Re: [TCLUG:9231] Observations and a non-Linux question

[Philip C Mendelsohn <mend0070@tc.umn.edu>]

On Sun, 17 Oct 1999, Christopher Reid Palmer wrote:

> To make sure it was stable, I did a batch compress of 11 wavs to mp3s.
> That took 2:40. Would anyone agree that if the CPU runs overclocked at
> 100% for 2:40 and doesn't fail, that it's safe to run it overclocked?

I bet that you can find someone to agree with you, but I don't think
that's a sufficient test.  Your best bet is probably to get the GIMPS
prime number searching software, which has a good processor test.  It
takes several hours to burn in, but more importantly, has varied
algorithms (i.e., makes use of the ALU differently) and *knows what
results it should get.*  Your test, in particular, is less than desirable
because the mp3 decoding is designed to handle quite a bit of bad data --
so you would probably correct your errors and pass that test even if your
CPU were on it's last legs.

> Are
> there any other tests I should perform?

Something where you can do a bit for bit comparison is a start.  The other
thing you can do is play with the heat gun.  If the processor stays cool
at your clock speed, even with a heat gun on it, your clock speed probably
isn't causing trouble!

> I plan to bump up the multiplier by 0.5 every few days or so until I get
> weirdness, then I'll back off. Any bets as to how high I'll get? :)

That depends entirely on how much cooling you have in your system, and how
well it's RF shielded.  (I've seen it happen where the system doesn't fail
at the higher clock, but it isn't resonant at the right RF freqs anymore
(due to case / PS / board layout) and wrecks havoc on TV reception, etc!)

-- 
if ($income > $expenses OR $time != $money )
	set hell_frozen=true;
asif