Intermittant memory errors. (Fishing for ideas)

[SunnyD]

Cliffs version, since my patience with this is wearing thin.

1. New build, CPU tested working from a friend, mobo NIB from eBay, ram "possibly" NIB from eBay
2. Put shit together, mobo default settings are whacked (automatically overclocking by default when they should be stock)
3. Manually set things to stock settings (as best as possible), and getting intermittant memory errors.

Okay, so now the longer version for #3. Yes... INTERMITTANT. Running memtest86+, the memory will sometimes pass with flying colors, and sometimes fail miserably - but not during the same run. See linked thread for more details.

The kicker - last night I got a new HDD in (larger, and much quieter to appease the wife), so I went to dick around with the machine and reinstall windows on the new HDD. Was getting some intermittent boot issues, apparent no-POST, but it seemed more to do with the monitor than the system after I poked around. Memtest was fine, installed windows without a hitch. Life is good.

Wife starts up the machine about an hour ago and IMs me saying it booted, loaded windows, logged in and BSOD'd. This is a typical MO for what's been going on with this machine now. She didn't catch the BSOD error, but I'm willing to bet that it's MEMORY_EXCEPTION again.

So with that said, given the fact that I can get memtest passing with flying colors at one moment, or failing miserably at another with absolutely no consistency... should I be looking at the memory, the motherboard or the CPU at this point?

Spec: Biostar TH55B HD, i3 530, GSkill DDR3-1333 2x2GB kit.

I'll be popping in a set of DDR3-1066 tonight and see what happens. I'm somewhat frustrated simply because there's no consistency to this issue.

 

[spinejam]

"I'm somewhat frustrated simply because there's no consistency to this issue."

-- sounds like the memory to me. try to loosen the timings instead of defaults (no auto settings).

 

[GlacierFreeze]

If you have a spare PSU, try swapping it. And instead of manually setting mobo settings back to normal, there should be an option to do it automatically. I'd rather do it that way.

 

[SunnyD]

If you have a spare PSU, try swapping it. And instead of manually setting mobo settings back to normal, there should be an option to do it automatically. I'd rather do it that way.

That's the problem. If you look at the link I provided, that is an issue all by itself.

I unfortunately don't have a spare PSU at the moment. That's something that was suggested already as well - which would suck since I'm in a slim case that doesn't use standard ATX PSU's.

 

[rivethead]

It sounds very similar to the "cold boot" problem with G.Skill memory. Check out the G.Skill forums for more details.

I am/was having similar issues. I would boot up, get random blue screen error (always the ntskrnl.exe driver....which is related to the RAM). I would have to turn off the computer and PSU, pull out the RAM, put them back in, power up and the miracle of miracles, the problems would be gone (would pass hours and hours of memtest86 and, more importantly, it would pass the Windows Memory Diagnostic Tool tests on "extended").

I'd use the computer fine all day no problems. However, if I shut down (as I do each night), the next morning the problems returned....random blue screen (ntskrnl.exe driver again), Windows Memory Diagnostic Tool tests would fail, sometimes memtest would pass and sometimes it would fail (at this point I concluded that the Windows Memory Diagnostic Tool was a better test than memtest).

GSkill tech support has alluded in a few posts in the GSkill forum that they know about this problem. Apparently there are settings that should be retained by the RAM even when the RAM is powered down. These are not being retained.

GSkill tech support will tell you to RMA the memory to them.....that's the only fix at this point. I did that yesterday....now I wait and pray I get sticks back that don't have this problem (many, but not all, have reported success on the GSkill forums once they got replacement sticks back).

Good luck.......I'm right there with you on dealing with a wife frustrated because she can't use the computer!

 

[rivethead]

If you have a spare PSU, try swapping it. And instead of manually setting mobo settings back to normal, there should be an option to do it automatically. I'd rather do it that way.

That's my next fix.....if I get the RAM back from GSkill and still have the cold boot problems, I'm going to try a new power supply (mine, while being a good-quality Antec, is old and not efficient).

 

[SunnyD]

It sounds very similar to the "cold boot" problem with G.Skill memory. Check out the G.Skill forums for more details.

