Date: Fri, 05 Aug 2005 18:52:15 -0500
From: Joy Freeman <[EMAIL PROTECTED]>
Sigh, now I'm starting to worry about what to do if none of these sticks are
good. So far I've had failures on four configurations of three of the
unknown chips using Jeff's sandwich method. And one of those configurations
consisted of the known good chip on one end with two of the unknown chips
that had previously passed 1000+ consecutive solo tests, run before I
switched to the sandwich method.
It is possible that one of your DIMM sockets is bad. I'll address
that further down.
RAM testing is tough because there are so many variables which can
play a part. Are your DIMMs good? Are the sockets good? Are the
crashes caused by some other issue such as an overheating processor?
Etc.
Your situation is very frustrating and I sympathize. I generated
this test method with the help of the folks on the SuperMac list
because I was in a similar situation some years ago with what seemed
like inconsistent results. I would think that I had DIMMs certified
okay and then test some more in other combination and get failures on
DIMMs that had tested okay by themselves. Once I started using the
Sandwich method things became consistent. In that old case I had
three bad DIMMs in a batch of ten though I didn't know it until I
formalized my testing method.
But the Sandwich is not the end-all and be-all of test
configurations. The trick is to know what a given test can tell you,
and then use your tests logically to eliminate (or at least make
improbable) possibilities.
For example these two cases are very important:
**Testing 1 DIMM at a time: A failure is really a failure. That
DIMM (or its socket) is bad. A success is not a success. You did
not test the whole DIMM. It could still have a defect.
**Testing 3 *Unknown* DIMMs in a sandwich: A success is only a
success for the middle DIMM. But it is a success. A failure does
not tell you where the defect lies. It could be in any of the three
DIMMs. Once you have two known good DIMMs this gets better.
Then again, it occurred to me that maybe my "known good" original 64MB chip
isn't good, either. Is it possible that I've been running on bad memory? Is
there a middle ground between memory so corrupt the machine won't boot and
somewhat flaky memory that causes occasional problems? What kind of problems
*does* bad memory cause?
Yes, your 64 MB DIMM may be bad. Yes to your two middle questions
too. Bad memory causes values in memory to be incorrect. This can
have effects from nothing, to a misdisplayed pixel on your screen, to
a command in a program becoming a completely different command,
resulting in unpredictable and possibly fatal (to the runtime)
results. Add in the possibility of very bad memory vs. subtely bad
memory which I mentioned in an earlier message and it gets messy.
Anyway, unless at least one of these tests good, I need to order at least
one 256 MB DIMM by tomorrow, hopefully to arrive Monday or Tuesday.
First, make sure that the pins on your DIMMs are clean. Examine them
and confirm that they're reasonably shiny. You may wish to clean
them with some rubbing (isopropyl) alcohol to remove any grease.
Some people will rub them with a pencil eraser, but if you do this,
make sure the eraser detritus lands far from your computer.
With five DIMMs there are 60 different ways you could test three
DIMMs. This is a pain, but I would formalize my testing order so
that I know I'm making progress.
The first thing I would do is test each DIMM all by itself. If you
test a DIMM all by itself, and it passes you do *not* know that the
DIMM is good. However, if you test a DIMM all by itself and it
fails, you *do* know that it is bad. If you can eliminate even one
DIMM as bad, then the number of possible test configurations drops
from 60 to 24.
So test each DIMM by itself (5 tests) and possibly eliminate one or
more from the testing as a bad DIMM. Also, if a DIMM fails, test it
again in a different slot. So I would test DIMM 1 in slot 1, DIMM 2
in slot 2, etc and retest any failed DIMM in a different slot. If a
DIMM is bad in one slot, and good in another slot, then the first
slot may be bad. Try some other DIMMs in that slot to see if they're
always bad in that slot. This will help you find any bad slots you
may have. If you have one bad slot out of four, you can still use
the sandwich method, leaving the bad slot empty.
If you number the DIMMs 1, 2, 3, 4 and 5 (cut up mailing labels?)
then there are ten different combinations of three DIMMs: (123),
(234), (124), (134), (125), (135), (145), (235), (245), (345). Using
the combination 123 as an example you can test them in the following
orders:
213
312
123
321
132
231
You only need one test to pass in order to find a good DIMM.
Unfortunately, a failure doesn't tell you which DIMM failed.
Now to speed up testing, I'd play the odds a little bit. For
example, rather than work through all the combinations above, in
order, I might argue that the middle DIMM is the most likely to be
the culprit in a given failure.
