How to fix memory management bluescreens.
March 4, 2011 8:15 PM   Subscribe

How do I identify exactly what my memory error bluescreen in Win Vista is.

I have been having bluescreen of death errors quoting MEMORY_MANAGEMENT.

However, I have replaced both hard drives in my RAID array and reinstalled windows (Vista) twice now.

I have run windows memory diagnostics as well as Dell diagnostics and both identify a memory problem but neither tell me exactly what is wrong.

I have 4GB of memory in the system - 4 x 1GB DDR RAM DIMMS. I presume it is one of these at fault, but how do I know which - is it likely to mean that I can replace one of the DIMMs, replacing all 4 would be very costly! Any advice appreciated.

System details:

The set up is a Dell XPS 420, Intel Core 2 Quad CPU ~ 2.4 GHz, running Vista Home Premium SP2
posted by inbetweener to Computers & Internet (13 answers total) 1 user marked this as a favorite
 
How long did you run the memory diagnostics for? Sometimes memory errors do not manifest themselves on only one or two complete cycles of the test. I suggest downloading MemTest86 and testing your 4GB for a good long time -- Perhaps start it when you go to bed one night, and then let it continue to run when you go to work the next day.
posted by Jinkeez at 8:34 PM on March 4, 2011


...Er, and to address your other questions, once I am certain that a stick is bad, I would pull half of them and test the other half for a similarly long time. If no errors, set the RAM aside in the "good" pile and test one of the remaining two -- if that one tests successfully too, your remaining stick is the bad one. If you get errors, be sure to test that one remaining stick anyway in case there are more than one bad stick.
posted by Jinkeez at 8:38 PM on March 4, 2011


In the past, I've used memtest86 for memory testing. It's a standalone tool that you burn to a bootable CD, so that it can run without an operating system getting in the way and preventing it from testing all your memory. memtest86 will report which address is failing, but that's not actually very useful when it comes to tracking down which module is faulty.

So what I would do is a simple process of elimination: Remove all the memory sticks except for one, and run the test on that. Repeat until you find the faulty stick.

Alternatively, if you're running the tests from within windows, and 1 GB isn't enough to boot, you could remove one stick at a time until the tests pass, then you know the stick you removed is faulty.

And once you've found the faulty stick, you can replace it. Or not: If 3 GB is enough for your needs, you could just run with three sticks of RAM. There might be a very small performance degradation, as DDR2 memory (which is probably what you have, not DDR) likes to run in identical pairs, so it can run in so-called "dual-channel" mode, but I doubt you could notice the slight difference in actual practice.

If you get results that seem odd (for example if every individual memory stick appears to be bad, which is a bit unlikely) it could also be a problem with the motherboard or CPU, but by far most common cause of memory corruption errors is (you guessed it) bad memory.
posted by jcreigh at 8:40 PM on March 4, 2011


I should second Jinkeez's comment about the need for prolonged testing to be sure. In my experience, bad RAM falls into two categories: Either it's "reliably" bad, and it will show up as bad within minutes of testing, or else it's only intermittently bad, and it might take hours for it to show up in testing. And hence long testing is required to be sure that you're not dealing with the second category.
posted by jcreigh at 8:48 PM on March 4, 2011


There's one diagnostic that seems to be missing from the above advice (of which is excellent). That is a problem with the motherboard or memory slot. If you find what you think is a faulty stick by doing memtest86, place a known or likely good stick in the memory slot where a faulty stick is found, then retest. If it is a faulty stick, the known good stick should test out. If it's a faulty memory slot, the known good stick should fail.

If it winds up being a bad memory slot, you would need to either replace the motherboard or leave that slot empty (empty slot is the cheap solution, and the one I personally would do).

Replacement DDR RAM is going for around $30/GB on newegg.
posted by Mister Fabulous at 10:53 PM on March 4, 2011


Sometimes the fault in a RAM stick that tests bad is as simple as contamination on the edge connector. If you find a faulty stick, look at the gold edge connector under a strong light, preferably with a magnifier. If you see anything even vaguely resembling a thumbprint, rub the edge connector over with a pencil eraser (don't use an ink eraser - too abrasive), then blow all the little rubber crumbs off with compressed air or no-residue contact cleaner, then test the stick again with memtest86. This method has saved me needlessly discarding several sticks. Works well on expansion-card edge connectors too.

