Strange RAM issues

First a little background.

The computer that this refers to has been my main production computer for the past 10 months with no hardware issues. It has run solidly with Vista SP1 x64, Vista SP2 64. For production, I have replaced it with another machine and this has become my test production machine.

Machine specs:

Gigabyte GA-EP45-UD3P motherboard
Q6600 Rev GO
Corsair 4G (2x2G) DDR2 1066 5-5-5-18 @2.1v
Corsair HX520 PS
EVGA 9600GT

All standard bios settings except as required to get RAM to run at rated speed and clocks (manually changed multiplier, and voltage - bios set timings). Otherwise no overclock.

Installed and ran 7100 x64 and then 7201 x64. Clean installs. Both ran without problems (except for a few bugs).

I then did a clean install of 7229 x64. Installed all my apps and everything seemed to work so I assumed the system was working.

I then had the idea to add 2 more sticks of RAM (Crucial 2G (2x1G) DDR2 5-5-5-18 @2.1v) to bring me to a total of 6G. Note these were removed from a thrid system when I upgraded it. The had worked flawlessly in this system for 2 years at the same timings and voltages).

As I am aware that mixing RAM has been known to cause problems, I tested. I ran memtest 86+ for 12 hours... no problem. I ran OCCT for 6 hours and prime 95 for 6 hours. All without problems.

I assumed that I was good to go.

At first, it seemed that way. I noticed big speed improvements in Photoshop and Premiere Pro CS4. Everything seemed to be okay. Ran for 8 hours of intense use (as a production machine).

Then Bluescreen. I had left the computer to make a cup of tea. Came back and saw a BSOD. First thought was that 7229 had a bug. Rebooted and finished my days work.

Left the computer on running utorrent - giving back 7229 (which prevents it from going to sleep). Next morning, I came back to another BSOD. Over the next couple of days. I would get BSODs 1 or 2 times per day, but the strange thing was, that it was only when the computer was unstressed. During hard use, it was rock solid. I noticed that preceding these BSODs, background appications such as sidebar or wmpnetwk, etc. would start to crash a couple of minutes before the BSOD (always different, but often in the Kernal).

Once again, I thought about my RAM. I reran memtest86+ for 12 hours and then the MS ramtest for 4 passes. No problems.

Next day. same pattern happened twice. Some app crashes during low activity periods and then BSOD.

I removed the 2G RAM and have run for a week with not a single app crash or BSOD. I really like 7229.

Does anyone have any ideas why the machine would have problems like this, expecially since the machine is typically idling when it crashes?

My thoughts to get us started.

1. 4 sticks are more demanding to run than 2 sticks. (but why only during unstressed periods?)
2. There is a strange bug in windows that does not like 4 sticks or a heterogenious mixture of sticks. (but why only during unstressed periods?)
3. My MB dynamically changes voltages in response to load. perhaps it relaxes too much under no load situations?

Thanks for your help.

Mark

Thanks for the suggestions.

I checked the Gigbyte site and they dad updated the bios (from f8 to f9). Even though there was no mention of memory issues in the update, I updated it anyway as a Gigabyte tech support person told me that the bios update comments only hits a "headline" issue and that there are often many more changes included.

I booted and loaded optimised defaults with the 4G RAM in. It suggested 1066 5-5-5-15 @2.1

I then booted and loaded optimised defaults with the 2G RAM in. It suggest 1006 5-5-5-15 @2.1 (Note, i had been running 5-5-5-18 with the previous bios).

I then put them both in a let it reload optimised defaults. It came up with 1066 5-7-7-24 @2.1v!

I will let it run like this and see what happens. It might be a while before I see anything.

Thanks again.

Mark

I did not have to wait too long. Two crashes ...

I then noticed in everest that the vDimm was 2.08 (and not the 2.10 that is the spec and the original reading with the previos bios).

As I had used optimised defaults in the new bios, i went to the RAm voltage settings and checked them. It was set to auto, so I went back and manually set it to 2.12v (asuming that the previous 2.10 was marginally unstable and 2.08 was unstable.

Anyway, overvolting DDR2 RAM by 0.02v should not cause a problem.

System is already more stable than it was 5 min ago. Let see how it does for the rest of the day.

Mark

Bumping the Voltage from 2.1 to 2.12v seems to have solved the problem. Rock solid for 3 days. (I still wonder why memtest did not pick up the issue).

Mark