So, for example, if 213 failed, I would set 1 aside temporarily and
try 324. If passed, this test will certify 2, which is different
from 1. It puts 3 in the front slot instead of the back slot, so if
3 was the problem, it's a different position now and may not cause an
error. It removes 1 which was most likely to be the problem, but
adds 4 which is an unknown. I don't know that trying this way would
really increase the odds, but I think it would. So here are all the
orders you can test in . View in Chicago so they'll line up:
213 214 314 215 315 415
312 412 413 512 513 514
123 324 124 125 325 425
321 423 421 521 523 524
132 234 134 135 235 435
231 432 431 531 532 534
243 142 143 145 245 345
342 241 341 541 542 543
152 153 253 154 254 354
251 351 352 451 452 453
Each column is a combination of three DIMMs. There are six orders
for each combination of three. Each pair of rows lists all the ways
to test a given DIMM in the middle. For example, rows 1 and 2 show
all the ways to test with DIMM 1 in the middle position. Of course,
for columns 234, 235, 245 and 345, those rows are empty, as 1 is not
included in those combinations.
I might try the following order of tests until I get a success:
213, 324, 432, 541, 153
If all those fail, I might just try DIMM 1 in the middle until it
either succeeds or fails six of the first two rows of twelve
possibilities. E.g.: 312, 214, 413, 512, 315, 514 and if all those
failed, I'd tentatively assume that DIMM 1 was bad and proceed with
testing DIMM 2. Since we're assuming DIMM 1 is bad, this leaves 324,
423, 325, 523, 425, 524. But for speed, lets eliminate the
differently ordered same combinations and just test 324, 523, 425.
Then if all those fail assume 2 is bad and test the 3-in-the-middle
rows excepting those configurations that include 1 or 2.
If none of those pass, you'll have to bring either 1, 2 or 3 back
into the mix to focus on 4 and 5.
As you run each configuration, check it off the list above.
Hopefully, some DIMM will test good long before you've tried every
combination. If three out five of your DIMMs have a defect, then you
may never find a combination that tests good.
Once a test passes, you know that the middle DIMM is good. Be sure
that DIMM is labeled. Now check off all the tests on the list above
with that DIMM in the middle. Then, only run further tests with that
DIMM in the front or back.
For example, if a test passes with DIMM 1 in the middle, then DIMM 1
is known good. There is no further need to run the first two rows of
tests above with 1 in the middle. Furthermore, limit your tests to
combinations that include DIMM 1, so you can skip columns 234, 235,
and 345.
Once you find a second good DIMM, you should only need to run three more tests.
Like David, I am wondering if there's a brand you recommend? Or that you
suggest we stay away from? (These are Smart Technologies DIMMs with Infineon
chips. Are they perhaps on that list?) Any advice on finding decent memory
that doesn't cost an arm and a leg? Anyone?
If Velocity Upgrades was still in business I'd recommend against them
with caveats (they were good about returns, I just had to make too
many returns). I have not bought that much memory recently, so
others will have much more experience in this area.
I bought one 256 MB PC100 DIMM from OWC and two from Frys with no
problems in any of those cases. OWC had a nice sale going but it was
over when I needed the next two, and Frys had the better price then.
One other thing. If it looks like all your RAM has a problem, think
about other things that might be wrong with your system. It's
possible that you have a bunch of defective RAM, but that's pretty
rare, so you may have some other problem that looks like bad RAM,
e.g. dirty or defective DIMM socket, CPU not seated properly, CPU
overheating (heat sink attached and heat sink grease fresh?), etc.
Jeff Walther
--
G-List is sponsored by <http://lowendmac.com/> and...
Small Dog Electronics http://www.smalldog.com | Refurbished Drives |
-- We have Apple Refurbished Monitors in stock! | & CDRWs on Sale! |
Support Low End Mac <http://lowendmac.com/lists/support.html>
G-List list info: <http://lowendmac.com/lists/g-list.shtml>
--> AOL users, remove "mailto:";
Send list messages to: <mailto:[email protected]>
To unsubscribe, email: <mailto:[EMAIL PROTECTED]>
For digest mode, email: <mailto:[EMAIL PROTECTED]>
Subscription questions: <mailto:[EMAIL PROTECTED]>
Archive: <http://www.mail-archive.com/g-list%40mail.maclaunch.com/>
iPod Accessories for Less
at 1-800-iPOD.COM
Fast Delivery, Low Price, Good Deal
www.1800ipod.com
