From: kutcha@eos.acm.rpi.edu (Phillip Rzewski) Subject: Crashing machine >16M: Update Date: 14 Apr 1993 03:14:47 GMT
Hello again everyone. For those that have been following the thread, I am the
same person who posted the bit about "Crashing machine with more than 16 megs
of memory" patch applied and all. I have gotten several replies and tried many
things and thought that those of you following the thread would appreciate an
update. Actually, I also think I should add a couple things I realized I left
out in my first post.
Oh, first the things I left out. I have been testing this under 0.99p8 (though
I also had made it crash under p6 and p7). Also, the memory itself is all the
same speed (80ns) so it's not possible that it could be a speed problem.
(Though I would be skeptical of that as a problem in the first place. Someone
explained to me that if you mix speeds it all just runs at the speed of the
highest in the mix, which makes sense from an architectural standpoint.)
As for the updates. Someone suggested that maybe it was a bad SIMM. I didn't
want to admit this as a possibility at first (it was easier to blame Linux :) )
so I cracked open the machine and ran a battery of tests. It all came up
negative. In the end I just SWITCHED the 4 SIMMS that have been working as my
16 megs for the past month with the other 4, and have been running fine with
them for days now. So, even though it would have been an easier solution to
the problem, it looks like my memory is fine.
Also someone pointed out to me that machines with a small cache and lots of
memory tend to run really slow. This explains a lot to me. I have just a 386/25
with a 64k cache, and I noticed that when I upgraded from 8 megs to 16 megs
that it spent less time swapping (of course) but it seemed to be running much
slower when doing normal work with the memory. So obviously this is linked to
the size of my cache. In other words, even if I COULD get all 32 megs to work,
with a 64k cache it's likely that it would be slower than I was at 8 megs with
swapping!
Nevertheless, these answers don't explain why it broke when I applied the patch
and tried to use all the memory. As another helpful soul pointed out, the patch
is amazingly simple and is only in place to watch out for things like the cache
problem. I guess the patch can't be blamed then. What is left? The motherboard?
Any other PD 386 owners try to go to 32 megs?