Inconsistent Memtest errors, possible compatibility issue?

[singularityflow]

System Information:
RAM: G.SKILL ECO 8GB (4 x 2GB) 240-Pin DDR3 1333 (Model F3-10666CL9Q-8GBECO)
Processor: AMD Phenom II X4 955 3.2GHz
Motherboard: GIGABYTE GA-MA770T-UD3P AM3 AMD 770 ATX


Problem
Ever since I built my computer in January, I have experienced periodic application crashes and blue screens. Memtest86+ would show errors, however, when I reseated the RAM, Memtest would no longer return errors. I would return to using my computer and go another week or two with no crashes before they would return.

Unfortunately, crashes in the past three days have increased in frequency, and reseating the RAM no longer makes them go away. Even more concerning is that Memtest does not return consistent errors. For example, when I started to experience crashes/blue screens most recently, I ran memtest and received an error. In an attempt to locate the problem, I ran memtest again only on the first two sticks of RAM (in the first and second slots) and did not receive any errors. Then I ran memtest again on the last two sticks of RAM (in the third and fourth slots) and did not receive any errors. Finally, I ran memtest again with all 4 sticks back in their positions and did not receive errors. Unfortunately, the next time I booted my computer, I received a BSOD ten minutes in; running memtest again returned errors again.

Note: for all the memtest runs above, I only allowed one pass to complete, rather than multiple passes. Also, I am not overclocking my system at all.



I?ve tried to brainstorm some possible reasons for the problem:

Incompatibility between AMD processor and the memory timings/speed?
After reading this thread on your forum, I am planning to lower my memory speed and timings to see if that provides stability (I am currently using the default 9/9/9/24 timings). Unfortunately, I am not extremely familiar with these settings and am open to suggestions on how to determine new settings to try. I believe I?d just drop from 1333MHz down to 1066MHz, but how should I determine the timings? (On another note, the RAM specs state that it is specifically designed for Intel i5 and i7 processors; is this another reason why I may have to alter the timings for an AMD processor or is this just an inconsequential marketing advertisement?)

Incompatibility between AMD processor and the memory voltage?
My RAM is currently running at 1.35V, and I believe the AMD processor I am using supports 0.875V - 1.5V, so I don?t believe this should be an issue.

Incompatibility between Gigabyte mobo and the memory speed?
After reading this sticky on your forum, I wondered if it applied to my RAM and mobo. I?m only running at 1333MHz, however, so I?m not sure if it is applicable.

Bad stick of RAM?
I don?t believe this is the case, as I assume that Memtest would return steady errors if a stick was bad, but please let me know if this assumption could be incorrect.




I appreciate any insight you can offer into this problem. If any other information could help with the diagnosis, please just let me know. Thanks in advance for your time!

[Tradesman]

Do you have the Command Rate at 2N? (Many AMD boards default to 1N), Might need to increase voltage to the CPU/MCH as the AMD MCs are on the weak side. What's the tRFC? will want 110 or 127 depending on options. Do you have the latest BIOS?

[singularityflow]

Thanks for the quick reply!

The Command Rate is set to 2T and the tRFC is set to 110ns for all four slots/DIMMS (both are the default values).

Since you mentioned voltage, I checked the voltage settings in the BIOS again and noticed that the DRAM Voltage Control is set to 1.6V by default (rather than the 1.35V noted on the product specifications). Could this be an issue? Should I manually change it to 1.35V?

Two BIOS updates have been issued, however the description of the updates are:
- Add CPU Core Control option
- Fix USB keyboard compatibility
- Support EuP Lot 6 function (for rev 1.1 only)
- Update AGESA 3.7.0.0 for AMD 6core CPU

Since none seem applicable to my issue, is it still worth updating other than as a last resort? I have never flashed a BIOS before, but if needed, I'll certainly learn how.


edit: CPU-Z is reporting my RAM voltage as 1.35V. Perhaps I'm misreading the BIOS settings?

[GSKILL TECH]

To try DDR3-1066, you can use 8-8-8-24 timings.

The reason why this kit was for Intel, is because only they are able to operate 1.35V. As you have already experienced, your minimum DRAM Voltage Control is 1.60V. So this may be too much for the memory. Try raising the HT and NB Voltage +0.2V to see if that helps. CPU-NB Voltage +0.1V may help as well. It simply depends where voltage is lacking, but once you supply enough, everything should run just fine.

A BIOS update could help, so it is worth a shot. As far as CPU-Z reading 1.35V, that is the suggested value, not what the system is actually operating at.

Thank you
GSKILL TECH

[singularityflow]

Thank you very much for the reply!

That's disappointing to hear I made a mistake with the RAM voltages, but good to know for the future. I'll be trying your suggestions over the next week to see if I still experience any crashes. I'll post if things go awry or I have any further questions, but hopefully the voltage changes will solve the issue. Thanks again

[singularityflow]

Unfortunately, after upping the voltages (using your suggestions of +.2V, +.2V, and +.1V) and the new speed/timings (1066 and 8-8-8-24), I am still experiencing unexpected crashes. Memtest is also still returning errors similar to before. (I've actually reverted to my original speed and timings as my system seemed even less stable with the 1066 and 8-8-8-24 timings.)

Since you advised that I need to find where the voltage is lacking to make the system stable, I assume I should try upping the voltages more. Are there any general guidelines to follow when making changes to the voltages? Would it be safe to up them another +.2V, +.2V, and +.1V each?

Thanks again for your time and help.

[GSKILL TECH]

You should not need any voltage changes for DDR3-1066, only DDR3-1333+. Try two sticks at a time, confirm each pair will operate at rated frequencies, then put them together. Otherwise, right now it seems like a module may be defective since you are encountering issues you should not be.

Thank you
GSKILL TECH

[singularityflow]

Thanks for your reply.

The past 2+ weeks I've continued to try to identify the source of the problem.

First, I tried testing pairs of memory like you suggested. When all four sticks returned memory errors, I removed two DIMMS and ran Memtest on the other two and Memtest did not report any errors. I then swapped out the DIMMS and tested the two I had removed and Memtest did not report any errors again. Lastly, I put all 4 DIMMS back in and ran a final Memtest, only to have it also not report any errors. It appears that something about reseating the DIMMS causes the problem to temporarily disappear. This doesn't appear to indicate a defective DIMM to me (though please correct me if I'm incorrect). I only ran Memtest through one full pass - should I try running an 8+ hour Memtest with multiple passes on each pair of DIMMs to make sure it's testing them thoroughly?

Second, I tried disabling the AMD Cool'n'Quiet technology, which apparently can affect RAM voltages at times. This appeared successful at first; however, 2 weeks after running with what seemed like a stable system, I received an unexpected blue screen and a subsequent Memtest run reported the same memory errors as before.

Finally, I flashed my BIOS and upgraded to the latest version. I've only upgraded the BIOS in the past 24 hours and will report back if this solves the problem; however, I did notice that the upgraded BIOS removed the option to adjust the SB/HT voltage. As a result, I only increased the NB and CPU-NB voltages by +0.2V and +0.1V respectively. Is there another voltage that I should increase now that I can no longer adjust the SB/HT voltage?

Thanks again, I appreciate your help in trying to get to the bottom of this issue!

[GSKILL TECH]

As long as your memory settings are configured properly, you should not have a problem. Keep me posted on the results.

Thank you
GSKILL TECH