How reliable is Memtest in finding errors?

J

JoeUser

2[H]4U
Joined
Mar 30, 2010
Messages
3,919
I ran Memtest the other day overnight...did something like 8 or 9 passes and didn't find any errors.

This is where my questioning its reliability comes in.

I tested the memory due to my system getting a BSOD once or twice a day. I have found using BlueScreenViewer that the cause of the BSOD seems to be "ntoskrnl.exe" which is the part of Windows that deals with memory management.

So...that leaves it pretty open to the issue. Considering my RAM tested OK I decided to install Windows on my known working HDD instead of my SSD (thinking the SSD was the source of the problem). Well as luck would have it I got another BSOD due to "ntoskrnl.exe" crashing on my HDD.

So I can reliably say it's not a storage disk problem. I have now reinstalled Windows a few times, three times on the SSD and twice on the HDD with the same problems.

I don't know if "ntoskrnl.exe" deals with GPU RAM though...so I'm basically stuck with the problem either being my RAM (which, again, tested fine with no errors) or my GPU (which runs fine with no errors).

So, is Memtest the end all of RAM reliability? Is it possible that it didn't find something after 8 hours? Again, if something is wrong with it it doesn't seem to be a MAJOR issue as I'm only getting a BSOD once or twice a day versus many times as was the case when I did have some truly bad RAM in the past.

Is there something else I can try to test the RAM?

Thanks!
 
I find it is pretty good at finding them.
Post all the information you have on the blue screen including all the info from event viewer.
 
I just remembered that a week ago or so I had a bad BIOS update so I ordered a new BIOS chip that came with the latest BIOS.

So I just now inserted my bad BIOS chip, reflashed it back to the OG BIOS I was using before all this started happening. All I can do now is see if maybe the new BIOS was causing issues. As far as I can tell my system is completely back to pre-BSOD specs.

Can't believe I didn't realize this earlier, guess I just assumed that the BIOS wouldn't cause issues...but that's yet to be seen...
 
Is it possible that it didn't find something after 8 hours?
Click to expand...

Yes. Very possible for marginal cells. These are the ones that exhibit rare bit flips. At work I had a system that would BSOD shortly after boot but took 72 hours to fail memtest86+. After replacing the ram the problem was gone. Also that was not the only instance of this that I have seen.

Try several simultaneous instances of hclmemtest in windows.
 
I'm not sure how you tested your RAM, but I did have to RMA a set a few years ago. When the RAM was tested as a set in memtest, I didn't get any errors, but when I tested the sticks individually, one of the sticks did fail.
 
It helps to remember that memtest, memtest86, and memtest86+ are all different programs.

The only one worth a crap nowadays is memtest86+, 4.20 is the most recent version. Don't trust anything else, run for 24hrs to get accurate results (sometimes they won't fail till 3-5 passes in)
 
HCI Memtest can sometimes find that rare bit that memtest86+ wasn't able to find. As I recall, HCI stresses the memory subsystem a lot more, so it can detect an unstable IMC as well.
 
Driver issue. Make sure you have the chipset drivers installed and all the latest drivers from the maker of the hardware for everything.
 
I have seen this before when windows update had a bad driver that kept on getting installed after a clean install which is why you need the drivers direct from the manufacturer.
 
drescherjm said:
Yes. Very possible for marginal cells. These are the ones that exhibit rare bit flips. At work I had a system that would BSOD shortly after boot but took 72 hours to fail memtest86+. After replacing the ram the problem was gone. Also that was not the only instance of this that I have seen.

Try several simultaneous instances of hclmemtest in windows.
Click to expand...

I had an openbsd router/firweall running on a pentium pro that would randomly kernel panic almost once a month....it took memtest between 24-48 hours to find the bad dimm.
 
stormy1 said:
I have seen this before when windows update had a bad driver that kept on getting installed after a clean install which is why you need the drivers direct from the manufacturer.
Click to expand...

I want to agree..I had a similar problem on an older build that was being caused by a Windows Update provided driver...Check for updates to everything manually, and report back..
 
Or just never install a driver through windows update. They show as optional for a reason.
 
Last edited:
Appparantly better than MemTest HCI, not as good as Gold Memory, and PHD's RST diagnostics beat everything, according to the most recent evaluation available, done back in 2001:

http://www.realworldtech.com/memory-diagnostics-1/

http://www.realworldtech.com/memory-diagnostics-2/

Does anybody know of something newer?

Try another diagnostic because nothing finds every error. One person ran MemTest86 or MemTest86+ for like over 12 hours, but it reported nothing. Then he ran Gold Memory, and it found an error in 76 minutes but took another 9 hours to find that error again:

goldmemorytestresult.jpg
 
You should of, but didn't tell us what Windows you are using.

A free option not mentioned: Vista and 7 both have a built-in pre-boot "Memory Diagnostic Tool" that's fairly decent. Has a few options for how complete/long the testing is (best to use "Extended" with multiple passes).

For further details on what Goros posted:

HCI Memtest is feature-crippled freeware and I'm not a fan of it running inside Windows. Can usually find memory problems more reliably running pre-boot utilities.

Memtest86+ is usually acknowledged as the third-party leader in free test utilities. Hasn't been updated since 2011, however.

And then there is Passmark's Memtest86 (free download). After two years, its finally been updated (Jan.).

"GoldMemory" is non-free shareware (30 day trial).

As noted by others, keep all your drivers up to date, especially video & sound. Download from the manufacturer's sites, _not_ Windows Update.
 
Fritingo said:
Memtest86+ is usually acknowledged as the third-party leader in free test utilities. Hasn't been updated since 2011, however.
Click to expand...
But it's in the process of being updated, most recently as 8 days ago: ver. 5.00 beta 4

It worked fine with my G41 mobo but showed lots of errors with my Z68 running the same memory or any other, unless I turned off MemTest86+'s parallel CPU core option. With the Z68 it also crashed with a green screen with text in it, as did MemTest86 ver. 4.1.0 (no errors shown with MemTest86 ver 3.5b).
 
I hate ram problems cuz its a pita trying to definitively diagnose it as a ram problem, esp if you have 8+ gig of ram, takes fkin ages usually.
 
EVIL-SCOTSMAN said:
I hate ram problems cuz its a pita trying to definitively diagnose it as a ram problem, esp if you have 8+ gig of ram, takes fkin ages usually.
Click to expand...
naw it isn't that bad, what is fun when they decide to keep the office at 65 degrees to go green all winter and 30+ 1year old dell workstations go all flaky with memory related blue screens when they ran fine when it was 72 in the office.
Yep most every dimm was under spec in width at the connector and every socket was out of spec the other way.
Dell said too bad out of warranty not our problem.
 
You must log in or register to reply here.
Back
Top