Has anybody encountered defective RAM that would still pass memtestx86+?

[francescob]

I have 4GB of DDR2 RAM taken from a machine where Windows would crash constantly until I replaced them with other RAM modules of another brand. Before that, I ran memtestx86+ on the machine several times but all the passes completed without any error. So, the question is, how high is the chance it wasn't the motherboard being incompatible but the memory being actually broken? Has anybody ever encountered RAM that would pass the test and still be without any doubt not working properly?

[Mindovermaster Moderator]

How many passes did you run? It won't detect all problems on the first pass.

[Aergan]

Yes, especially Crucial Ballistix DDR-400 RAM modules.

 

That being said, I find that most of the time adjusting the timings to the manufacturer's specifications and manually setting the voltage resolves a lot of issues with new RAM modules (Especially if not a 'matched pair').

[Phouchg]

Memtest does not give a 100% definite answer, and it might not be memory that is broken unless it shows so on some other (preferably different, to avoid incompatibilities) system, too.

 

For me, however, time is often ticking. Goldmem (it costs, though) has given results once where Memtest didn't.

[francescob]

How many passes did you run? It won't detect all problems on the first pass.

I don't remember the number but I left it scanning for days.

[francescob]

Yes, especially Crucial Ballistix DDR-400 RAM modules.

 

That being said, I find that most of the time adjusting the timings to the manufacturer's specifications and manually setting the voltage resolves a lot of issues with new RAM modules (Especially if not a 'matched pair').

I tried changing the RAM settings several times, changing the frequency and timings to the lowest settings, but it didn't help. I also tried removing and reinserting the banks to make sure they were placed correctly. It was like the RAM was broken but I never managed to prove it in any way.

 

 

Yes, especially Crucial Ballistix DDR-400 RAM modules.

 

That being said, I find that most of the time adjusting the timings to the manufacturer's specifications and manually setting the voltage resolves a lot of issues with new RAM modules (Especially if not a 'matched pair').

 

The RAM brand was Corsair but the motherboard brand was Asus, and Asus motherboards are famous for having all sort of issues with RAM. I searched on google if anybody had issues with the same RAM model but unfortunately I didn't find any result to confirm it was incompatible.

[Biohead]

Yep - not too long ago I bought 2 x 8GB sodimms (Corsair) for my MBP.  Passed all the memtest checks no problems, but no matter what I did, when taking the MBP out of sleep mode it would just Panic and crash. Subsequent hours of memtesting didn't result in any errors found, and the error only occured with one of the sodimms. Got that specific one replaced and problem went away.

[n_K]

Yeah I had some unbranded DDR RAM years ago, windows would just freeze and bluescreen randomly, did a memtest on it and it passed with flying colours, tried a live linux and that too just messed up, changed RAM and problem solved.

[lars77]

Yes, I've seen bad RAM make it through less than 10 passes easily, e.g. it would only fail at around 8 passes.

 

re: your specific issue, did you check the capacitors on the motherboard?  Something weird like you're describing could indicate system instability if the old RAM vs the new RAM pulled different levels of voltage.  Of course if bad caps are the problem then your system will fail again eventually..

 

The only other thing I could think of is that your specific motherboard/PSU combination did not like your old RAM for some reason (again could be voltage).

[francescob]

Yes, I've seen bad RAM make it through less than 10 passes easily, e.g. it would only fail at around 8 passes.

 

re: your specific issue, did you check the capacitors on the motherboard?  Something weird like you're describing could indicate system instability if the old RAM vs the new RAM pulled different levels of voltage.  Of course if bad caps are the problem then your system will fail again eventually..

 

The only other thing I could think of is that your specific motherboard/PSU combination did not like your old RAM for some reason (again could be voltage).

The crashes were pretty often, several times every day so it was very strange that memtestx86+ was still not finding anything even when left running during all night. I though it could have been a voltage issue too, with the RAM screwing up when the motherboard was switching phases due to Intel's speedstep, but I never tried setting the CPU to run at full speed to verify that though I did change the RAM voltages. I'd love to send the RAM back since it had a 'lifetime' warranty but without being able to prove that it's actually broken it will be quite a problem.

[Karl L.]

Something to keep in mind is that Memtest86+ is conservative. A good rule of thumb is that if Memtest86+ says the RAM is bad, it is definatively faulty; if you have run at least three passes with Memtest86+, the RAM is probably not bad, but there is still a slight possibility that it is actually faulty.

[V9s]

I have 4GB of DDR2 RAM taken from a machine where Windows would crash constantly until I replaced them with other RAM modules of another brand. Before that, I ran memtestx86+ on the machine several times but all the passes completed without any error. So, the question is, how high is the chance it wasn't the motherboard being incompatible but the memory being actually broken? Has anybody ever encountered RAM that would pass the test and still be without any doubt not working properly?

 

Yep yep yep. Once left memtest running overnight on a pair of RAM sticks and they went through 20 passes without a problem, but my new PC would bluescreen constantly for some reason. Someone advised me to run memtest on a coldboot after leaving the PC off overnight and wallah...memtest showed the RAM was faulty....so try that too.

[UseLess]

I had a strange one, my RAM would pass any test (that i threw at it), never crash while running...BUT couldn't wake from sleep for some reason! When I replaced the RAM it was ok (so we all assumed it was a RAM problem). This was some corsair DDR2 sticks (brand new).

 

Probably the most bizarre computer problem I've run into in my time...

[srbeen]

I had an AMD X2 3800+ that didn't like a specific version of memtest86+. I used an older linux boot image and it failed as it should have, whereas the newer version passed without problem. It was test 7 I think that falsely passed IIRC, something to do with the chipset not reporting properly. Narrowed down the suspect stick and system has no more RR, freezing or SOD's.

 

Suggest verifying in another system with different processor if you are suspcious of the memory.

[n_K]

I had a strange one, my RAM would pass any test (that i threw at it), never crash while running...BUT couldn't wake from sleep for some reason! When I replaced the RAM it was ok (so we all assumed it was a RAM problem). This was some corsair DDR2 sticks (brand new).

 

Probably the most bizarre computer problem I've run into in my time...

That sounds like the RAM is fine but it's not getting enough power when in sleep mode. Possibly a PSU problem but more likely a motherboard problem, different RAM sticks require different amounts of power in sleep mode and if it undercuts it, the RAM will corrupt so when you try to power it back up, it's completely unusable.

[123456789A]

I have a couple of bad 30-pin 1MB SIMMs, but I don't know if this memtest thing will run on my 386 to try testing them.

[Noir Angel]

Do you have another computer you could test the ram in, and did the ram you replaced it with work properly? It is also possible that a faulty memory controller could cause similar problems, or that your board isn't compatible with the RAM

[lars77]

One other thing you can try.  If you see that memtest86+ is passing, yet you're still getting OS crashes & whatnot, try running a stress test on the system.  (e.g. Prime95)  That will at least reveal if there's another problem somewhere, maybe heat related, PSU, bad capacitors, etc.

 

(in your case not sure if it's worth it, but it might be good to run Prime95 as a sanity check to make sure everything actually *is* fixed)

[LittleNeutrino Veteran]

Al the time at work we run into defective ram that can pass memtest but will fail in voltage tests. And simply won't function.