Memtest86+ Passes

[Grinder]

So, does Memtest86+ just run continuosly, or does it eventually stop
after a fixed number of passes?

 

[John]

So, does Memtest86+ just run continuosly, or does it eventually stop
after a fixed number of passes?

Im pretty sure it runs continuously. Ive been running it over night or
longer the last few times Ive tried it but errors usually show up
right away or after few errors almost all the times Ive had problems.

 

[Grinder]

Im pretty sure it runs continuously. Ive been running it over night or
longer the last few times Ive tried it but errors usually show up
right away or after few errors almost all the times Ive had problems.

So if I have zero errors after a half dozen passes, I can afford to be
confident?

 

[kony]

So if I have zero errors after a half dozen passes, I can afford to be
confident?

No, it is only an assurance that the "odds" are your system
is stable enough to have no errors in (same type of memory
accesses) 2 X the length of time that memtest had ran.

AT a minimum, the length of time it should run is several
hours. That's on a system with a modest amount of memory.
With more memory, it may take longer for each pass (assuming
same memory speed/technology) and thus needs to run longer.
For a system with more than 512MB of memory it's good to let
it run at least overnight, ~ 8 hours. For even more
assurance, leave it running even longer, till you "need" to
start using the system again.

 

[John]

So if I have zero errors after a half dozen passes, I can afford to be
confident?

I would have said sure until recently. I never had memory problems
that I know of until recently. I used to run memtest over and over
when I had problems in the past but it never was the mem. I used to
test it a few loops . Then Id read posts where people insisted on
doing it over night. I tried that and nope no errors so I thought it
was a waste of time. Obviously I didnt have mem problems so it didnt
matter if it did it for 1 min or 10 hours.

However as Ive posted many times the last few months --- one of my
Kingston sticks seems like it was damaged when I was putting them in
and out testing my system with some new sticks. Maybe static elec.

Anyway, what happened was I didnt know it at first that it was my
memory --- my system acted BIZARRE. Overtime my system acted flakier
and flakier and windows would get corrupted eventually and start
rebooting my system and freezing. I would reinstall and everything
would be fine and then two weeks later crash ! I thought it was
drivers or my HD. My HD was a primary suspect cause I assume my
memory effected every piece on my system. My WD actually after the
initial usual partitioning and formatting for a new WIN XP install ,
started clicking and eventually refusing to bootup. And I developed a
partially bad pixel on my LCD ! One pixel was stuck on blue ! After
finally DUHHHHH testing everything else I decided to really test my
mem.

The thing I found in this case -- unlike a few times in the past with
neighbors systems etc where I found errors right away , these were
more subtle and took overnight or even two days !!!!! It was
maddening. I must have tested it 2o times cause it would come out OK
for hours then ZING error. Sometimes within 4 hours, other times 10
hours or longer. So like Kony says the odds are usually theyll show up
right away but its no guarantee. If you still have problems do it
overnight. Or always do it overnight to be more confident about it.

Anyway - after I low leveled formatted my new WD right before I was
about to send it back to WD which would be a hassle , it actually
showed up as bad with their diag utility software --- after low level
formatting its been fine ever since. And weirdly the bad pixel on my
LCD also vanished. have no idea if it was all a coincidence or not.

Ill mention something since its been posted at another website. I
posted that my old 3200 memory DDR branded as Kingston/Centon/PNY and
even Kbyte was all CL2.5 which works beautifully with my old ASUS
nforce2 board in dual mem mode and in my new 939 socket amd 64 system
now. Ive posted ad nauseum that I had huge problems with the new crop
of 3200 mem on sale - they (Kingston, PNY etc ) all changed it to CL 3
type. I blamed it on that. Well I dont know now.

A lot of people bought the compaq deal - an AMD 754 socket at a
website and people were buying a lot of the new rebate deals on the
CL3 mem. Lots of people but not everyone has had big problems with it
too. No one has pinned definitively to this but it may be the culprit.
Someone read at Toms Hardware on one motherboard review that he Tom
pointed out that double sided memory often caused problems with many
of the boards (not all) of the AMD 64 systems. These were 754 sockets.
Unfortunately no one has definitively tested this theory but it sounds
plausible as it was defintely a problem then according to TH. So maybe
my claim it was CL2.4 vs CL3 causing the problem is totally off (or
maybe not). I dont know. If people are having a problem check that
out ---- try single sided mem sticks and see if that cures it.

Im going to take a look at my old sticks tomorrow and see if they are
all single sided sticks. The thing that makes it hard is some people
with the compaq system claim the new cheapo mem they bought runs fine
in their systems but people have pointed out theres a mix of double
and single sided mem in the same Kingston brands so that may explain
it or maybe not who knows. Also wondered if a bios update can fix
these problems.

 

[Grinder]

I let it run for 4 hours, against 160 Mb of RAM. When I have time, I'll
let it run even longer. For now, though, I'm following up on another
promising source of this machine's freeze-ups -- the video card.

Thanks John and kony for your consideration though, as you would be the
two whose advice I would most value.

 

[kony]

I let it run for 4 hours, against 160 Mb of RAM. When I have time, I'll
let it run even longer. For now, though, I'm following up on another
promising source of this machine's freeze-ups -- the video card.

Thanks John and kony for your consideration though, as you would be the
two whose advice I would most value.

It is rare for a system to be so close to stable that it
takes several hours to reveal the first error, but it is
possible, I've observed it a few times myself... mostly when
overclocking, when the actual goal WAS to find the threshold
of stability, what it did take to first cause errors so a
margin could be planned.

Another factor is heat. Memtest is fairly light on
energy/heat so if a module is subject to higher temps during
certain high-load uses that can also have an effect- though
it's a bit harder to test since one doesn't generally try to
add heat to a system to simulate such loads... and I don't
either... but if you wanted to do so, I suppose you could
put a heating element or a light bulb a reasonable (safe)
distance from the components to elevate the temp some.

The other thing I like to do when possible is deliberately
o'c the memory bus- even if the system won't be overclocked
in use later. Providing you confirm that the system uses
same (Manual) memory timings at either bus speed, testing
the memory at 5% over the target (everyday) frequency is a
nice way to ensure an extra margin of stability.

 

[CBFalconer]

.... snip ...

Anyway, what happened was I didnt know it at first that it was my
memory --- my system acted BIZARRE. Overtime my system acted flakier
and flakier and windows would get corrupted eventually and start
rebooting my system and freezing. I would reinstall and everything
would be fine and then two weeks later crash ! I thought it was
drivers or my HD. My HD was a primary suspect cause I assume my
memory effected every piece on my system. My WD actually after the
initial usual partitioning and formatting for a new WIN XP install ,
started clicking and eventually refusing to bootup. And I developed a
partially bad pixel on my LCD ! One pixel was stuck on blue ! After
finally DUHHHHH testing everything else I decided to really test my
mem.

The thing I found in this case -- unlike a few times in the past with
neighbors systems etc where I found errors right away , these were
more subtle and took overnight or even two days !!!!! It was
maddening. I must have tested it 2o times cause it would come out OK
for hours then ZING error. Sometimes within 4 hours, other times 10
hours or longer. So like Kony says the odds are usually theyll show up
right away but its no guarantee. If you still have problems do it
overnight. Or always do it overnight to be more confident about it.

This is the sort of problem you will never have if you simply
install ECC memory (assuming the chipset can handle it). That sort
of thing would have been corrected and the actual correction
logged, so that you can take steps long before serious data damage
occurs.

However, you should turn off ECC operation before using a memory
tester.