On Tuesday 21 October 2003 10:06 pm, Terence J. Golightly wrote:
> Tom/List,
>
> I bought the memory according to a newer post to you sent (I
> can't find it :( ). Got the Two 512 MB sticks in the mail and
> stuck them both in (using proper static aviodance procedure) and
> tested.
Barefoot on a tile floor is my method ;) If your not familiar
with bios ram timing options, set the ram to auto or bios defaults.
Cas 3, precharge 3, banking disabled are the safest (but slowest)
settings.
> Memtest gave me errors in the five figure area. I
> shutdown and tested each stick individually. One stick is fine
> and the other gave two errors in test 5. Why would I get 5 figure
> errors when both sticks are installed and only two when the "bad"
> stick in installed. If you want the particulars of the output,
> I'll post them it you like.
OK, I've been havin hardware problems myself (due to
overclocking too far. IOW's user error, ME). I narrowed it down to
either my ram or the cpu has become weak. Figuring it would be
better to try ram first, Crucial delivered a new stick of pc3200
this afternoon to replace my existing Kingston (both 512 MB
sticks). I figured if ram wasn't the problem, then at least I'd
have twice as much ram ;)
But, I took some precautions. First I d/l'd a 1 GB ram capable
kernel (2.4.22-18mdk-i686-up-4GB) an installed it. You'll need to
do the same, or add mem=860M to your lilo append line for the
kernel you're usin. The 'regular' kernel can't address 1024MB of
ram. Next I underclocked my XP 3000+ cpu to less than a 2200+ to
try an mitigate it as a factor when I installed the new ram and
tested with memtest86. First with just the new stick.
It took 3 tries to get the new ram properly seated and the system
to even boot. I also had to swap slots in order to find a combo
that worked. By now I've determined that it's my cpu that's
fragile, not the ram. So I'm usin both sticks now. Memtest86 checks
both sticks together with no errors (an an underclocked cpu).
BUT, a word about memtest86. There is no such thing as a software
ram tester. While memtest86 is runnin, your whole system is being
used, particularly cpu/cache/ram and motherboard. In memtest86's
configuration tho (press the <c> key while it's runnin), you can
turn cpu cache off. Press c, then 1, then 2, then 0. It could be
the cache areas of your cpu generating the errors.
So, install both sticks of ram and the kernel I did (or any
'enterprise' kernel) or add mem=860M to lilo. Don't forget to
run 'lilo' to make the change effective. Then get
ftp://mersenne.org/gimps/mprime2212.tar.gz and unpack it.
cd to the directory you unpacked to and run './mprime -m' Then
choose 17, the torture test. You should be able to let it run for
at least an hour or more. If it doesn't stop on hardware errors,
you done a better cpu/cache/ram test than letting memtest86 run for
a week.
> oh btw I didn't notice a selection that would allow me to save
> output to disk. Is there one?
Read the docs, look at all the options when you press <c>.
I believe the error output would be shown in the bottom half of
memtest86's screen, and there is an option to also write to file.
But since I've never had any errors, I can't verify that.
--
Tom Brinkman Corpus Christi, Texas
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
