Memtest86 reports no error at the store, but reports errors here!!

[Admin]

I don't know what is going on with my DDR Ram memory here.
Earlier this week, I have run a memtest86 in the office and found out that
the memory was bad (lots of red error messages on the screen).
I took it to the computer store, they tested it, and they found the same
errors. The guy said he cleaned it, set it back into place, ran the
memtest86 again, and it went through without any error.
Back into the Office, I have the memory problems again, I run memtest86
and... The red warnings showing the errors again.
The memory and the mobo are all brand new, they're not dusty at all.
I did the test with another memory, and there are also these errors.
Should I change both of them? Is there a way to "clean" them, or to fix
them?

 

[Alceryes]

errors. The guy said he cleaned it, set it back into place, ran the

memtest86 again, and it went through without any error.

That sounds like BS to me. Were you a witness to him running the memtest? Is
there a substantial (10º +) difference in temperature between your office
and the computer shop? Did he say that the system's airflow was cut off by
dust?
I'd bring the computer back to the guy (in the afternoon, when it's hottest)
and have him run the memtest in front of you.
--


"I don't cheat to survive. I cheat to LIVE!!"
- Alceryes

 

[kony]

I don't know what is going on with my DDR Ram memory here.
Earlier this week, I have run a memtest86 in the office and found out that
the memory was bad (lots of red error messages on the screen).
I took it to the computer store, they tested it, and they found the same
errors. The guy said he cleaned it, set it back into place, ran the
memtest86 again, and it went through without any error.
Back into the Office, I have the memory problems again, I run memtest86
and... The red warnings showing the errors again.
The memory and the mobo are all brand new, they're not dusty at all.
I did the test with another memory, and there are also these errors.
Should I change both of them? Is there a way to "clean" them, or to fix
them?

Memtest checks for memory errors. That doesn't mean the
memory is bad, necessarily, only that the specific module as
installed at the moment in a _specific_motherboard_ is not
stable. It can be (and often is) the motherboard itself
that is partly to blame.

It doesn't matter what the "guy" determined, if the memory's
spec is supposedly suitable for the system but your system
isn't stable running it, you should return it for a
refund... most preferribly NOT an exchange for exact same
brand/model of memory because of this potential for the
motherboard or motherboard-memory interaction (compatibility
issues) problem potential.

Check your bios to see if it's running "auto" or "SPD"
timings. If bios settings are misconfigured, correct that.
You might try clearing CMOS, if you can set any necessary
bios settings yourself afterwards. "Sometimes" a bios
update can help. If you know how to manually set memory
timings that can help too, but is generally best done when
you "need" to keep/use a particular module for some reason,
not when you have the possiblity of being able to return the
module instead.

 

[w_tom]

Memory must run fine in every type of bus condition from 20
up. This is an arbitrary number provided only to demonstrate
the process. Your computer may be a 40. His tester may be a
70. Therefore the memory will work everytime in his tester
AND memory is still defective - it will not work on a 50 bus
or less.

One way to make errors obvious is to use a hairdryer. Did
this once with memory. It would work just fine during test.
Computer would crash periodically. Vendor refused to admit
memory was bad until I warmed it with a hairdryer on high.
Then it failed in his tester every time. Warm memory changes
defective memory from a 50 to a 20. If memory was good, it
would remain at above 100.

Memory calls it "pig's heaven" - thinks 70 degree F and 100+
degrees F are ideal and comfortable temperatures. If you
don't think so, then get the manufacturer's data sheets -
provide numbers that say otherwise. Intermittent memory tends
to show its pathetic nature only when heated to 100+ degrees -
a hairdryer on high.

BTW what do we do for flight hardware? We run those ICs
right up at the maximum spec temperatures, cycle temperatures,
and test again. Contrary to myth, this causes no damage, no
significant stress, and finds intermittents before they morph
into hard failures. Learn from how testing was done even in
the 1960s. Your memory should remain at well over 100 even
when heated by a hairdryer.

BTW, if the memory vendor said he cleaned the memory pins,
then you know his technical knowledge is fully based in
myths. Those electrical connections are self cleaning. And
yet still some naive will take an eraser and clean those
contacts. Classic myth purveyor who would make a good FEMA
Director.

