many baby-size 80486 motherboards fail with ISC 2.2.1 & 1542B & HPDD

leo at aai.com leo at aai.com
Thu Jun 13 08:27:40 AEST 1991


I have spent a lot of time trying to track down a problem with a
486 system.  We tried changing power supplies, motherboards, disk
controllers, I/O boards, memory, processors, etc.  I experimented
with wait states, bus speed, and everything else the bios let me
twiddle, to no avail.  The problem was tracked down to the
motherboard.  When I plugged all of my boards, memory, and
processor into a full-size motherboard using an Opti chipset, it
works fine.  Out of about 10 baby-size motherboards from
different manufacturers, with different bios, only ONE so far
doesn't have the problem!  The motherboards have used a variety
of chipsets including Opti, Eteq, and TI.  The only one that
seems to work properly uses an Intel chipset and Award bios.

The symptom is that the system spontaneously reboots, often
during heavy disk activity.  The disk activity light stays lit
for about 20 seconds, although no disk activity can be heard.
The next time any key is pressed, a reboot happens.  Sometimes it
just reboots without a key being pressed.  No panic, no errors,
just the beep of the power-on self-test.  It's as if you hit the
reset switch.

I found that I could exercise the problem by running two
disk-intensive processes simultaneously.  What I did was to use a
couple of little scripts that continuously loop, making and
cleaning both g++ and libg++.  I run each of these in a separate
virtual terminal and then do time stamping in a third.  Sometimes
it will fail after only a few minutes, and sometimes it will run
an hour or so.

What does this problem sound like?  My guess is that it has to do
with the High Performance Disk Driver using the 1542B in bus
mastering mode.  Does ISC do this?  If so, have all these board
manufacturers made the same mistakes?  Other than this
problem(!), the systems run fine.  If you have a baby-size
motherboard that you use with ISC 2.2.1 and a 1542B, you might
want to try a similar test.  Am I the only one to see this?
-- 
Leo	leo at aai.com   leo%aai at uunet.uu.net   ...uunet!aai!leo



More information about the Comp.unix.sysv386 mailing list