4 Memory Sticks Installed = Errors; Individually Tested = Zero Errors

[Route44]

Last night our second family PC Blue Screened with an 0x4E PFN_LIST_CORRUPT error which is a very strong indicator of corrupted memory.

Memory is Crucial Ballistix PC3200 (DDR 400) in a 4x512 (2 gigs) configuration in a socket 939 system Epox 9NPA+Ultra motherboard.

The specs on this memory are as follows: 2.8 voltage which the motherboard has it set at correctly as well as in 2T. The CAS is a designed 2-2-2-6 but mobo has it set at 3-4-4-8 which is probably due to the 4x512 configuration and 2T

Ran memtest and into the fifth PASS received 212 errors all in Test 5. I saw enough so I tested each stick individually for the minimum Passes of 7 and no errors on any of them even in test 5.

So I reinstalled all four sticks, checked the BIOS settings and ran Memtest again and in the very first PASS I received 20 errors in Test 5.

Any ideas?

 

[red1776]

maybe i misread your testing method, but could it be a bad dimm?

 

[Route44]

Hey red, thanks for looking in. I ran all four sticks together and the errors occured. I ran each stick individually with memtest and no errors whatsoever but I did only test in slot one.

I have the painful feeling I'm going to need to test in the other slots. If one is going what, from your experience, is the chances the motherboard itself may be going as a whole?

By the way, two Passes with all four sticks and 141 errors in test 5. 20 in the first Pass 121 in the second Pass.

 

[Tedster]

memtest rarely is wrong. If you are getting errors, then the ram is bad.

 

[captaincranky]

Hey red, thanks for looking in. I ran all four sticks together and the errors occured. I ran each stick individually with memtest and no errors whatsoever but I did only test in slot one.

I have the painful feeling I'm going to need to test in the other slots. If one is going what, from your experience, is the chances the motherboard itself may be going as a whole?

By the way, two Passes with all four sticks and 141 errors in test 5. 20 in the first Pass 121 in the second Pass.

You do realize that in all cases slot one needs to be populated. So the minimum number of DIMMs installed for testing would be 2, for sockets 2, 3, & 4. Running with Channel "A", socket "0" empty could possibly damage the board. That's my story, and I'm stickin' to it. The symptoms you're describing could also be a defective northbridge (MCH) (!) or (?)

I am somewhat suspicious that good memory doesn't test bad simply because it's placed in a different socket. At least not if the voltage is correct.

 

[red1776]

Hey Route,
the reason i brought it up is that while it most likley a module, I have seen a dimms short out from debris in them. I would blow them out and inspect closely for burn marks along the contacts (pins). its been known to happen,more commeonly in DDR (1) boards, and your results were irregular when testing in one slot only.

 

[captaincranky]

OK, I concede that socket corrosion is also a distinct possibility.

 

[Route44]

I did blow out the slots with air and slot one has always been populated. I am even getting more errors with second time through with it in 5x512 than the first time through.

 

[dividebyzero]

If you've never populated all fout DIMMs before then there's a real possibility that the board has stability problems re: providing enough voltage through the MCH for four modules.
Gigabyte P45 boards are somewhat notorious with this also, as are some Abit boards.

Edit: You might want to try raising the MCH (northbridge) voltage a notch or two and retest. If the MCH is having trouble maintaining voltage across four DIMM's then Memtest86 is likely to throw errors on the two most intensive tests (tests 2 and 5). Contrary to some supposed experts view, just because Memtest throws up errors it doesn't automatically point to a bad module. Crucial are lifetime warranty in any case-if you don't mind potentially getting the same modules back and paying the postal service to schlep them around the country, theres always RMA.

 

[Route44]

For over a year and a-half all four DIMMs have been populated without a hitch.

I also realize that it could be due to other issues than bad modules.

Pardon my ignorance but where would I find the MCH in the BIOS to raise the voltage a notch or two? Thanks.

 

[dividebyzero]

Sorry, haven't worked with an nVidia board for a while- MCH would be the chipset voltage (nForce4 ?).
I assumed that since the module in the first DIMM was passing that that the problem would likely be in DIMM2 and 3. Longer memory traces are sometimes more susceptible to voltage drop, the thinking being that raising the chipset voltage by 0.1 or 0.2 v might produce a more stable Memtest result, which in turn might point to a failing power phase/capacitor.

 

[Route44]

Well it's my memory.

One stick failed Test 3 and another stick failed Test 5. The problem is that one of the sticks came in a set which means I am going to have to RMA the set plus the one which leaves me with the blistering speed of 512 megs of RAM. :blackeye:

Thanks to all of you for the help.

Update: Contacted Crucial and they'll exchange all four of my 512 sticks for four 1GB sticks because there is no way to replace my Ballistix. I have always found Crucial's customer service A+ and this time was no exception. I love their RAM. Used it for many upgrades but their Ballistix line needs to be avoided like the plague.

 

[dividebyzero]

...but their Ballistix line needs to be avoided like the plague.

In my experience The DDR3 modules are pretty average and a bit overpriced, also sub-par on the overclocking.
I have used Ballistix and B.Tracers in many of my builds. With regards he original Ballistix DDR2-800/-1066 kits, the first revision were (and continue to be) fantastic modules- they are characterised by being dual sided 2.0v modules with the much sought after D9GMH chips (suffix .16FD3 (2Gb).8FD3(1Gb) ). During the production run of this (and the lower rated series) Crucial changed production to a less costly D9 chip (suffix 16FE5 (2Gb), .8FE5 (1Gb) ) at 2.0v (early series) and 2.2v (current series), these modules have a somewhat variable (to put it kindly) quality control.
Just as Corsair RAM has a huge variance depending upon revision* so do Crucial and it pays to reseach just exactly what you're buying.

(* Corsair RAM versions use chips sourced from Nanya, Micron, PQI, Elpida, Qimonda, Infineon, Winbond, Samsung, Hynix, ProMOS amongst others)