 

[Shep©]

I don't know what is going on with my DDR Ram memory here.
Earlier this week, I have run a memtest86 in the office and found out that
the memory was bad (lots of red error messages on the screen).
I took it to the computer store, they tested it, and they found the same
errors. The guy said he cleaned it, set it back into place, ran the
memtest86 again, and it went through without any error.
Back into the Office, I have the memory problems again, I run memtest86
and... The red warnings showing the errors again.
The memory and the mobo are all brand new, they're not dusty at all.
I did the test with another memory, and there are also these errors.
Should I change both of them? Is there a way to "clean" them, or to fix
them?

Software memory testers aren't worth the cost of the download let
alone running.
My 10 penn'th

 

[philo]

I don't know what is going on with my DDR Ram memory here.
Earlier this week, I have run a memtest86 in the office and found out
that the memory was bad (lots of red error messages on the screen).
I took it to the computer store, they tested it, and they found the
same errors. The guy said he cleaned it, set it back into place, ran

<snip>

Is the RAM...the "generic" type that the store has on special?
I'd return it for a know-good name brand...even if it's more expensive.


Or... try clocking down the RAM a bit.

 

[larry moe 'n curly]

Earlier this week, I have run a memtest86 in the office and found out that
the memory was bad (lots of red error messages on the screen).
I took it to the computer store, they tested it, and they found the same
errors. The guy said he cleaned it, set it back into place, ran the
memtest86 again, and it went through without any error.
Back into the Office, I have the memory problems again, I run memtest86
and... The red warnings showing the errors again.
The memory and the mobo are all brand new, they're not dusty at all.
I did the test with another memory, and there are also these errors.
Should I change both of them? Is there a way to "clean" them, or to fix
them?

You deserve a refund because you paid to have the computer fixed or at
least diagnosed exactly, not to have a clueless guy tinker with it and
do something useless. I can't imagine dust on a new mobo and memory
being thick enough to affect anything, so why didn't the guy simply
chant and pray? In the repair report, what are the values of the
voltages, not just the PSU's but also for the CPU core, memory bus, and
AGP slot?

I'd remove and reseat the memory, but if that doesn't help (it probably
won't), change the memory parameters in the BIOS. Sometimes the
automatic settings (SPD) are wrong for a module because it's rated
faster than the mobo's memory speed (i.e., PC3200 memory in a PC2100
mobo) or the module maker was lax about screening its chips for speed
and quality and used marginal ones (often the case when the chip
maker's part numbers are missing and the chips have only the module
maker's marks on them) that may require underclocking. I've
experienced this even with some national brands, most recently
Kingston.

A memory diagnostic has to be run for several hours, maybe overnight,
to find all errors, and while MemTest86 has gotten high marks (see
www.realworldtech.com review), you should also use Gold Memory
(www.goldmemory.cz) to find errors it misses. Gold has also missed
errors for me, so it's important to use both diagnostics.

 

[larry moe 'n curly]

Software memory testers aren't worth the cost of the download let
alone running.

How much does a good hardware memory tester cost, and do most computer
shops with hardware testers have the good ones? I've seen some selling
for $500 US, and one store said that theirs cost $700.

If software testers are useless, why have they found defects in some of
the modules I've bought (only those with no-name chips, not those with
chip makers' part numbers on them) and have done so consistently with
each pass?

Also how do you explain this:
www.realworldtech.com/page.cfm?ArticleID=RWT052001232443 and this:
www.realworldtech.com/page.cfm?ArticleID=RWT120901222920 , where some
software testers did as well as the hardware testers from Ultra-X?

 

[larry moe 'n curly]

Is the RAM...the "generic" type that the store has on special?
I'd return it for a know-good name brand...even if it's more expensive.

Or... try clocking down the RAM a bit.

What's a good name brand? Because even Kingston, PNY, K-byte, Centon,
and maybe even Corsair now use no-name chips in at least some of their
modules. I buy locally from stores with 100% money-back guarantees
because of the high defect rates I've experienced with some of these.

 

[Bob]

What's a good name brand? Because even Kingston, PNY, K-byte, Centon,
and maybe even Corsair now use no-name chips in at least some of their
modules. I buy locally from stores with 100% money-back guarantees
because of the high defect rates I've experienced with some of these.

