Subject: Re: lessons learned...
To: None <port-hp300@NetBSD.ORG>
From: Steve Peurifoy <firstname.lastname@example.org>
Date: 03/31/1998 17:50:11
> Recently I learned a couple of hp300 lessons. I thought those of you who
> didn't already know this would be interested:
> a) overclocking an 040 is not necessarily koshure... I had a 425T that I had
> overclocked to a 433T. It worked fine for the longest time until we tried
> to install a newer version of perl and the install failed. Then we noticed
> that 'ps -axwl' was printing out numbers incorrectly. as in "1.B" and "1.;"
> and so forth. We bumped the oscillator back down to a 50Mhz oscillator and
> everything started working again... So it seems just the FPU was unhappy
> but other than that, everything else had worked for over a year...
It may be worth noting that if one attempts to overclock an 040 using a
run-of-the-mill oscillator, one is likely pushing the envelope a little
harder than usual. The PCLK input has a very tight duty cycle spec
(50% +- 2.5% for the 25 MHz part, 50% +- 3.3% for the 33) and I'm given to
believe that they Really Mean It. I think most of the 380/4xx machines
had oscillators that were either binned NEL parts or a special compensated
job that Kyocera designed specifically for driving 040s. These sorts of
things tend to be difficult to find and relatively expensive. This might
be the reason for the 4x oscillator in the "S" - if PCLK is derived by
dividing a raw clock by two, the designer has more control over things
like duty cycle.
Personally, I've never had enough faith in my luck to run a processor
past its spec anyway.