From: Nancy Haitz <[EMAIL PROTECTED]>
Date: Thu, 4 Aug 2005 21:32:50 -0400
Joy,
There are probably a number of ways to test RAM but I like a program
called RAMometer. It is part of Newer Tech's Gauge Pro group of
programs, and is a free download from Version Tracker here <http://
www.versiontracker.com/dyn/moreinfo/macos/15583&vid=61005>
After you download and install RAMometer, remove all of the RAM chips
except your original 64MB chip, which is your "known good" one. Then
add one of the four new chips. Then launch RAMometer and let it do
its thing. Overnight is always good, but you can run it anytime.
Let it run for at least 1000 passes/cycles. If you get any errors
the chip is not suitable for use. Errors are never acceptable.
However, removing and reseating a chip that caused an error is
usually a good idea.
Test each of your four new chips that way (alone with the 64MB chip)
To add a bit of explanation and one modification to Nancy's recommendation.
RAMometer and any other memory checker cannot test all of your
memory, because some of your RAM will be occupied by the operating
system at the time of testing. So if the only memory installed is
the suspect memory, you will not test all of the suspect memory, only
a portion of it.
Hence, you should install your 64 MB DIMM in hopes that it will
provide the space for the OS while the suspect DIMM is tested.
However, on other machines, we (those who discussed this issue on
other lists) were never certain which end of the DIMM sockets gets
used first and there are indications that some memory is used at both
ends.
So we originated the RAM Sandwich method (term coined by Peter in
Japan, now in Sydney). In this test method you need two known good
DIMMs. You put the known good DIMMs in your two outer DIMM sockets
and the DIMM to be tested in the middle. If you don't have two known
good DIMMs you can still get there from here.
Put your 64 MB DIMM in the first socket and one of your new DIMMs in
the last socket. Put a DIMM to be tested in one of the middle
sockets. You may wish to have some small labels (I cut up mailing
labels) available to stick to the DIMMs as it is easy to get them
mixed up during testing.
The thing to remember is that if the test fails at this point, you do
not know the cause. A failure could be caused by the middle or by
the "unknown" end DIMM. However, if the test succeeds, then you know
that the DIMM in the middle is good, but you do not know if the DIMM
at the end is good. Working from these premises you should be able
to certify another DIMM as good (unless 3/4 are bad or something) and
proceed to test from there.
After inserting your DIMM sandwich, run RAMometer for about 1300
iterations. Most failures will occur within the first 200
iterations. However, I have seen DIMMs that would consistently fail
between 1200 and 1300 (always at the same number for a given DIMM),
so at least 1300 iterations are necessary for a thorough test. This
may require overnight testing, unfortunately. If you are in a hurry
(deadline on your project?) you may wish to just run 300 iterations
for now because that will catch the non-subtle defects.
Once you have two known good DIMMs, just leave them at the ends and
insert your suspect DIMMs in the middle for testing. Run RAMometer
and see if they pass or fail. Use the "Shut Down Background
Applications" and "Run Continuously" settings. Ramometer will stop
when it detects an error. There is no way to set the number of
iterations. You just have to stop by from time to time and see how
far it has gotten. It displays the number of iterations on its
window. Of course, you can time ten iterations, multiply by 30 and
have a pretty good idea of how long it will take to run 300
iterations.
Even if you buy new DIMMs it is still a good idea to run this test on
them. On older machines 128 MB FPM DIMMs were $120 until Velocity
Upgrades came along and offered them for $80. The price fell from
there, and while their early shipments were good, later stuff had a
high failure rate and required testing. I found three of ten bad in
one shipment and seven of eight bad in another shipment. And while
their more reputable competitors did better, they shipped some bad
memory too. When the prices get low, quality often suffers even from
the guys who have been around forever.
Even if a DIMM does not cause consistent system failures, it may
still have a defect and this testing will probably catch it.
Obvious defects are the type where a cell (a bit) in the RAM is stuck
to 0 or 1, regardless of what is written to it. So, if it's stuck at
1 and your computer writes a 1 to it, everything is fine. The
computer will read back a 1 from that cell. But if the computer
writes a 0 to that location, it is still going to read a 1 from it
later, which will cause some kind of unintended result, the severity
of which depends on what that 0 represented. These are the kinds of
errors that are probably caught in the first 50 iterations of
Ramometer.
Non-obvious defects depend on the surrounding cells. Physically, RAM
is a bunch of tightly packed structures made of layers of doped
silicon, oxidized silicon and metal. Their operation is dependent on
electrical charges in those structures. It is possible for the
electrical charges in surrounding cells to affect a nearby cell,
though they shouldn't. So there are sometimes subtle defects where
an error is only produced in a given cell, if the surrounding cells
have a particular pattern of data in them. It isn't possible in any
reasonable amount of time to try all the patterns of data that may
occur, but my hypothesis is that Ramometer changes test pattern as it
runs extra iterations and this is why some defects are not caught
until late in the testing and are always caught on the same iteration.
The good news is that these subtle defects aren't very likely to
occur or will seldom occur during use of your computer. Still, the
chance exists.
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
