Memtest Question

M

msny

2[H]4U
Joined
Sep 5, 2001
Messages
2,197
Can a system have one error in Memtest+ and still run fine,
even under load?
 
Yes for some time if your system never uses the bad cell. This can happen if you have more ram than your programs use. Although this can lead to data corruption and random crashes.
 
Which "memtest" are you running?

If it is some variant of memtest 86+ and you received a single error, you had better replace that RAM ASAP or figure out what is causing the RAM to be unstable.. voltage, timings, heat, etc.

Memtest 86+ is very unreliable form my experience... and if it even picks up a single error, then your RAM is not stable at all.

It does not put enough load on the system to detect any but the mot flagrant problems.

HCI memtest on the other hand, detects errors a lot easier and quicker than memtest 86+.
http://hcidesign.com/memtest/

I am talking seconds or minutes compared to memtest 86+ which usually takes hours or days to detect a single error unless the RAM is so horribly unstable that the OS has a hard time loading and/or being stable for more than a few minutes.
 
Even a single error deserves a replacement. I have never had an issue getting RMA's through G.Skill, Crucial, or Kingston, and listing the most recent version of Memtest86+ as my reason for declaring it failing.
 
cyclone3d said:
Which "memtest" are you running?

If it is some variant of memtest 86+ and you received a single error, you had better replace that RAM ASAP or figure out what is causing the RAM to be unstable.. voltage, timings, heat, etc.

Memtest 86+ is very unreliable form my experience... and if it even picks up a single error, then your RAM is not stable at all.

It does not put enough load on the system to detect any but the mot flagrant problems.

HCI memtest on the other hand, detects errors a lot easier and quicker than memtest 86+.
http://hcidesign.com/memtest/

I am talking seconds or minutes compared to memtest 86+ which usually takes hours or days to detect a single error unless the RAM is so horribly unstable that the OS has a hard time loading and/or being stable for more than a few minutes.
Click to expand...

Memtest86 +.

I have one error.
Ran it twice, each for about 10 hours, and it came up with the same error each time.
 
Maverick0984 said:
Even a single error deserves a replacement. I have never had an issue getting RMA's through G.Skill, Crucial, or Kingston, and listing the most recent version of Memtest86+ as my reason for declaring it failing.
Click to expand...

Ya thanks, I have 4 stick x 4gigs of G. Skill, see sig.

I will test each one, and see if I can find the bad stick.
Hopefully the error, is just on one stick alone.

I think I should have bought, matched pairs for duel channel.
 
If you are only seeing 1 error, it is 1 stick. Honestly, I place little value in the "matching" of pairs when getting 4 sticks.

I've always bought 2 pairs to get to 4 and never had any issues. I think people have just been unlucky and want to point the finger at them not being matched as a set of 4. As long as they are the same brand, model, and specs, you'll be fine.
 
msny said:
Memtest86 +.

I have one error.
Ran it twice, each for about 10 hours, and it came up with the same error each time.
Click to expand...

Do you have the timings set properly? Auto SHOULD work on 1333, but the board may not be setting it exactly right.

Have you tried bumping up the RAM voltage by .1 or .2 ?
Sometimes it takes a little more juice to get the RAM stable if you are filling all the slots.

You may also need to bump up the memory controller voltage by one notch.
 
cyclone3d said:
Do you have the timings set properly? Auto SHOULD work on 1333, but the board may not be setting it exactly right.

Have you tried bumping up the RAM voltage by .1 or .2 ?
Sometimes it takes a little more juice to get the RAM stable if you are filling all the slots.

You may also need to bump up the memory controller voltage by one notch.
Click to expand...

While correct. You shouldn't have to run the sticks above their voltage spec to get their spec speeds and timings. I would still probably replace it, it'd be a sign of degradation.
 
cyclone3d said:
Do you have the timings set properly? Auto SHOULD work on 1333, but the board may not be setting it exactly right.

Have you tried bumping up the RAM voltage by .1 or .2 ?
Sometimes it takes a little more juice to get the RAM stable if you are filling all the slots.

You may also need to bump up the memory controller voltage by one notch.
Click to expand...

Yes, all the timings are exactly the same, and set to auto..
 
OK, here are the results.

I retested each 4gb stick in bank A, slot one separately.
BIOS set to default optimize.
I ran 2 passes on each.

I labeled each stick 1 to 4 first.
Bank A 1-2.
Bank B 3-4

Bank A stick 1 - one error was found.- bad
Bank A stick 2 - no errors found - good
Bank B stick 3 - no errors found - good
Bank B stick 4 - stick would not boot the system, even after BIOS reset - bad
(the board has a compatibily mode for RAM and this stick failed that also)

I then put the two good sticks in Bank A and reran the tested again,
no errors.