If you find a bad slot, as opposed to a bad card: once again, the fault may be as simple as gunk lodged in the contacts, and a good wash out with a no-residue contact cleaner spray might well resurrect it.
posted by flabdablet at 4:55 AM on March 5, 2011


When you clean the edges of a card be SURE to do it up/down not across. You don't want to rub across and potentially break loose several circuit traces. Also look down into the sockets and make sure there's no debris in them. A strong flashlight helps here. If you see anything start by using a can of compressed air to blast it clear. Trying to straighten any bent connectors is rather difficult but possible with a steady hand. A dentist's pick and/or strong fine-point tweezers are handy tools for this. But it's definitely not an easy task.

I also suggest using a memtest boot CD. Since it runs linux it'll work with only 1GB of memory (or even less). Try running it with each stick individually and then in pairs. Some motherboards will allow using individual sticks in whatever slots you prefer, while others are particular about which slots have to be used in order or pairs. Can't say whether that Dell would require that or not. Your problems may be the memory stick(s) or slot(s). Swapping things around and testing is pretty much the only way to do it.
posted by wkearney99 at 7:04 AM on March 5, 2011


I've eraser-cleaned a lot of edge connectors, I have always rubbed them along the connector (which I'm guessing is "across") and I've never broken a circuit trace. You want to be rubbing the gold-plated connector pads, not anything else, and if the circuit traces on a card are so crappily bonded that a gently deployed eraser can take them off, the card is not of sufficient quality to be worth cleaning anyway.

The major hazard with doing it across-ways is increased eraser wear and rubber crumb production, but then, you really need to clean the crumbs off thoroughly regardless of rubbing technique. Holding the stick over the mobo while cleaning it is an error :-)

Memtest86 is a completely self-standing 120K executable. It doesn't need an OS, just some form of boot loader. It will indeed run on machines with less than 1GB RAM - much less; doesn't even need 1MB.
posted by flabdablet at 8:04 AM on March 5, 2011


Memtest86, and/or do the diagnostician thing. Pull all sticks, and one at a time put them in slot 1 until you get the error. If no error, you know it's a bad slot and not a bad stick.

Now, put 1 stick in slot 1, and then try all sticks in slot 2. Rinse/repeat until you identify the bad stick or the bad slot.

Bad slot...sucks, could be indicative that board will begin to die soon, but use it for now.

Bad stick...replace it!

Move on.

Sidebar---to determine if your board needs parity (multiples of 2, ergo 2 sticks or 4, not 3 or 1), visit crucial.com and do their memory scan (if the system will run long enough, else just type in make/model unless it's homebrew.) It will say "needs to run in pairs" or not.
posted by TomMelee at 10:14 AM on March 5, 2011


Sidebar: that's not what "parity" generally refers to when talking about RAM. Parity-checked RAM has extra chips that cross-check the RAM data content and allow single-bit storage errors to be detected on memory reads. Each such stick has its own parity chips. Parity-checked RAM is rare on desktop computers - it's generally used in servers that need more reliability.

Whether or not RAM sticks should be paired, so enabling them to run in dual-channel mode, is a separate issue.

You may well encounter RAM sticks that work fine in single-channel mode but fail when run as half of a dual-channel pair. I know of no faster way to detect these than systematically working through all possible pairings.
posted by flabdablet at 5:46 PM on March 5, 2011


flabdablet's right, sorry, typed faster than my brain worked.
posted by TomMelee at 7:06 PM on March 5, 2011


Thank you to everyone who answered this question. The simple logic of just testing each slot and card worked. All the advice was really helpful.

Thank you.
posted by inbetweener at 10:14 AM on March 28, 2011


So, was it a bad slot, a bad card, or just a dirty edge connector?
posted by flabdablet at 6:20 PM on March 28, 2011


« Older vomit after eating bread   |   Which Movies Are They? Newer »
This thread is closed to new comments.