Stability testing

decto · 6 Mar 2006 at 21:13

Stability testing S&M

Hi,

I'm a little puzzled about a few issues I'm having with stability testing.

System
Opteron 165
MSI Neo4-F
Thermal Take Big Typhoon
2GB Gskill PC4000
Westen digital 250GB 16MB cache SATA
Enermax 465W PSU
POV 7800GTX 256MB
Antec P180

I run prime95 Dual (max heat) at at almost any OC from 2Ghz to 2.6Ghz and it runs fine mostly, sometimes 12hrs without fault and sometimes 2-3hrs before detecting an error. Usually a rounding error. This even happens at stock 1.8Ghz. Blend always runs fine !!!

I run memtest86 and it will run continuously at mem 270Mhz 3-4-4-8 1T 2.6v without any errors at all. 19hrs max in one go! that was around 38 passes of the full test (CD ISO downloaded = OCZ Memtest86 v1.00)

S&M always fails on the second core (FPU test) regardless of clock speed and voltage, and always on the second loop. This doesn't matter if I run full tests or just run the FPU test on a loop. Also the memory test reports 2 persistant errors.

I'm now assuming this is some quirk, the memory is clearly clean as Memtest will run overnight without a hitch even at much higher settings than I routinley run at. I haven't had a single blue screen or system restart in month. CPU temps are mostly below 45C load but can hit 50+ with 10% volatage. Even ran memtest overnight with the CPU @ 2.8Ghz, HTT 312 and mem around 260Mhz - no errors.

I'm thinking my windows install may be stuffed, anyone give me a clue?

AD

decto · 6 Mar 2006 at 21:31

Yeah Yeah .... was typing S&M and somehow I posted the first half !!!

decto · 7 Mar 2006 at 18:22

Thanks for the info, I'm at 5 hours of small FFT prime... long way to go.

Didn't release small FFT would isolate the CPU. Hopefully I'll get to the bottom of my issues some time soon.

AD

decto · 8 Mar 2006 at 21:52

After 27hrs of dual Prime small FTT it seems the CPU is fine.

Unfortunately the memtest 1.65 shows my Gskill memory is toast so looks like an RMA ahead. I've tested it at 2.5, 2.6 and 2.8v with default timings 3-4-4-8 and slack 3-5-5-10 @ both 200 and 250Mhz, the result is a fail after several hours followed be repeated failures every few minutes.

I place a temp probe in the case and the temp is 23C, the Dimms barely feel warm.

Thanks for the help

AD

decto · 10 Mar 2006 at 16:54

After some more testing it's confirmed as a bad stick of ram. Strange how the OCZ version of memtest didn't find it.

The mainboard manual states a single dimm shoul be installed in slot 1 or 3.

Dimm A in slot 1 - errors after 40 mins

Dimm A in slot 3 - errors after 25 mins - (assume as the dimm was alread warm)

Dimm B in slot 3 - 21.5 hours and counting. Didn't test slot 1 as is tight with a big typhoon.

Timings manually set to 3-4-4-8 1T @ 250Mhz divider for each test.

Interestingly memtest reports single channel bandwidth as ~1500MB/s while Dual channel is only ~2000MB/s. I would have expected more of a difference.

Thanks again for the help

AD