Re: System test (Was: Dave Anderson BOINC Lecture)

f/f george wrote:

I know several crunchers that overclock and do NO CHECKS what-so-ever!
When I ask them about it they are not concerned. These kind people do
not seem to understand the potential problems with overclocking. They
are just happy they got it to work faster than when they bought it! They
also refuse to try to understand!!!


Guilty.  Yes, in the past I have used the S@H program itself as a way of
checking the system, since it does stress the CPU, FPU and the I/O system
pretty good.  On occasions, system stability has wained when approaching
the limits (e.g. especially going through my liquid nitrogen and
Fluorinert� phase) and WUs have aborted early.  In my defense, in such
cases I did NOT return those WUs.  Besides, S@H do redundancy checking, so
I don't really think it matters a rat's ass.

Furthermore, I think most OCers are more aware of cooling issues than
most, and thus are less likely to return garbage than the average joe who
NEVER cleans out the dust or has even thought about taking off their
system cover!  Yes, there are one or two OCers who want to play Doom or
Quake faster and don't care if the PC catches on fire while they do it,
but those are in the minority.  Here's a re-post from a well-meaning guy,
just to illustrate that even people vehemently opposed to overclocking can
generate garbage:

<Quote>

Date: Sun, 05 Mar 2000 09:59:43 -0500 From: Bob Keeter
<bkeeter@netway.com> Newsgroups: alt.sci.seti
Subject: In the heat of the battle!  (Warning for 24/7 S@H'ers)

This probably applies to all of the S@H volunteers out there who have
taken to leaving their machines on 24/7, but it bit me in the Mac!

Running an old rusty Performa 6400/180 and since S@H came up to speed have
been banging along essentially 24/7 (except for surfing time and the kids
computer games!).  Well, in several years before S@H I had not really
bothered to service the computer.  Had to replace a disk drive and add
memory so I cleaned up stuff then, but no real periodic maintenance.

Well, Friday afternoon the computer started to act wacky.  It would not
exactly quit but started getting all sorts of subtle errors that I had
never seen before.  Ran through all of the anti-virals, checked everything
with a couple of diagnostics packages including a Norton scan and really
didnt find any problem.  Finally just gave up and took it to the local
computer shop for a checkout.

It was quite humbling to see the rats nest of dust that the technician
pulled out of the box!  Running 24/7 had just plain sucked the bottom of
the unit full of trash, effectively cutting off the cooling air. Given
that Im prone to rant and rave about everything that can produce errors in
S@H processing, its very humiliating that by not servicing my CPU I may
have also been contributing to just the type of "overheated" computational
errors I've suggested to be possible when people choose to overclock.

If you are running 24/7, and particularly if you have a box that doesnt
cool well or if (god forbid!) you choose to overclock, MAKE SURE TO CLEAN
OUT THE DUST!  When the cooling ports were cleaned out, the machine runs
just fine, no errors, no glitches, etc, etc, etc!  ;-) Restarted the WU
that was in progress and it seems to be crunching fine now, cant do
anything about the ones already in, but hopefully if they start
crosschecking they will find any that my CPU goosed up.

Guess I got lucky when the hardware didnt smoke!

Regards
bk

</Quote>