Need help with difficult hd problem: was Gremlins...

[alfrede_neumann]

(was "Drive Gremlins continuing saga" I posted a reply to
"(e-mail address removed)" this below twice but it never showed.)

Further information since last post: I benchmarked the hard drives on my
system. Drive 0, 20 gb appears to be working normally, around
20MB/second both read and write. Drive one, the slave for which this
machine seems to have destroyed two drives in this slot, shows read at
45MB/sec and write at only 2.5 MB/sec (8GB). Both are WD on primary ide
channel. WD diagnostic program shows both set to UDMA-66 and OS also set
to this, win98SE. Both appear to be on an 80 wire cable (from it's
looks).

The 2.5MB/sec is suspect from the little I know about this subject. Re
the two drives that seem to have had head crashes, killing one and
putting a bad spot in the second, does anyone have any further tips in
how to diagnose this problem? Next I will replace the cable with a
better round cable, twisted pair. Maybe this post is overkill since I
seem to have generally brought two areas of suspicion from the good
replies I've gotten (much thanks), the power connectors and the power in
general. Anything else that could show these symptoms?

Dell reports that "some" of their models had a problem accepting udma-66
settings from the bios, but they do not say which models and their
"solution" is to reset the bios to auto and reboot. Maybe this is a crap MB
or bios design?

----
Subject: Re: Drive Gremlins continuing saga
From: (e-mail address removed) replied:

Reseated all the power connections, emoried the power
connector to make sure the metal contacts were not oxidized.

Don't do that; it can remove the plating and lead to oxidation problems
in the future. Just use alcohol, preferrably 90% or higher purity, and
plug and unplug the connector several times or wipe the contacts with a
popsicle stick or toothpick. Allow alcohol to evaporate completely
before reappling power. But more often the problem is loose contacts,
which can be bent slightly (emphasis) to retighten. Bad contacts,
crimps, and broken wires can be detected by measuring voltages at the
circuit board and comparing them to the voltages on unused connectors.

***
I actually didn't emory it, I used a micro screwdriver to lightly scrap
the female connectors. I don't know how you can bend them since the are
recessed into the plug-how do you do it? Also how can you get a vom in
there while the comp. is running to test the voltages, seems very risky
to me?

Next step is replacing the cables as soon as I can. There are so many
brands out there, it's hard to figure which is the best value; I was
unable to find any test data on the 40 vs 80 wires, or anything
related to IDE cable testing.

With anything faster than ATA/33, an 80-wire cable is a must for
reliability, but round cables vary in quality, and many are made
incorrectly, with straight wires instead of twisted pairs (TPO). The
safest choice is a $1 generic flat 80-wire cable.

***
I was suprised to find it difficult to find any 18 inch flat cable. All
the ones I saw at egghead were 24 inches and up. Suprisingly, this
interface must be more robust than I thought, since there are so many
mfgs. who are selling cables in the 24 inch to 36 inch range. The one
test I found showed better transfer rates between round twisted pair and
flat cables, claimning alot of interference comes from outside the
cable, but this test looked like it could have been sponsored by the
round cable mfg (no suprise).
***

 

[Rod Speed]

(was "Drive Gremlins continuing saga" I posted a reply to
"(e-mail address removed)" this below twice but it never showed.)

Further information since last post: I benchmarked the hard drives
on my system. Drive 0, 20 gb appears to be working normally,
around 20MB/second both read and write. Drive one, the slave
for which this machine seems to have destroyed two drives in this
slot, shows read at 45MB/sec and write at only 2.5 MB/sec (8GB).

What are you measuring that with ?
It may be just an artifact of a poor test.

Both are WD on primary ide channel. WD diagnostic program
shows both set to UDMA-66 and OS also set to this, win98SE.
Both appear to be on an 80 wire cable (from it's looks).

It doesnt take much to count the wires to work
out whether its got roughly 40 or roughly 80.

The 2.5MB/sec is suspect from the little I know about this
subject. Re the two drives that seem to have had head
crashes, killing one and putting a bad spot in the second,

Thats unlikely, real head crashes shouldnt produce that sort of read rate,
particularly if its over the entire drive, not just one patch of the platter.

does anyone have any further tips in how to diagnose this problem?

I'd check carefully that the write rate is real. Try HDTach.

