Kainaw's Journal: Computer Restart Woes

Who has had a computer that restarts without warning at completely random times (even while editing the BIOS)? My computer had issues with restarting a few months ago. It lasted for a few days and stopped after I updated my kernel. So, I figured it was something wrong in the kernel and left it at that. Then, three weeks ago, it restarted. When I say it restarted, I don't mean that it threw some error, shut down, and began booting up again. I mean that it made a loud click, the power shut off, the monitors went black, and then the happy BIOS report began as it powered back up. What would you do? Blame a power spike? Blame a software bug? Wonder if there's a ghost pressing your reset button?

My first suspicion was the memory. A few days earlier, I added another gig of RAM. So, I reseated the memory sticks and ran memtest86 all night. Memory checked out just fine and memtest86 never restarted. Perhaps reseating the RAM was the fix. Then, it restarted again.

OK. It isn't the memory. Perhaps it is power. I went out and purchased a brand new 800W UPS to ensure a nice steady flow of electrons into my power supply. I turned it on everything was great. I went back to doing work and figured I fixed the problem. Then, it restarted again.

So, memory is OK. Incoming power is OK. I know the video card sucks more power than it should. I reconfigured the power cables so it had a line all to itself and the hard drives shared another one. It restarted right away.

Maybe a component went bad. I removed the floppy and DVD drives. I removed the extra network adapter. I replaced the power supply and fans with spare ones I had from when I initially put in super-quiet fans. It ran fine for many hours. I began to wonder how I could detect which device was causing the restarts. Then, it restarted again.

At this point, I lost my temper. I unplugged everything that could be removed. I even unplugged the reset switch, just in case it was causing the trouble. Nothing worked, so I took the computer to a repair shop before I broke something. They checked it for three days and only had it restart once. They blew out the dust and gave it back to me, suggesting I get a new keyboard and mouse. I went a step further. I took the computer to my office and plugged it in there. I found a trick to force it to restart: "ping localhost > ping.log". Apparently, high network traffic while writing to the hard drive forced a restart. I decided to pound the hard drives to see if I could get one to throw an error. I tried "cat sda", but that ran fine - no restart. Then, I was told to try running bonnie++. I've never heard of it, so I downloaded it and tried it. After severely abusing my hard drives, I found DMA error warnings. Further investigation indicated that my BIOS was autodetecting my 333MHz RAM as 400MHz. I set it to 333MHz and suddenly the "ping localhost > ping.log" trick wouldn't force a reboot. The bonnie++ program ran without any DMA warnings. I solved it! I took my computer home and worked all night without a single restart.

The following day, I was back to doing work and, without warning, a restart. I checked BIOS - RAM was still set to 333MHz. I checked my tricks to force a restart and bonnie++. It wasn't RAM. So, I completely disassembled the computer. I took absolutely everything apart, cleaned every part, put new goop on the CPU to keep it cool, blew out dust from every crevice, and then I found something odd. When I removed the CPU's fan, I found thick white dust caked all over the main GPU. I washed it off with rubbing alcohol and there was some discoloring, possibly from overheating. Could it be that my assumption about the video card causing the problems was actually the video card overheating? I put it all back together and now I've been working for three days without a single restart. Please pray for computer that it doesn't restart again.

The beahvior...

[johndiii]

The behavior does sound like a heat problem. I've never had s problem with a video card, but then again I don't buy the high-end ones. But I did have very similar symptoms from CPU overheating. I ran a monitor for a while after I reapplied thermal goop and the heatsink, but I didn't have problems after I did that.

Perhaps the board manufacturer has a status monitoring utility. And you might try and figure out why the GPU was caked with dust like that. Sometimes the sir flow in the case is weird, and it

Re:

[nizo]

Yeah I second the overheating comment; sounds like a heat problem because it is so sporadic (best kind to track down). Which reminds me, I have a computer that has been resetting too *sigh*

Oh and btw, they sell a cheap power supply tester (look on newegg; good ones start at about $15). Well worth having if you piddle around with machines often. Though it probably won't show you a problem if your PS is overheating because it is overloaded (I need to see if there is a newer tester that loads the PS while test