What ever happened to Micron?

http://www.micron.com/



--

Greatest Movie Line Ever
http://home.houston.rr.com/rkba/MovieLine.wmv

"What is history but the story of how politicians have
squandered the blood and treasure of the human race?"
--Thomas Sowell

 

[D & B]

Shep©'s posts aren't worth the cost of the download let alone reading...

Software memory testers aren't worth the cost of the download let
alone running.

How much does a good hardware memory tester cost, and do most computer
shops with hardware testers have the good ones? I've seen some selling
for $500 US, and one store said that theirs cost $700.

If software testers are useless, why have they found defects in some of
the modules I've bought (only those with no-name chips, not those with
chip makers' part numbers on them) and have done so consistently with
each pass?

Also how do you explain this:
www.realworldtech.com/page.cfm?ArticleID=RWT052001232443 and this:
www.realworldtech.com/page.cfm?ArticleID=RWT120901222920 , where some
software testers did as well as the hardware testers from Ultra-X?

 

[w_tom]

When a poster can say why he made a statement, only then do
we have a post of merit. Unless one can tell us what tests
the hardware tester and software tester perform, only then do
we have minimal information on the reliability or value of
either tester. Statements made without significant reasons
why akin to wasting bandwidth.

I was writing memory tests long before PCs even existed.
Many of the traditional tests include marching ones and
marching zeros, full one and full zeroes, the checkerboard
test, and other 'standard' tests. In one case, the board has
passed every marching zero test and every marching one test -
then failed a 'next to last' test. Why did that memory
location pass every test previously? Memory location only
failed when enough adjacent memory locations were opposite
logic state long enough to leak charges. Even the
manufacturer tests would not detect this memory failure. Not
only did I detect a failure, BUT, I also must know why.
Without also knowing why, then I would not have sufficient
knowledge to post.

Meantime, defined was how to make even a cheap tester
superior to more expensive testers. Heat memory with a
hairdryer on highest heat. This from someone who was writing
computer memory tests even for core memory. When did you work
on core memory computers? This from someone who would find
defective memory after standard test said memory was OK - but
computer still crashed. I was writing memory tests probably
before most lurkers even existed. I cannot say enough about
testing under elevated temperature - which is a delightful and
normal temperature to functioning memory.

Again, once a poster demonstrates a grasp of how memory
fails, how the many testers actually work, and provide the
"why" details, only then are we ready to enhance our
knowledge. Currently almost nothing has been posted here
about a good or inferior tester .... because reasons why and
underlying grasp of the concepts are not included in that
post.

The author of that cited review (realworldtech.com) feels
that 'poor quality or slow memory' caused problems. Nice.
But it tells us nothing. Define 'poor quality'? Define
'slow'? He makes no attempt to even discuss basics of why
memories fail - what symptoms cause memory failure. I suspect
he may be just inventing these definitions because he did not
actually learn what was defective in each memory.

To provide useful information, the author must have a grasp
of basic concepts - the underlying theory - AND support his
theories with experimental evidence. Both conditions are
necessary to understand - to know. For example, how long
after the memory address is provided does memory then provide
data from that location? Specific numbers exist for this
parameter. How long can memory hold data before requiring a
refresh? Again a parameter that memory testers would test. Is
voltage driver for that data output sufficient to drive a
fully loaded bus? Not hard to test for and could be listed as
one of the many tests for a good hardware tester.

So many parameters to test. So many ways a memory can
fail. So few of these parameters are tested for let alone
even mentioned. Author makes no effort to define each
possible failure condition let alone suggest he even knows why
memories fail.

IOW his tests provides nothing more than speculation based
upon random defective memories. Defects that we don't even
know why they are defective.

The problem. This citation that is not based upon even
minimally acceptable technical facts is offered as the best
proof of good memory testers. Unfortunately among many
technically knowledgeable computer professionals, that
woefully inadequate article constitutes a proof - rather than
nothing more than outright speculation.

Meanwhile, one condition that tends to aggravate many
marginal conditions into a hard failure mode - heat. This is
also why so many naive computer assemblers instead recommend
more fans for the system - to cure the symptoms rather than a
hardware problem. Heat is a most effective diagnostic tool
for electronics - especially memory.

 

[MrGrumpy]

memory test may report errs if the memory is not fully compatible with both
mobo and cpu, apart from if the memory is itself faulty.