Soliciting Computer Hardware Opinions

[dariusk]

My Windows (gaming, music) box has been giving me shit for about three months. It's a Shuttle case, and it keeps overheating. At first I thought it was a fan problem, but I checked every fan in my box (video card, power supply, CPU/case), and they're all working fine. Then again, when I'd reboot, my BIOS would tell me that I had only been running

Comments

[grfpopl]

1. 100 F is the temperature when it's done rebooting. It can cool down a fair bit in that time.

2. Most likely what happens is a few bits get flipped here and there, which can go unnoticed for a good long while. The more cpu work you're doing, the higher the chance of catching a problem. That's more or less the problem I had for a while, before I replaced the mb/cpu/ram.

You can try running memtest86 or prime95 to get a clue.

[dariusk]

I'm thinking you're spot on with #2. I guess I'll run memtest86.

[purple_dj]

The concept of a processor or RAM that is marginal dependent on multiple conditions isn't common, but I wouldn't think it impossible. (... this speaking from the PoV of someone working in the hard drive industry and hearing how our drives can fail under such precise conditions as when the temp is between 50 to 55C and something is shaking the test rack at exactly 73Hz...)

I second the opinion of beating on the machine for a day or so with memtest86. It will prove to you that something is wrong w/ the hardware (as opposed to your OS/virii), might narrow down whether it is a RAM or CPU problem, and will give you a more solid testbed if you want to try swapping parts out to fix the problem.

Minor questions:

• you've got thermal paste between the CPU and fan, right?
  
• have you noticed what sort of temp the machine gets to just before crashing?
  
• have you tried backing down the CPU settings in the BIOS to a slower rating than the processor actually is?

[mpgalvin]

HSF (or heatpipe assembly thingy if it's one of those shuttles) may have become less than perfectly secure? (do you move the machine a lot, like for LANparties?)

did the case fan or cpu fan (if there is a seperate one) go bad? if the fan died, that would do it.
did you move the machine to an area where backpressure could fight the efforts of the exhaust fan? (just pinging random ideas)

---
underclocking and undervolting are both ways to make less heat with the thing, but if the problem is that heat isn't escaping, that'll just make it work a little bit longer before dying.

random reboots could also mean your PSU is dipping under heavy load? also possible is a mobo V-reg dying. i *think* that's what's happening on my machine, but i haven't nailed it down yet. the reboots part (rather than the hard crashes i've been having) lead me to guess voltage, but it's still a guess. i was thinking my problem was heat for a while, but now that i've "solved" that, i'm on to the next least expensive fix. :)

[dariusk]

The answer to your questions is "no", but thanks for playing :)

[purple_dj]

you've got thermal paste between the CPU and fan, right?

It's got one of those heat pipe systems, so the fan is about 5 inches of pipe away from the CPU.

have you noticed what sort of temp the machine gets to just before crashing?

Nope. I should check.

have you tried backing down the CPU settings in the BIOS to a slower rating than the processor actually is?

Yes, I underclocked it, but my BIOS only allows something like a 10% drop in speed.

Heat problems

[runexe]

When my main server down here went up in smoke (based on my 6 year old Dell Dimension - 300MHz PII - that I got as a frosh), I switched the drives over to my other desktop - the now only 3 year old Athlon-based system. But I couldn't keep it running for more than 5-10 minutes before it'd reboot. It took me a little while to figure out it was heat related, and fortunetly my BIOS allows plenty of tweaking in the system clock category - although it was only about a 10% drop in clockspeed before the system would run stable in the hot/humid eastern caribbean enviornment.
So, my money is on the CPU is still fine, something is just getting too hot (either CPU, or possibly the northbridge/memory). Good luck. Hope memtest or similar gives you a clue as to whats up.

[qedrakmar]

The spotting on the CPU is mostly baking of the production environment and the enamel, not the "chip" itself. I've seen this a lot, and while it's not a particularly good thing, it's not particularly bad, either. You know faster chips are more susceptible to heat issues (tighter packed traces = lower energy needed to jump traces), so you shouldn't be surprised by a fast chip having issues at what seems like a normal temperature.

Does it die when you're working on it, or while you're away?

[dariusk]

It only dies when I'm doing something computationally intensive. I can leave it running idle indefinitely, with say Windows XP with Firefox open and my usual background apps.

[qedrakmar]

How does the CPU usage here stack up against the memory usage (specifically Cache usage)?

speedfan

[ilp]

if you run speedfan, it will tell you what your temperatures and fan speeds are, so you can log the temps right until the freeze.