[SOLVED] Memory behaving oddly, no errors despite random crashes

[LeoF2]

Hey all,

I run a 16GB DDR3 system in a 4x4GB setting. The memory has been like this for well over a year now (before then I had 8GB in a 2x4GB configuration and these two sticks have worked for seven years now, I'm 100% positive they're good) and only recently the system has been giving me some odd behaviour which seems to be tied to memory... Except I can't troubleshoot it at all.

From time to time, something will crash. Mostly it's the browser, but often times memory-heavy games will crash, too. This happens especially when the memory usage is nearing 80%+. The errors are generic, the usual 'Whoops, this tab has crashed' or something. I have a 4GB pagefile in a working SSD, and placed it there just in case it was pagefile related... Nope. The crashes are random and the system itself never crashes or bluescreens. The few games that give me errors instead of just crashing all point to memory, such as 'couldn't write/read to memory at address xxxx'. Surely one of the sticks is bad, right?

I've ran Memtest86+ for a few hours on the system - nothing. Previously I've had a bad stick of RAM and it detected errors in like 5 seconds of Memtest. I have also tried OCCT's software tool for error detecting in memory, which runs off the OS and might find things Memtest86 misses, but also nothing. Nothing crashed while the test ran and I've ran it multiple times, considering it's much easier than setting up Memtest86. Hell, I'm running it right now and memory is at 95% usage. I had a spare 2GB stick sitting around and so I've swapped it for these not-so-recently acquired 4GB sticks, one at a time, guessing that something was off with either one of them, or both. The crashes still happen and both memory tests still report no errors whatsoever.

Other than a bad memory slot, which I unfortunately have no way of testing but don't think would cause this sort of trouble (might be wrong on this one tho), and a GPU upgrade I've had for well over a year now, I can't think of anything else that could be related to the issue. The crashes are very rare and random and sometimes even at heavy memory loads, don't happen. But they've been happening for months now and I don't think it's a coincidence. Any ideas?

 

[SamirD]

A bit of a toughy for sure. If it passes through a while run of memtest with zero errors that will add to the mystery.

Since it is crashing in windows, try using hci design's memtest:

MemTest: RAM reliability tester for Windows

It's just a nice single exe and you'll have to run it multiple times to get all your memory occupied. Watch the usage in task manager and don't go too high or it will actually start swapping and testing that, which might not be a bad idea for a little bit of overlap in case that's the issue.

The other thing I would try is using just 8GB of memory and see what happens. Or better yet, use 12GB removing just a single module. And keep trying that removing a different module each time if 12GB works without a crash. Eventually, you'll either find the bad module or this test didn't prove anything, lol.

One thing I would also consider if your system can handle it is upgrading to 32GB of memory while you can. If a new set of this doesn't work right, then it almost certainly a motherboard/cpu/power issue.

Speaking of power issue, I had a weird issue similar to this where a system would just randomly crash and reboot and it turned out to be a power supply issue.

I can't think of any more ideas atm, but hopefully this helps a bit!

 

[LeoF2]

A bit of a toughy for sure. If it passes through a while run of memtest with zero errors that will add to the mystery.

Since it is crashing in windows, try using hci design's memtest:

MemTest: RAM reliability tester for Windows

It's just a nice single exe and you'll have to run it multiple times to get all your memory occupied. Watch the usage in task manager and don't go too high or it will actually start swapping and testing that, which might not be a bad idea for a little bit of overlap in case that's the issue.

The other thing I would try is using just 8GB of memory and see what happens. Or better yet, use 12GB removing just a single module. And keep trying that removing a different module each time if 12GB works without a crash. Eventually, you'll either find the bad module or this test didn't prove anything, lol.

One thing I would also consider if your system can handle it is upgrading to 32GB of memory while you can. If a new set of this doesn't work right, then it almost certainly a motherboard/cpu/power issue.

Speaking of power issue, I had a weird issue similar to this where a system would just randomly crash and reboot and it turned out to be a power supply issue.

I can't think of any more ideas atm, but hopefully this helps a bit!

Thanks for the reply.

I've let this test run for about 35 minutes now, no crashes or errors. Part of what makes troubleshooting this difficult is that the crashes aren't common, but have been happening for enough time now that I'm positive something, somewhere, isn't working well.

I'd suspect a faulty motherboard/PSU if the crashes were system related, but they're not, which again makes this issue really odd to me. Only software crashes - the OS never bluescreens and the computer never freezes. I've had explorer.exe crash, once, without the system freezing or rebooting. It's really odd how all issues point to faulty memory, yet any and all memory testing software passes without issue. And only applications crash, like browser tabs and (very rarely) games, the system is yet to give me a hard reboot/BSOD.

I've also considered a faulty GPU, and to be specific, faulty VRAM modules to be the culprit. Modern browsers and Windows' DWM use video memory and perhaps bad VRAM was to blame... But I find no artifacts in games and no other issues except for the odd (memory-related) crash and VRAM testing software, and I've only used OCCT's inbuilt tool for this, passes without issue, too.

I'll try formatting the system if this keeps up, since this seems to be the kind of niche, bizarre thing that a system reset might just fix. Any other ideas are welcome!

 

[SamirD]

Ooooo....this isn't memory at all! It's storage! If applications are getting bit errors from the drive, they can crash like that (I've seen it).

I'd download the ultimate boot cd and run parted magic from it and then run a short smart test on your drive. If it fails, there's your culprit. If it passes, I'd run the long smart test. You can also run the manufacturers tools, but they typically need windows hence why I like using parted magic.

 

[LeoF2]

Ooooo....this isn't memory at all! It's storage! If applications are getting bit errors from the drive, they can crash like that (I've seen it).

I'd download the ultimate boot cd and run parted magic from it and then run a short smart test on your drive. If it fails, there's your culprit. If it passes, I'd run the long smart test. You can also run the manufacturers tools, but they typically need windows hence why I like using parted magic.

Thanks again for the reply and sorry for the late answer. I hope I'm not clogging the forums too badly with this thread, but hopefully I can help some poor soul out with this issue.

TLDR: You were right. This was not a memory issue at all - it was storage.

I had already ran some HDD testing for bad blocks/SMART readings before and did it again, nothing came out of it and the HDD reported being at 100% health, which is a SMART reading IIRC. However, what I hadn't checked was that the poor drive was sitting at a whopping 41% fragmentation rate. The drive wasn't bad, but fragmented to hell and back and I figure this fragmentation was causing all sorts of delays in reads that shouldn't be there and some applications (like browsers) couldn't handle that and just crashed. It took me over one day of defragmenting using Defraggler, since Windows' tool seems not to defrag 'large' files (over 64MB), but it's been over a week so far and I've had zero crashes or hangs.

The drive itself is fully functional despite the 7 years it's been used for, but it hadn't been defragmented in a long, long while... If it ever was. So just to recap, if you're having:

• Web browser (Firefox, Chrome) crashes without error messages, particularly at high memory usage;
• Game crashes pointing to bad memory reads/writes;
• Applications, including system stuff (explorer.exe) hanging, but not crashing, particularly while downloading/decompressing large files at a high speed.

Your issue might not be memory, but a bad/fragmented drive.

Thanks for the help!

 

[SamirD]

Glad I helped you in the right direction! But I don't think fragmenting alone would have caused that. I would still run the smart tests on the drive as there may be weak sectors that will cause enough of a delay on sector reads to cause this same problem--and then the problem will come back.

 

[LeoF2]

Glad I helped you in the right direction! But I don't think fragmenting alone would have caused that. I would still run the smart tests on the drive as there may be weak sectors that will cause enough of a delay on sector reads to cause this same problem--and then the problem will come back.

I'll be on the lookout. But so far I've had zero crashes or freezes.

I, too, believe fragmentation wouldn't usually cause this... Then again, 41% seems pretty absurd to me. I'd bet the fragmentation in some files/dependancies was so high that it caused unexpected slowdowns to some applications that crashed on the spot. It was like this for browsers and games - explorer.exe never crashed per se, only froze. A few other programs, too, only froze and never crashed. SMART reports a healthy drive with no bad blocks or lost sectors, which is lucky for a 7 year old drive, or the SMART is simply reporting wrong information. Either way, a read/write benchmark reports good values and never reports a single bad read, so I don't think the HDD is dying right now - though it might be and SMART is reporting nonsensical information.

Once more, thanks for the assist. This was a tough one to crack. Cheers!

 

[SamirD]

I'll be on the lookout. But so far I've had zero crashes or freezes.

I, too, believe fragmentation wouldn't usually cause this... Then again, 41% seems pretty absurd to me. I'd bet the fragmentation in some files/dependancies was so high that it caused unexpected slowdowns to some applications that crashed on the spot. It was like this for browsers and games - explorer.exe never crashed per se, only froze. A few other programs, too, only froze and never crashed. SMART reports a healthy drive with no bad blocks or lost sectors, which is lucky for a 7 year old drive, or the SMART is simply reporting wrong information. Either way, a read/write benchmark reports good values and never reports a single bad read, so I don't think the HDD is dying right now - though it might be and SMART is reporting nonsensical information.

Once more, thanks for the assist. This was a tough one to crack. Cheers!

You're welcome! Yeah, it really depends on how long a software will 'wait' for the drive. Back in the day we got to choose that timing (Abort, Retry, Fail?).

I've seen drives even with a good smart result actually have problems, and the best non-destructive way I've found to check these is to boot parted magic off a cd and run the smart long test. If a drive passes that then I've found that it is fine.