TWIN3X2048-1800C7DFIN - Faulty Module Identified [Asus P5E3 Premium, X3350]

LstThngYouC · September 2, 2008

This is my first time posting on this forum, although I have been a frequent visitor and have greatly benefited from other user's experiences over the last few months. I felt it only fair that I should share my own experience with the hope that someone else might take something from it. :biggrin:

I built a new system back in March. The system has been pretty stable since day one, especially with BIOS 0402 (Asus P5E3 Premium), but it was never quite 100% stable. To quantify that statement, I rarely encountered any BSODs (WinXP or Vista), could play games hours on end without major problems (Crysis, Bioshock) & encode/decode video and audio. I did however encounter the occasional application crash, RAID array failure (happened once but was recoverable) and corrupted Vista file system (2 times, had to use the DVD to fix).

I never have run my memory at 1800MHz because I like to keep my component voltages low and haven't had the need as of yet. Below are the speeds I have always run this system at:

Intel X3350 @ 3.4GHz (425x8)

Memory @ 1700MHz 7-7-7-20, CR 2, Perf. Level 7, 1.94-1.96V

After upgrading to BIOS 0505 in July, the system became unstable. Ever since it has been nothing but the endless loop of tweaking settings--> run memtest 2x--> run prime95 25.6 --> fail anywhere between 3-10 hours --> repeat process.

To make a long story short, there was no voltage on the CPU (core, pll, vtt, ref. voltages), NB or DIMMs that would result in a stable system. I also played with the CPU, NB and memory skews to no avail. I would always have a system that would work, but any prime95 run would always fail between 3-10hours. Finally I found a faster way to troubleshoot the problem: Memtest86+.

With memtest I quickly discovered my system always errored on Test #4 (moving inversions, random pattern). So I dove in this past weekend to really figure out what was going on. Starting from stock speeds and moving up (increasing voltages as needed), I hoped to find a stable point. What I found was a definitive pattern in my memory failures at any fsb/cpu speed:

(1) Test #4

(2) Memory address 1282.8-1283.1MB

(3) Error bit was always 000X0000

After reading some other forums it seemed that having the same address & bit fail could most likely mean a memory hardware issue. So tonight I finally decided to test each module individually, and sure enough memtest86 passes with flying colors with 1 module and fails repeatedly with the other in the same slot (A2). Also, the failing memory address is at exactly half (641.5-641.9MB) and the failing bit is the same.

I requested an RMA this afternoon. RAMGUY, do you think this is the appropriate action or is there something I have overlooked? Has anyone else had similar experiences? If anyone with a similar system is interested in the settings that have worked for me in general, i.e. voltages vs. fsb minimums, please feel free to respond.