[G4] Memory

Garth Fletcher garth at JacqCAD.com
Thu May 25 06:37:50 PDT 2006


SLarsonIH at aol.com wrote:
> Subject: [G4] Memory
> 
> I just bought a QS for my mom, and I have a question about memory. How do I 
> tell if the memory sticks are good or bad, for sure? The QS came with three 
> 512MB sticks. They show up in ASP. One of the disks that came with the QS was an 
> Apple Hardware Test disk. I was curious what was on it, so I ran the test 
> program. It says two of the sticks have errors. I ran the test numerous times with 
> one stick, in different slots, pairs in different slots, in different orders, 
> etc. Same results. I then found and downloaded MemTest, an OSX program for 
> testing ram and it says the same thing. I called the guy and he says the QS 
> never gave him any problem. He's willing to buy me ram if that will make me happy. 
> I don't want him to waste his money if these tests don't really mean 
> anything. A trusted Apple tech friend of hos said that the tests are not reliable. So, 
> outside of using the computer for months, is there any sure fire way to know 
> ifthey are good or not?
> 
> Thanks
> STeve

What the trusted friend meant when he said that "memory tests
are not reliable" is the following:

   passing the test does NOT reliably guarantee that the memory is OK.

Finding a "stuck" bit is easy; all test programs go through setting
every bit and checking that it sets, and clearing every bit and checking
that it got cleared.

Unfortunately, it is possible for bits to be temperature sensitive - so
that they work perfectly until the chip has reached a higher temperature
at which point they become unreliable.  That means that running the test
soon after startup may not spot problems which might be caught after the
machine has been running hard for hours on a hot day...

The big problem is that memory chips can be "pattern sensitive", meaning
that a failure can depend on a specific pattern of reads and writes to
*nearby* bits.  There are simply too many possible patterns to fully
test each one - it would take decades.

Sometimes one test program will find no problem while another one will -
because they use slightly different patterns.  Sometimes neither test
program will find any problem, but a certain program will misbehave -
again because the program "found" the right pattern to trigger the error.

Conclusion:
  if *any* memory test finds a problem you DO have BAD memory
  if no memory test finds a problem you MAY have good memory


Also, bear in mind that problems with the motherboard and/or power
supplies can cause memory to malfunction.  It seems you did the
right tests to check by moving around the memory sticks - if the
failures moved with the sticks then you can be confident that
the problem lies in the stick rather than in the slot. And if
some sticks are always OK then that suggests the motherboard and
supplies should be OK...

HTH
-- 
Garth Fletcher


More information about the G4 mailing list