What could be wrong here?

[joemadeus]

[Specs for the machine in question are posted at the end of this message.]

It's a darn odd problem: every so often the machine will suddenly 'freeze' with no warning, usually under load, sometimes not. If sound was playing the last 'frame' of the sound will keep on playing. The display usually doesn't change at all, but sometimes shows a 'bar' of vertical lines, one DOS-prompt font in height (if you catch my meaning), usually blue on black. The laser in my optical mouse is no longer illumiated (USB model). On an average day:

SETI@Home runs 30 minutes - ~6 hours
Prime95 runs ~10-20 minutes
Rome: Total War runs 2 minutes - an hour (rather variable)
SageTV: Hasn't failed while running this yet
Idle: Haven't waited it out, yet.

The memory passes memtest, and the drives pass Hitachi drive fitness tests.

It might be related to temperature, I think: ambient temps have been in the 90's F in my house lately, and it'll freeze within 10 minutes when running Prime95 on those days, but the parts for which I have temps seem to be within spec:

(Running Prime95, latest version, low priority but nothing else exec'ing)
Ambient: ~34C
CPU (on-die sensor): ~56-60C (varies)
Power supply (thermistor in the 'fins' on my Phantom 350): 42C
Case (on-MB sensor): ~42C
Northbridge (thermistor in the fins in the heatsink, passive): 42-46C
HD 1 & 2 ('SMART' sensor enabled): ~42C
6600GT (from the video control panel): 67C (ouch!)

Today the ambient is ~28C, almost no load, 45C CPU core temp, 42C case temp, 40C NB heatsink temp, 35C power supply temp, 41 & 42C drive temps. I've been using it for about 45 minutes without a problem.

I've tried all sorts of BIOS tweaks. A subset of what I've tried:
* Undervolting and/or underclocking the CPU doesn't help
* Raising the voltage on the NB to 1.5v (from 1.05v) makes it fail faster (which I find rather suspicious, but I feel like I'm biasing, here)
* Modifying memory timings doesn't help
* Changing the type of CPU speed throttling doesn't help

The moment I try overclocking the CPU by adjusting the FSB frequency, it fails hard, even in BIOS. If I go from 133 -> 134MHz, forget it, I have to hit the reset button.

What else should I test? Could it just be a bad motherboard or video card?


Thanks for the help... this is rather frustrating.
-j


SPECIFICATIONS:
* AOpen i915GMm-HFS motherboard
* Pentium M, 2.0 GHz, 533MHz FSB, 1.308v, 'Speedstep' enabled (I've replaced AOpen's lousy stock heatsink with a modified Thermaltake Volcano 10, Artic Silver 5 interface, Panasonic 80mm fan with an 80mm/60mm adapter)
* OCZ DDR2 memory, 2x512Mb, PC4200, 3-2-2-8
* Gigabyte 6600GT, PCIe-16, standard voltage and heatsink, standard core speed and mem speed
* Hauppage PVR-150 TV-Tuner
* 2x Hitachi/IBM Deskstar 7k80, SATA-I interface
* Sony DW-26a DVD-RW/+RW drive, primary IDE interface
* Phantom 350 power supply
* Aspire X-QPack case
* Single 120mm case fan, no-name brand (read: NOISY), blowing inward
* Logitech USB mouse, wired; standard PC keyboard, PS2

[darthan]

This sounds like it could be a heat problem. The easiest way to test is to simply get a large desk or room fan, take the side off of your computer, point the fan in and run it and see what happens. If this solves your problems you need better cooling. I'm especially inclined to think that you have heat problems because a Pentium M just should not generate enough heat to get to 60C in a decently cooled computer. Also, having just one fan blowing into your case is unlikely to successfully cool it unless you have a rather amazingly carefully designed airflow path. If heat is your problem, I'd recommend getting an exhaust fan installed so that it pulls hot air off your motherboard and away from that power supply.

[merovingian]

So, no psu fan and a 120mm fan blowing inward? You have an exaust fan right?

[joemadeus]

This sounds like it could be a heat problem. The easiest way to test is to simply get a large desk or room fan, take the side off of your computer, point the fan in and run it and see what happens. ... I'm especially inclined to think that you have heat problems because a Pentium M just should not generate enough heat to get to 60C in a decently cooled computer.

I've tried running it with the case open and a fan blowing on it (I have a nice, small one that moves a lot of air) but it doesn't seem to make a difference. I agree that 60C for a Pentium M is pretty hot -- it's 27w TDP -- but this is the very top end of a four degree range, on an extremely hot day, running Prime95 full-tilt. Actually, I've seen it spike hotter -- 63C -- while running Prime, but that was on a 95F day. Figure ambient alone accounts for ~35C.

I forget what Intel rated this chip to, thermally.

So, no psu fan and a 120mm fan blowing inward? You have an exaust fan right?

I've read that the Phantom power supplies work best when there's positive case pressure; apparently they don't cool correctly if there's negative pressure. The fan blows into a clear area in the case about 4x4x5", with the two 7k80 drives at the opposite end of that 'box'. This is about an inch to the left and an inch & 1/2 above the CPU nad northbridge heatsinks. The flow is almost certainly scattered before it gets to the CPU fan, fwiw, since it flows over the drives first.

There's no exhaust fan; there're two strips of openings, about an inch wide by 12" tall, ~50% blocked, one on each side of the case, near the front (the 12cm fan is in the rear of the case.) When I put my hand next to the openings I can feel air rushing out.

I suppose you can see how well the case is being vented by looking at the case temps: is 42C at full Prime95 load acceptable, given that the ambient was 34C?

I wrote earlier that the machine had been running about 45minutes without problem; it's now up to about three hours. Ambient temp is 29C (85F), case temp is level at 39C, CPU temp level at 47C (seems high), NB heatsink thermistor reads 40C, power supply thermistor reads 37C, drive sensors level at 37 & 38C.


-j