Memtest Results? What Do They Mean?

dragonborn123

Distinguished
Nov 19, 2013
247
0
18,680
So I started a memtest last night, to check the stability of my RAM (been having lots of BSODs that can be caused by faulty memory). Anyway, I check this morning and I see ~1800 errors, but also 6 passes. Does this mean one, or both, of RAM sticks is good? Or is this a sign of dysfunctional memory?
 
Solution
Bad memory. The sticks are just not running at the rated speed and timings, you really need to RMA them or you will never have a stable reliable system. Often bad memory can be made stable by running it at higher voltage or looser timings, but that isn't what you paid for when you bought them.


I figured as much. Thanks for letting me know, I've been having all kinds of weird BSODs and results on my stress tests (couldn't reach max cpu load) which I think have been related to bad memory. I guess I'll have to RMA and wait to use my new pc until after the holidays.

EDIT: Although, couldn't it be a bad memory slot that's causing the issue? Anything I should check before I send the RAM back?
 
Since you can't send it back yet, run memtest on one stick and then the other -- use whichever one works for now. As soon as you see an error quit and try the other stick.

It's not a bad slot -- you would have more errors, if the board would even post.
 
Ok, so I ran the memtest again, just to verify my previous results. Oddly enough, this time I got no errors. Any idea why this would be the case? The only thing I can think of is that the previous results are because the memory overheated (I memtested them right after running Prime95), but I honestly have no clue. Thoughts?
 
Memtest can show errors due to incorrect bios settings for memory, like voltage too low, timings too tight, etc. I have never seen memory suffer from heat related failure when running at rated settings, or even when overclocked at high Vdimm.

If you have no more BSODs and pass memtest though, you're good.
 


Well, I think I do still have issues with BSODs (UNEXPECTED KERNEL MODE TRAP and SYSTEM SERVICE EXCEPTION have each appeared twice in the last four days). Apparently faulty memory can be the cause of both of these. So I'm just trying to figure out if the RAM errors I experienced were a false alarm, or actually indicative of some kind of problem. Because otherwise I need to start searching for other possible causes for the BSODs.
 
One problem that I've run into a number of times is that if you do an OS install with bad memory (or some memory issue) then you really need to do a new clean install to fix all the errors that the bad memory caused during the install. Run memtest overnight for Santa, and if all is good you can do a clean install tomorrow.

My build routine for the last few years (about 100 builds) is to run memtest right after finishing the build and before the OS install. It's saved me on 4 builds that had memory problems, a couple were bios settings where the SPD incorrectly set timings and a couple that were just bad memory.
 


I guess it wouldn't hurt to do a fresh install, but how would that help determine if my memory is faulty? If the RAM is bad, wouldn't I just be making another bad OS install? Is there any fool-proof way to figure out the condition of my RAM?

Also, regarding the BIOS, what exactly should I check to make sure the settings are correct?
 
It is not worth doing a clean install unless you can get comfortable with the memory being intact, hence running memtest again for a prolonged period first. If the memory testing comes up with any errors don't waste time with another install until you get a new set.

Memtest pretty much is the "fool proof" way to test your memory, and I remain suspicious until you get 5 or 6 passes with zero errors. Take a look at your memory related settings in the bios to check that they agree with the detailed timing information from the memory manufacturer for that set.
 


Ok, thanks. I've had memtest on for the last five hours, four passes. Should I just keep it running all night? If the test comes back clean, what should I assume caused memtest to register those errors originally?
 


Ok, I tested all night, finished with 12 passes and zero errors. However, following your previous recmmendation, I checked the BIOS settings to see if they were properly set. Turns out, I don't think they were. The frequency and timings were slightly off, but I wasn't able to edit them. Here's my build and some photos of my BIOS, so you can see what I mean, and maybe advise me on how to correct this:

http://pcpartpicker.com/p/2m2xf
http://s1257.photobucket.com/user/tacallaway/library/?view=recent&page=1
20131225_125650.jpg.html

20131225_125630.jpg.html

20131225_125600.jpg.html
 


Ok, I adjusted the timings and frequency automatically with XMP, and that fixed all the settings. However, I just ran another memtest to verify everything was functioning properly, and unfortunately I got about 4k errors. Any idea what could cause these inconsistent results?
 
Bad memory. The sticks are just not running at the rated speed and timings, you really need to RMA them or you will never have a stable reliable system. Often bad memory can be made stable by running it at higher voltage or looser timings, but that isn't what you paid for when you bought them.
 
Solution


Ok, I guess I'll do that. But it's a funny thing, I kept memtest running all night, and I got 12 passes and no more errors. Really odd.

I do have a couple of old sticks of RAM lying around, I guess it wouldn't hurt to give them a try (DDR3 1066). Hopefully they come back clean.