Strange RAM issue (Chronic)

[BarkingGhostar]

Three years ago I built a W7 PC for the wife. A month into it the RAM tested bad (4x2GB OCZ costing $330 at the time). I identified the bad pair and removed them using the Windows 7 memory test. System seemed to be more stable, but since this wasn't my PC and my wife tends to 'live with it' conditions, this got ignored.

Then about six months or so ago I replaced all of the OCZ RAM with Corsair XMS (4x4GB CMX16GX3M4A1600C9). The problems of stability and BSOD didn't go away. heck, I even replaced the hard drives, stopped using the graphics on the Core processor (employing a standalone), terminated the RAID to standalone, etc.

Then last week I happen to be running Office Diagnostics and it reported bad memory (RAM). I then ran the W7 memory test with the same bad results. But this morning I happen to be testing it via a USB flash boot into memtest86+ and so far it is finishing up its second pass and no errors.

BTW, this is supposed to be 1600MHz, 9-9-9-24, 1.5V etc., but the Gigabyte BIOS is showing it configured as 1333 with 8-8-8-24 timings. I have read plenty of folks in several forums over the past year claiming that this RAM is not being read 'as advertised', but I wouldn't imagine it being a problem running slower.

Thoughts? If the RAM is good I plan on attempting to recycle it in a new build. I am buying a new CPU and motherboard, case and PSU, etc. tomorrow for a dual-boot W7/OSX (yes, Hackintosh). I just wonder why the H55-USB3 board would employ this memory in such a manner. Is it the mobo or the RAM?

 

[BarkingGhostar]

After 4.5 hours of running memtest86+ without errors I booted into the BIOS and manually set the RAM to 1600 9-9-9-24. Boots faster but will she be stable?

 

[Insert_Nickname]

It could sound like a wrongly configured sub-timing. That can produce random/non-consistent errors like you describe. Some board/RAM configurations also have trouble running with the CR set to 1. Might want to check that first.

The best thing I can come up with, is to use a program that can read the SPD timing/subtimings (Aida64/HWinfo or similar), write them down, then go into the BIOS and make sure that ALL timings (main/sub/CR) are set as specified by the SPD.

 

[KingFatty]

Just to confirm, the SPD timing/subtimings include all those many many settings in the BIOS under memory timings/advanced? I'm having similar issues as OP, which arise when I flashed to a more recent BIOS, where it can't do "auto" reliably and I need to manually configure. I will give Aida64 a try if I can't figure out how to access this level of super-detail for the SPD timings using HWinfo that is already installed, but any specific tips are welcome.

As for OP, I'd suggest you distrust any "auto" settings, even if you have the latest BIOS for your motherboard installed (but make sure you install the latest anyway).

 

[Insert_Nickname]

Just to confirm, the SPD timing/subtimings include all those many many settings in the BIOS under memory timings/advanced?

Yes. Some BIOS just can't reliably auto configure timings. But SPD has never been that reliable to begin with. Sometimes its just better to manually do it yourself. Or at least check the auto configuration.

Aida64 is pretty reliable for exposing sub timings BTW.

Motherboard->Chipset and SPD is where you should look. First is for your current config, second is the SPD/XMS of the installed RAM.

 

[BarkingGhostar]

OK, so after the changes to the timings and frequency the Windows 7 memory test run under Normal mode is not finding any errors, but Office Diagnostics is still reporting them.

When I try to run Windows 7 memory test under Extended mode something strange occurs. It gets to a point of 21% in the first pass and never seems to go beyond it. I've a couple of hours I gave up and ran the Normal test. I know there are some additional tests in the Extended, and I'm guessing this is where Office Diagnostics is detecting the issue.

Still, I have no idea why memtest86+ didn't find anything. Also, I have no idea what the sub-timings need to be for this memory. I guess I need to do some serious research on the matter.

 

[BarkingGhostar]

It could sound like a wrongly configured sub-timing. That can produce random/non-consistent errors like you describe. Some board/RAM configurations also have trouble running with the CR set to 1. Might want to check that first.

It was set to AUTO, with a reported automatic setting of 2. I changed it to 1. When I read this and it suggested 2 was for stability (1 being normal in the Intel world when the number of banks per channel are limited to 4.

I'm not sure if maybe this is a board level voltage regulation issue, and if it was how to conduct some sort of monitoring.

 

[bryanl]

BTW, this is supposed to be 1600MHz, 9-9-9-24, 1.5V etc.,

It's supposed to be but isn't, considering the chips are probably rated 11-11-11-28 or slower, perhaps just 1333 MHz.