I am/was having similar issues. I would boot up, get random blue screen error (always the ntskrnl.exe driver....which is related to the RAM). I would have to turn off the computer and PSU, pull out the RAM, put them back in, power up and the miracle of miracles, the problems would be gone (would pass hours and hours of memtest86 and, more importantly, it would pass the Windows Memory Diagnostic Tool tests on "extended").

I'd use the computer fine all day no problems. However, if I shut down (as I do each night), the next morning the problems returned....random blue screen (ntskrnl.exe driver again), Windows Memory Diagnostic Tool tests would fail, sometimes memtest would pass and sometimes it would fail (at this point I concluded that the Windows Memory Diagnostic Tool was a better test than memtest).

GSkill tech support has alluded in a few posts in the GSkill forum that they know about this problem. Apparently there are settings that should be retained by the RAM even when the RAM is powered down. These are not being retained.

GSkill tech support will tell you to RMA the memory to them.....that's the only fix at this point. I did that yesterday....now I wait and pray I get sticks back that don't have this problem (many, but not all, have reported success on the GSkill forums once they got replacement sticks back).

Good luck.......I'm right there with you on dealing with a wife frustrated because she can't use the computer!

Well, good to know. I'll have to poke around about that. Right now I'm sitting on 8 hours and change of 0-error memtest. I left it going for the day while I'm at work too.

As far as the wife goes, she doesn't really care - she has a laptop to use. It's still frustrating.

 

[n7]

When in doubt, i'd always assume RAM first.

Definitely see what happens with different RAM.

BTW i've seen similar situations w/ bad RAM where it'll sometimes test okay, & other times not.

I'd also really suggest using HCI Memtest alternatively.
I find it's more thorough, particularly for higher end RAM configs where there's more RAM to stress.

 

[SunnyD]

When in doubt, i'd always assume RAM first.

Definitely see what happens with different RAM.

BTW i've seen similar situations w/ bad RAM where it'll sometimes test okay, & other times not.

I'd also really suggest using HCI Memtest alternatively.
I find it's more thorough, particularly for higher end RAM configs where there's more RAM to stress.

Link for the lazy?

Thankfully, I have extra ram laying around. In fact, I have to swap some memory around between my i7 anyway - selling off a bunch, which ironically leaves me with a spare 4GB of memory. So that's happening tonight.

One of the thing that bothers me though is that I'm stuck with at least a portion of reserved memory for the onboard GPU that memtest won't touch. It's moot though, since everything seems almost exactly symptomatic of this "GSkill Cold Boot Issue" anyway. I can't reproduce the error once the machine has been up and running for a while. Just from cold boots after it's been off for ~24 hours at this point. It makes the most sense.

 

[RebateMonger]

A failing capacitor on the motherboard can also cause problems that look like memory problems. If it's a new motherboard, that's unlikely, but it wouldn't hurt to do a one-minute visual scan of all the capacitors for bulging tops.

 

[n7]

http://hcidesign.com/memtest/MemTest.zip

Run as many instances as you have cores, or in your case, threads.

So 4 instances @ ~ 700 MB each.

 

[SunnyD]

A failing capacitor on the motherboard can also cause problems that look like memory problems. If it's a new motherboard, that's unlikely, but it wouldn't hurt to do a one-minute visual scan of all the capacitors for bulging tops.

I did that on install. Brand new board, but I was disappointed it wasn't all solid caps. Everything looked good though. I'll do a recheck tonight when I swap ram out anyway.

http://hcidesign.com/memtest/MemTest.zip

Run as many instances as you have cores, or in your case, threads.

So 4 instances @ ~ 700 MB each.

Thanks. I'll try that out tonight.

 

[SunnyD]

Relayed my story to GSkill as per what rivethead said. They say RMA the stuff. Back it goes.

 

[BTRY B 529th FA BN]

Memtest HCI is the test you should start with. LinX will also help too.

Good Luck

 

[SunnyD]

For what it's worth, this is resolved. I had two sticks of Samsung OEM ram sitting around which I swapped in while the RMA on the GSkill stuff processes. Haven't had a single issue with the system since.