The system- MB: MA790FXT-UD5P
- CPU: Phenom II 965 black edition
- Hard drives: Hitachi_HDT721010SLA360_STF607MH3SL8YK, Hitachi_HDT721010SLA360_STF607MH3U8WAK, TOSHIBA_DT01ACA100_Y2RTLRHPS
- HD IF: SATA, non-raid
- Video: GeForce 9500 GS
- P/S: Antec SP-500, 500W
- RAM: G-Skill F3-10666CL7Q-16GBXH (running at 9-9-9-24); formerly F3-10666CL8D04GBHK (two sets)
- OS: Debian Wheezy, 64-bit, kernel 3.2.0-4-amd64, up-to-date
- Overclocking: none
- A/C power: via UPS
CPU temp almost never exceeds 60C; when it does, I blow the dust out of the heat sink and it drops to a reasonable 57C. Hard drives almost never get above 44C.
The problemThe system works fine until memory is filled and the system has to start dropping cached disk blocks. At this point, I almost invariably get a kernel panic or other GPF.
For a while, I thought it was a RAM issue, so I bought the 16GiB memory (4 sticks) and sent the 8GiB (4 sticks) back to G-Skill for testing. The new RAM wouldn't work well until I set the timings to 9-9-9-24.
I can fill the memory building Smoothwall v3.1 (either 32-bit or 64-bit), or manually read enough files to fill Linux's disk cache. Every time, as soon as the system needs more RAM, it crashes. I can even build in a KVM with 512MiB assigned and the system will still crash.
I really have no clue as to where to look next. Or how to look....
So what is likely to be the problem?
- Linux? I imagine I could build a custom 3.4 kernel, in the event 3.2 has a problem with paging.
- mainboard? Would there be any visual clues?
- CPU? Have the '965s' ever been known to have VM problems?
- Power supply? It's a fairly new supply; I doubt it's the source since the failure seems to happen only when RAM is filled, But I could put a scope on the various outputs just in case power is dicey.
Basically, I ask here to see if anyone has ever encountered similar problems with this mainboard/CPU combination.
Thanks!