So I have 8 gigs presently, and will RMA the two bad ones to G. Skill.
When I get two new sticks, i will repeat the tests on them.

Thanks all, for your input.
 
Last edited:
Maverick0984 said:
While correct. You shouldn't have to run the sticks above their voltage spec to get their spec speeds and timings. I would still probably replace it, it'd be a sign of degradation.
Click to expand...

Agreed, if there bad, there bad, i want new ones, no use wasting time
with them.
 
Maverick0984 said:
If you are only seeing 1 error, it is 1 stick. Honestly, I place little value in the "matching" of pairs when getting 4 sticks.

I've always bought 2 pairs to get to 4 and never had any issues. I think people have just been unlucky and want to point the finger at them not being matched as a set of 4. As long as they are the same brand, model, and specs, you'll be fine.
Click to expand...

Sure, just thought those would have a better chance of being good.
I've built several systems in the past and have always using matched pairs
as preferred. For some reason, i didn't buy them this time??

Appreciate your advice.
 
Last edited:
Several months ago, a person reported crashes whenever he used his sound card, but MemTest86 or MemTest86+ found no errors. Then he ran Gold Memory 6.93, and it reported one bad location, and then another 9 hours passed before another error was found (the same one). Replacing the faulty module or fixed the problem.

Many years ago, we had some computers that ran one program all the time, and every several weeks, the program would return strange results. The memory in those machines passed testing, but one day somebody decided to run the test much longer, and every machine had 1-5 bad memory locations. After the memory was replaced, those computers never acted up again. Also some machines that never had trouble with the same application software had their memory tested, and all of it passed.

Diagnostics that have never found errors for me:

MemTest86+
DocMem
MemTest (Windows based, AKA MemTest HCI)
Microsoft memory diagnostic
Gold Memory 6.93

Only Gold Memory 5.07 and MemTest86 have reported errors in modules I've tested. That doesn't necessarily mean the other programs are bad (except DocMem -- the worst), only that you may have to try several tests before one works well enough.
 
larrymoencurly said:
Several months ago, a person reported crashes whenever he used his sound card, but MemTest86 or MemTest86+ found no errors. Then he ran Gold Memory 6.93, and it reported one bad location, and then another 9 hours passed before another error was found (the same one). Replacing the faulty module or fixed the problem.

Many years ago, we had some computers that ran one program all the time, and every several weeks, the program would return strange results. The memory in those machines passed testing, but one day somebody decided to run the test much longer, and every machine had 1-5 bad memory locations. After the memory was replaced, those computers never acted up again. Also some machines that never had trouble with the same application software had their memory tested, and all of it passed.

Diagnostics that have never found errors for me:

MemTest86+
DocMem
MemTest (Windows based, AKA MemTest HCI)
Microsoft memory diagnostic
Gold Memory 6.93

Only Gold Memory 5.07 and MemTest86 have reported errors in modules I've tested. That doesn't necessarily mean the other programs are bad (except DocMem -- the worst), only that you may have to try several tests before one works well enough.
Click to expand...

Good observation.

I had run Memtest for Windows, ASUS diagnotics memory test and
Windows 7 Memory test. NONE of these others found errors
except for Memtst86+..

Lesson learned, run more then one.
 
cyclone3d said:
Which "memtest" are you running?

If it is some variant of memtest 86+ and you received a single error, you had better replace that RAM ASAP or figure out what is causing the RAM to be unstable.. voltage, timings, heat, etc.

Memtest 86+ is very unreliable form my experience... and if it even picks up a single error, then your RAM is not stable at all.

It does not put enough load on the system to detect any but the mot flagrant problems.

HCI memtest on the other hand, detects errors a lot easier and quicker than memtest 86+.
http://hcidesign.com/memtest/

I am talking seconds or minutes compared to memtest 86+ which usually takes hours or days to detect a single error unless the RAM is so horribly unstable that the OS has a hard time loading and/or being stable for more than a few minutes.
Click to expand...

Interesting...memtest86+ from CD found the error, memtest for windows did not
find anything....:eek:

And it found it in the first pass...
 
Did you run several simultaneous instances of HCI memtest so that it covered all of your ram?
 
drescherjm said:
Did you run several simultaneous instances of HCI memtest so that it covered all of your ram?
Click to expand...

Yes, per instructions, I ran 4 simultaneous instances of HCI, overnight.
I did this for two nights in a row.
It found no errors.

I dont question why this test did not find it and the other did, I'm just glad
I found something that led me to the issue.

Over the last couple of weeks my system would become unresponsive
from time to time. The display would go black, but the system stayed up & running.
I'm thinking it was due to one bad stick of RAM now (see post above).
 
Last edited:
You must log in or register to reply here.
Back
Top