Next I will replace the cable with a better round cable, twisted pair.

Dont do that, those flout the standard and the last thing
you need right now is another variable. A new standard
80 wire ribbon cable costs peanuts. Make sure its no
longer than the standard 18" long.

Maybe this post is overkill since I seem to have generally
brought two areas of suspicion from the good replies I've
gotten (much thanks), the power connectors and the power
in general. Anything else that could show these symptoms?

Its more likely that the write rate is an artifact of a bad test.

Dell reports that "some" of their models had a problem accepting
udma-66 settings from the bios, but they do not say which models
and their "solution" is to reset the bios to auto and reboot. Maybe
this is a crap MB or bios design?

Its possible. The obvious way to test that is to see how
the 8G drive performs in a completely different system.

 

[alfrede_neumann]

What are you measuring that with ?
It may be just an artifact of a poor test.

I tested the drive by timing it on a data transfer or 500MB. The
benchmarks came from fresh diagnostics. Both confirmed each other.

It doesnt take much to count the wires to work
out whether its got roughly 40 or roughly 80.

It's 80, went by pictures of both types, didn't actually count them.

Thats unlikely, real head crashes shouldnt produce that sort of read
rate, particularly if its over the entire drive, not just one patch of
the platter.

Well, all I know is that neither DLG zeroing program nor scandisk could
transverse that spot on the drive. Had to part. to 7GB to avoid it
trying to read that spot. The first drive it ruined now makes a loud
clicking sound in the bad area, like a skipping record needle.

I'd check carefully that the write rate is real. Try HDTach.

I tried to install one version of HDtach and got: winNT required. So I
must not have the right ver. or support files.

Dont do that, those flout the standard and the last thing
you need right now is another variable. A new standard
80 wire ribbon cable costs peanuts. Make sure its no
longer than the standard 18" long.

Well, I'll have to search further. Egghead didn't have one that I could
find. Either 12" or 24" and up for the flat cables.

Its more likely that the write rate is an artifact of a bad test.

Don't think so. Manual timing confirmed that rate.

Its possible. The obvious way to test that is to see how
the 8G drive performs in a completely different system.

Don't have one now. Probably going to buy a seagate anyways, but am
afraid to put a new drive into that box before I know what's going on.
Thanks for answering.

 

[Guest]

Reseated all the power connections, emoried the power
connector to make sure the metal contacts were not oxidized.

I actually didn't emory it, I used a micro screwdriver to
lightly scrap the female connectors.

My mistake for assuming you had used emory paper when you wrote
"emoried the power connector." But don't scrape the connectors
either. At most, rub them with a pencil eraser that contains no grit,
but it's safer and about as effective to wet them with alcohol and
simply plug and unplug them several times.

I don't know how you can bend them since the are recessed
into the plug-how do you do it?

The drive connectors on the power supply can be tightened by prying the
small screwdriver between the metal socket tubes and the holes in which
they sit.

Also how can you get a vom in there while the comp. is
running to test the voltages, seems very risky to me?

The circuit board of the drive has to be accessible, so you may have to
remove the drive to touch the meter probes to the circuit board (just
behind its power connector). You're right it can be risky, especially
if a meter probe slips, and to lessen that risk cover the circuit board
with a piece of cardboard with holes cut out for the places you want to
measure. Use plain brown cardboard, not anything glossy, which will
usually generate more static electricity.

Next step is replacing the cables as soon as I can. There are so many
brands out there, it's hard to figure which is the best value; I was
unable to find any test data on the 40 vs 80 wires, or anything
related to IDE cable testing.

With anything faster than ATA/33, an 80-wire cable is a must for
reliability, but round cables vary in quality, and many are made
incorrectly, with straight wires instead of twisted pairs (TPO). The
safest choice is a $1 generic flat 80-wire cable.

***
I was suprised to find it difficult to find any 18 inch flat cable. All
the ones I saw at egghead were 24 inches and up. Suprisingly, this
interface must be more robust than I thought, since there are so many
mfgs. who are selling cables in the 24 inch to 36 inch range. The one
test I found showed better transfer rates between round twisted pair
and
flat cables, claimning alot of interference comes from outside the
cable, but this test looked like it could have been sponsored by the
round cable mfg (no suprise).