run a longtime HW test using a known quality testprogram? Always do,
when modifying beyond specs. - Your cpu's are doing 20% beyond...
Anyway, now that you know the MB+cpu's runs stable at specs, and _almost_ stable
when OC'd to 200Mhz, why not check to see if that MB will do 75Mhz on the bus and
scale the cpu's at 2.5 to 187.5Mhz. Overall performance will be even better that
way.
OC'ing should gererally be limited to 10-15% - *if* at all! _And_ extra cooling
should be provided; did you do any work here, or still using stock (undersized?)
cpu-coolers ;/
BTW, do your cpu's show the same bogomips when running at specs? Else they may
differ somehow. They are of same stepping, though. Others have in various
linux-smp entries pointed out that cpu's _must_ be exactly identical, when doing
smp.

--
Best regards,
Mogens Valentin.  Networking & Programming
Web Site: http://www.danbbs.dk/~monz
mailto:monz@danbbs.dk or mailto:monz@usa.net
"Lord, give me today my daily Linux login"
  --Monz

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu