How to test RAM when Memtest and Windows memory test don't work?

[anon5123]

So I'm trying to diagnose the BSODs I've been getting the past two days, with most of the error messages pointing towards it being a RAM issue. I first tried the built-in Windows Memory Diagnostic thing, and that ran for about an hour and a half, getting all the way up to 99%.....and then crashing to a screen that told me that my Windows installation needs to be repaired. Even stranger is that IMMEDIATELY when the test began, it said it detected a hardware error.

I tried Memtest, and while it booted fine through USB, as soon as the test started, it just exploded with thousands and thousands of errors. Like seriously it racked up 10,000+ errors in just a few seconds. So that's obviously not working either.

Also tried Memtest+ and...it won't even boot.

So what other memory test programs can I use? Googling for "windows RAM test" every site just keeps suggesting Memtest.

It was a bad RAM stick.

[mr120]

Is the next step not to try different combinations; use different memory slots, 1 stick, etc and if those all fail, try new memory modules?
Rather than to say the software saying the ram is faulty must be wrong?

[Temp name]

Memtest works. If you're getting thousands of errors, your RAM is majorly fucked. I don't understand how you can try 2 different well respected memory testing programs, have both spit out a shit ton of errors, and assume they're both broken instead of your hardware.

Find out if it's just 1 stick, or both, or the slots.

[anon5123]

Memtest works. If you're getting thousands of errors, your RAM is majorly fucked.

Thousands of errors in A FEW SECONDS. I didn't even let it run very long, it was running for only like 20 seconds and had already gotten up over 15,000 errors found. Does that sound like a valid result, or does it sound like the test is glitching out?

I mean, which is more likely: The RAM test is glitched, or "majorly fucked" RAM is somehow working perfectly fine outside of a random BSOD once or twice per day? Think, man. Googling around a bit, it seems other people have gotten similar results, one guy getting "70,000" errors on his test, and another getting "over one million" after letting it run for a couple days, but after testing one stick at a time it seemed to work. I guess that's what I need to do, just test one stick at a time?

I did read that Memtest86 doesn't work on some newer computers, hence why I tried Memtest86+ which is apparently newer and updated, but that one wouldn't even boot at all. I'd just get "Operating system not found".

Find out if it's just 1 stick, or both, or the slots.

I kind of need to run a memory test to find that out. But all of the ones that I try aren't working.

[Temp name]

Thousands of errors in A FEW SECONDS.

Because your RAM is majorly fucked.

I didn't even let it run very long, it was running for only like 20 seconds and had already gotten up over 15,000 errors found. Does that sound like a valid result, or does it sound like the test is glitching out?

If you have fucked up RAM, it is a valid result.

I mean, which is more likely: The RAM test is glitched, or "majorly fucked" RAM is somehow working perfectly fine outside of a random BSOD once or twice per day? Think, man. Googling around a bit, it seems other people have gotten similar results, one guy getting "70,000" errors on his test, and another getting "over one million" after letting it run for a couple days, but after testing one stick at a time it seemed to work. I guess that's what I need to do, just test one stick at a time?

RAM leading to crashes is incredibly unlikely. It needs to corrupt the data in a way that is unrecoverable, which is rare. It's more likely that all the errors are just seriously hurting your performance. Especially considering you yourself said that your BSOD messages are related to memory.

I kind of need to run a memory test to find that out. But all of the ones that I try aren't working.

Try HCI memtest
you'll likely get the same results though.

[zhero]

I kind of need to run a memory test to find that out. But all of the ones that I try aren't working.

why? you just pull the sticks of ram out, try it in different combinations and then try starting it up

[Ringthane]

Thousands of errors in A FEW SECONDS. I didn't even let it run very long, it was running for only like 20 seconds and had already gotten up over 15,000 errors found. Does that sound like a valid result, or does it sound like the test is glitching out?

I mean, which is more likely: The RAM test is glitched, or "majorly fucked" RAM is somehow working perfectly fine outside of a random BSOD once or twice per day? Think, man. Googling around a bit, it seems other people have gotten similar results, one guy getting "70,000" errors on his test, and another getting "over one million" after letting it run for a couple days, but after testing one stick at a time it seemed to work. I guess that's what I need to do, just test one stick at a time?

I did read that Memtest86 doesn't work on some newer computers, hence why I tried Memtest86+ which is apparently newer and updated, but that one wouldn't even boot at all. I'd just get "Operating system not found".


I kind of need to run a memory test to find that out. But all of the ones that I try aren't working.

No, the ones you try ARE working. If Memtest finds thousands of errors with your memory, your memory is bad. If your RAM was working properly, Memtest would find nothing, no matter if you let it run for five seconds, five years, or five decades. And if you're getting blue screens related to RAM errors, your RAM is bad. I've been running the same memory for four or five years for hours and hours per day. Everything from streaming Netflix to playing games, and I can't remember the last time my computer crashed.

Trust us on this - if your computer is crashing with blue screens related to memory, and Memtest is finding thousands of errors, your memory is bad (or the slot on the motherboard is bad). Your next step is to try them, one stick at a time, to see if it's the slot or the RAM itself, and go from there.

[anon5123]

Because your RAM is majorly fucked.

Why does it work just fine then, if it is "majorly fucked" ?

I've been running this RAM for about 10 days and have only had two bluescreens, one of which was actually caused by a program and not randomly. That doesn't sound "majorly fucked" to me.

What DOES sound "majorly fucked" is a memory test apparently detecting 15,000+ errors in the span of 20 seconds. That doesn't seem right in the slightest.

Try HCI memtest
you'll likely get the same results though.

The one that runs in Windows, yeah? That one only lets me test 2gb of ram at a time, any more and it gives an error saying that Windows limits the amount of RAM it can test.

I opened 4 instances of that, testing 2GB each, and let them run for about an hour; 0 errors found.

- - - Updated - - -

No, the ones you try ARE working. If Memtest finds thousands of errors with your memory, your memory is bad. If your RAM was working properly, Memtest would find nothing, no matter if you let it run for five seconds, five years, or five decades. And if you're getting blue screens related to RAM errors, your RAM is bad. I've been running the same memory for four or five years for hours and hours per day. Everything from streaming Netflix to playing games, and I can't remember the last time my computer crashed.

Trust us on this - if your computer is crashing with blue screens related to memory, and Memtest is finding thousands of errors, your memory is bad (or the slot on the motherboard is bad). Your next step is to try them, one stick at a time, to see if it's the slot or the RAM itself, and go from there.

I'm not saying that my RAM isn't bad, I'm saying that Memtest finding 15,000+ errors in 20 seconds looks very much like it is glitching/not working. Google searching this found that other people were having similar issues, even one guy saying Memtest reported over a million errors on his RAM, but then when he tried it one stick at a time, there were no errors.

[Twoddle]

Try Prime95.

[anon5123]

Try Prime95.

I read that Prime95 is more for stress testing, not memory error testing.

- - - Updated - - -

Okay, testing out Memtest86 again.

First attempt, all 4 sticks: It gets about 2 seconds in and then just starts throwing endless errors. Stopped around 27,000 errors and said it aborted due to too many errors.
Second attempt, just 1 stick: Same result. Threw so many errors that it aborted due to too many errors.
Third attempt, 1 stick, different stick, same slot: ...and it actually decides to work now. It’s been running for about 10 minutes, 75% on pass 1, and has had no errors thus far.

So maybe it’s just one of my RAM sticks that is fucky.

On a similar note, is there a chance that my RAM is incompatible with my mobo in some way? I used PCPartsPicker which usually ensures that the parts you choose are compatible with each other. Mobo is an MSI Z490-A, and the RAM is Ripjaws DDR4 (4 x 8 GB), says it’s 2400mhz or something like that.

https://www.newegg.com/msi-z490-a-pr...82E16813144306
https://www.newegg.com/g-skill-32gb-...82E16820231889

[Twoddle]

I read that Prime95 is more for stress testing, not memory error testing.

Yeah it's a stressing tool but you could still use it to positively identify a fault in that if it gives errors you can suspect faulty RAM or CPU but not the other way round, ie. if you don't get errors it doesn't necessarily mean a dedicated RAM test won't find anything wrong.

[anon5123]

Okay, Memtest completed a full pass on the one RAM stick, 0 errors found. I know that one pass isn’t completely conclusive, but it’s a lot better than “27,000 errors, aborted due to too many errors”.

Went ahead and tested the remaining two sticks, all in the same slot, tested them for about 5 minutes, and they also didn’t throw a million errors, so for the moment I can assume they’re fine.

Just to be entirely sure, I put the “bad” RAM stick back in and...yeah, immediate bajillion errors in Memtest86, so I guess this one stick is garbo and I’m gonna RMA.

[zhero]

On a similar note, is there a chance that my RAM is incompatible with my mobo in some way? I used PCPartsPicker which usually ensures that the parts you choose are compatible with each other. Mobo is an MSI Z490-A, and the RAM is Ripjaws DDR4 (4 x 8 GB), says it’s 2400mhz or something like that.

some times electronics are at the bottom of the bin and barely function long enough for you to have no problems for a few days.

[pansertjald]

I read that Prime95 is more for stress testing, not memory error testing.

- - - Updated - - -

Okay, testing out Memtest86 again.

First attempt, all 4 sticks: It gets about 2 seconds in and then just starts throwing endless errors. Stopped around 27,000 errors and said it aborted due to too many errors.
Second attempt, just 1 stick: Same result. Threw so many errors that it aborted due to too many errors.
Third attempt, 1 stick, different stick, same slot: ...and it actually decides to work now. It’s been running for about 10 minutes, 75% on pass 1, and has had no errors thus far.

So maybe it’s just one of my RAM sticks that is fucky.

On a similar note, is there a chance that my RAM is incompatible with my mobo in some way? I used PCPartsPicker which usually ensures that the parts you choose are compatible with each other. Mobo is an MSI Z490-A, and the RAM is Ripjaws DDR4 (4 x 8 GB), says it’s 2400mhz or something like that.

What BIOS version are you running whit? Version 7C75v23 has Improved memory compatibility

https://www.msi.com/Motherboard/support/Z490-A-PRO

[anon5123]

some times electronics are at the bottom of the bin and barely function long enough for you to have no problems for a few days.

I had no problems for an entire week. I built this new setup on the 17th, and the BSODs didn't start happening until the 25th.

Even so, I don't think Ripjaws is "bottom of the bin", considering all the good reviews it has on Newegg. Likely just another freak case of 1 RAM stick out of the bunch being bad.

What BIOS version are you running whit? Version 7C75v23 has Improved memory compatibility

https://www.msi.com/Motherboard/support/Z490-A-PRO

I didn't check while I was in the BIOS, but I'm thinking that the problem here is the one bad RAM stick, now that I've run Memtest on every individual stick.

- - - Updated - - -

Furthermore, looking at the order of the DIMM slots on this mobo, the bad RAM stick was in the 4th slot, so I guess it makes sense that it just never got touched unless Windows went over a certain amount of memory. Or something like that, I have no idea how Windows actually manages memory.

[zhero]

Likely just another freak case of 1 RAM stick out of the bunch being bad.

thats what bottom of the bin silicon is.

[C20]

Just in case anyone else finds this thread looking for answers, I've added some more info below on why you shouldn't overlook what might seem like an unrelated BSOD, and how memory in Windows works (and why RAM test tools aren't throwing you errors unless an actual error happened).

Furthermore, looking at the order of the DIMM slots on this mobo, the bad RAM stick was in the 4th slot, so I guess it makes sense that it just never got touched unless Windows went over a certain amount of memory. Or something like that, I have no idea how Windows actually manages memory.

Honestly, if it's a single bad stick, RMA it. Bad hardware does exist, and sometime it makes it past the QC and ends up in the retail channels. This is why retailers have RMA processes in place.

If it's happening in a single DIMM slot with each stick (or more than one stick, at least), then that's a sign of either an incompatibility or stability issue, or a motherboard issue. This might be fixable with a BIOS update, but there's no guarantees. With that said, the only time I'll update my BIOS is if it's a memory compatibility fix and I'm having issues, or a CPU microcode update to support newer chips.

Windows sees memory/RAM as one big storage area. It doesn't do any splitting of data over different sticks - the memory controller on the board or in the CPU decides how best to allocate it. This is simply due to the wiring of each architecture being different, and the manufacturer's knowing how best to allocate and split the load. On some boards, this will allocate it evenly across each stick, and in others, it'll progressively fill up one stick before it starts to fill another. We'd have to look into the technical docs for a specific controller to know for sure - ultimately though, that's not going to fix the problem.

Why does it work just fine then, if it is "majorly fucked" ?

I've been running this RAM for about 10 days and have only had two bluescreens, one of which was actually caused by a program and not randomly. That doesn't sound "majorly fucked" to me.

What DOES sound "majorly fucked" is a memory test apparently detecting 15,000+ errors in the span of 20 seconds. That doesn't seem right in the slightest.

How do you think these programs work? Do you think they have some logic in them that randomly decides if you're going to get an error or not?

That's not how they work. They exist to diagnose problems. If a problem is detected, they tell you. That's all there is to it. In an ideal world, you should have zero errors with memory testing. A few errors generally isn't a sign of an immediate problem, but it's something to keep in mind.

A program causing a blue screen is not normal, and that's a sign of it either being built to intentionally cause them, or there was an error caused by a hardware or software fault. A software fault usually won't trigger a blue screen unless that software is a driver or it's trying to access privileged instructions or resources which it doesn't have permission to (e.g. areas of memory that Windows protects, kernel-space memory, etc). Drivers are the worst culprit for this because they integrate more deeply with the OS than regular software.

Drive issues can also throw up some very weird problems. For example, I kept getting the odd BSOD for such a long time that didn't immediately point towards a drive problem. As I upgraded my system, the only things I hadn't replaced was the drives but they still happened. Upgraded the drives in due course, and the issues seemed far less prevalent.

[anon5123]

If it's happening in a single DIMM slot with each stick (or more than one stick, at least)

Nope, I tested each stick individually in the same DIMM slot, only one of them spazzed out with errors, the other 3 were fine.

Do you think they have some logic in them that randomly decides if you're going to get an error or not?

I never said that.

I had just read that Memtest86 does not work on "some newer PCs", so I assumed it was glitching out in some way.

[JustaWarlock]

Why does it work just fine then

had two bluescreens

detecting 15,000+ errors in the span of 20 seconds.

Not sure if trolling or just dumb?

Infracted

[chazus]

Even so, I don't think Ripjaws is "bottom of the bin", considering all the good reviews it has on Newegg. Likely just another freak case of 1 RAM stick out of the bunch being bad.

I mean, it's sort of accurate. All ram, regardless of maker, is going to have less than like, 1 in 1,000,000 bad. If you buy 900,000 sticks of $30 ram, and 900,000 sticks of $60 ram.. and they're all 100% good, which do you think is better? Which do you think people will buy more, and review as 'great'. The $30 will have a a ton of great reviews, and the $60 one will have significantly less. That doesn't mean it's less reliable, or the cheap one is more reliable. It just means that ram has a very low failure rate, and higher purchase rate.

Why does it work just fine then, if it is "majorly fucked" ?

I've been running this RAM for about 10 days and have only had two bluescreens, one of which was actually caused by a program and not randomly. That doesn't sound "majorly fucked" to me.

What DOES sound "majorly fucked" is a memory test apparently detecting 15,000+ errors in the span of 20 seconds. That doesn't seem right in the slightest.

That's sort of how it works. 2 BSOD in 10 days is 1000% more than what I would expect from 'normal operation'. Windows doens't hit ram like an actual test does. Just like 'being achy in the morning' isn't nearly as conclusive a test as a doctor checking you out in the office. Ram testers do specifically non-organic, non-standard methods in order to root out rarely used methods to determine RAM is bad. That's why I often use a windows install as a 'litmus' test on a HDD to see if its good. I've had plenty of drives pass various tests, and yet windows wont install. Bad drive.

I had just read that Memtest86 does not work on "some newer PCs", so I assumed it was glitching out in some way.

One of them wasn't updated for a while, but I believe they're both compatible with just about everything now.

That said, a clean bill of health from Memtest of any version doesn't mean it's good. But seeing errors almost definitely means something is bad.