CPU or Motherboard?

[jimux]

I have a new build setup, and it has been behaving erratically and inconsistant. I believe with the motherboard or the CPU needs to be RMAed. I'm not sure that the exact location of the problem can be narrowed down from the symptoms, but any help would be nice.

Motherboard: MSI P67A-G45
CPU: Core i5-2400S BX80623I52400S

Symptoms:

memtest86+ failed once. I had both sets of ram (totaling 16GB) installed and it had a fail on test 6. I was unable to recreate the failure with hours of testing. This leads me to accuse the CPU, as the memory controller is onboard. A single error out of hours tests wouldn't account for the numerous other issues faced. I also had the same symptoms listed below when using individual sticks and sets, which all tested fine.

Windows only boots most of the time. No idea what causes it to frequently restart during boot. Once up, applications frequently crash. Installers for applications frequently complain about corruption. Though the last few times, explorer.exe crashed on startup and I couldn't get to my desktop.

The installer for windows boots, but can't install (it gets to extracting files then complains it can't access needed files).

Ubuntu live-cd never boots. None of my live CDs boot, except for stresslinux in failsafe with the safe kernel. Backtrack falls back to a busybox prompt.

With Prime95:
Blend yields this error: FATAL ERROR: Rounding was 0.5, expected less than 0.4.
Small ffp runs for a minute or so before the system crashes and reboots.

Adjusting vcore and memory voltage has no impact. I have not overclocked anything.

Firmware on everything is up to date.

Tried three different power supplies, no impact.

 

[jaquith]

Welcome to Tom's Forum!

Anytime I see 'erratic' I first think a failing PSU or sometimes a short.

Having an error with Memtest with 16GB (4x4GB) doesn't surprise me too much especially if it's not consistent; see -> http://www.memtest86.com/tech.html IF you were running into Test #7 errors then that would indicate the CPU's IMC. Further, if your RAM is DDR3 1600 or faster and/or a 'non-matched quad-set' meaning it wasn't a 4x4GB set e.g. CMZ16GX3M4A1600C9B -> http://www.newegg.com/Product/Product.aspx?Item=N82E16820145347 then that can be addressed with either a QPI/VTT/VCCIO Voltage, DRAM Voltage increase or CAS Timings adjustment.

There are a lot of things that can cause 'FATAL ERROR: Rounding was 0.5, expected less than 0.4' since you're not OC'ing a PSU failure can cause a 'vDroop' and crash your system.

Tried three different power supplies, no impact IF the PSU is undersized this can still be a problem.

----
Q - How long does it take for Prime95 to fail?
Q - What were the Temps?
Q - Have you tried a Breadboard?
Q - What HSF?
Q - What PSU?
Q - What GPU?

 

[jimux]

Prime95 fails in about thirty seconds.

I wasn't monitoring temps as much as I should have, but the highest I saw the CPU go was 42c.

I am uncertain about your meaning of a breadboard beyond basic electronics.

HSF: Noctua NH-U12P SE2 120mm SSO w/ Xigmatek PTI-G4512 thermal.

PSU: Rosewill RV2-600, HEC Raptor R500, and BFG BFGR450WGSPSU. All have operated in previous systems without error. No new PSU was used in this build.

GPU: Currently installed is my PNY Geforce 9800 GT. I also used a SPARKLE Geforce GTS 450.

 

[jaquith]

Okay so much for 'Over-Heating'

The Noctua NH-U12P appears to have a plastic composite backplate so probably no shorting.

I'm now confused about "Tried three different power supplies, no impact" - please elaborate. If you tried all 3 PSUs Rosewill 600W, HEC Raptor 500W, and BFG 450W and all yielded the same results = FAILURE then it's probably a short.

Simple first: Run MSCONFIG -> select Diagnostic and reboot
* Drivers e.g Chipset Driver were all downloaded from MSI and installed
* Device Manager has no {? or !} listed devices

Test

Failure:
Assuming the backplate is plastic -- try a Lazyman's breadboard:
* Disconnect ALL headers {USB + Front Panel}; leave only the PWR+/PWR-
* Disconnect ALL peripherals ; try a diffident keyboard/mouse if USB
* Unscrew all of the MOBO and PCIe screws ; support the MOBO with a towel away from anything conductive including the I/O shield
* /I assume the GPUs are tested and good/
* Run MSCONFIG -> select Diagnostic and reboot

Test

 

[jimux]

Most astute! I recalled while building that the bases of the screws I used to secure the motherboard were wider than I thought healthy, but thought nothing of it once the system first turned on. A great catch considering the myriad of disjointed symptoms! Thank you very much! I have only run Prime95 so far, but both small ffp and blend mode work minutes on end without issue.

 

[jaquith]

Let me know...