Site Update: Code tidying ►

◄ Firefox 1.0 is released! Tell everyone you know! Go now!

2004-11-09 📌 So my computer spent a couple of months slowly dying...

Tags All Tech

...hence the fewer updates. The following was written over the course of a week, though the problems it refers to had been creeping up for weeks previous to that.

SYMPTOM/PROBLEM:

Display freeze, no response to ctrl-alt-delete but screen output continues (monitor can be powered-off, unplugged and re-plugged, with the same hung screen evident.)

System is irrecoverable, but boots fine after a freeze.
Happens whether using onboard graphics or separate AGP card.
Happens whether using PCI or serial modem.
Happens with no expansion cards in the machine.
RAM checks out okay with MemTest-86 when using for graphics.
RAM checks out okay with MemTest-86 when using AGP card.
Motherboard, processor and hard drive report safe temperatures.
Machine seems fine when left on overnight to defragment, encode MP3s for several hours, or just run a screensaver.

I've replaced the power supply (which wasn't great quality) with a better one, just for the additional peace of mind that it's unlikely to fry the hard drive. Voltages are steady.

DEDUCTION PROCESS:

The first inference is a software fault, as the display doesn't disappear. The problem *seems* connected to a live net connection being in use; one or twice it's hung suspiciously straight after dialing. It's hung whilst downloading in either Internet Explorer and Firefox, so it isn't browser-specific. The frequency of freezes *seems* to have reduced since taking out cards and switching from Kerio Personal Firewall 2.1.5 to Sygate 5.5, and I'm noticing frequent port probes although no verifiable correlation between intrusion attempts and hanging. On the other hand, most freezes *seem* to occur when connection is tied up with incoming and outgoing requests combined with high CPU usage. The system also *seems* far more stable during early hours of the AM. That's the evidence towards a connection/firewall problem.

However, no freeze occurs on any GRC.com port scans at 100% CPU usage with heavy system load, downloading other things and running a couple of videos on repeat simultaneously. And if it were all down to remote attack, I'd assume my sister's PC (also running 2000, which I used until I could replace the PSU in mine) would suffer similar freezes.

So, I'm back to thinking it could be the motherboard, RAM or CPU. Which is bloody frustrating... looking at www.badcaps.net it's possible, even likely, that the K7MVA is from a period affected by dodgy capacitors. All of which doesn't link it with the perceived instability of the system increasing at certain times of day, but would explain why the other PC in the house with a similar software setup behaves fine...

One thing pointing to the motherboard is that the keyboard/mouse LEDs seem to start losing responsiveness on a freeze. Removing and replugging the keyboard regains that for a short while, before they start locking in state again. I'm thinking that removing all of the cards has assisted the motherboard in not building up some kind of charge, which is dissipated by a power reset. This would explain why it's most flaky a) when just switched on from booting up cold in the afternoon, and b) after running for a good few hours. I'm guessing that whatever part is defective is tied up with the PCI and serial buses, as simply leaving the system on hasn't resulted in a hang. It's also possible that the cheapish power supply in there before may have hastened the demise of the board.

Now, this presents a problem: can I get a Duron-suitable motherboard which wouldn't be of the same vintage that had endemic capacitor problems? I really could do without needing to replace board, processor and RAM. Hell, I could do without a steadily dying board. I also ideally need a replacement with integrated graphics and sound, which means shopping around in the same manufacturer and type category.

Research to do: are motherboards which will take Duron chips still manufactured, or are retailers dealing with old stock which would suffer the same problems? That should reduce risk of duff capacitors.

SOLUTIONS, PROBLEMS, SOLUTIONS:

Well, I tracked down a board... the PC Chips M825 has onboard AGP/audio/LAN, 4 USB 2.0 ports, an AGP slot for expansion and a couple of PCI slots. It takes PC2100 RAM, which is an additional expense but pretty much unavoidable, and it'll take the Duron and processors up to about an Athlon XP2000 if I want to give it a bit more power later on. 256Mb of DDR turns out to cost more than the board, although that is from Crucial (never, ever buy dirt-cheap RAM as it can be a troubleshooting nightmare.) Whilst still frustrating, the situation did seem to be looking up.

Of course, things rarely are that sodding simple.

Switched everything over, double-checked jumpers, power connectors and everything else... motherboard doesn't post.

Take everything off, leaving RAM, VDU cable, keyboard and case power button plugged in... motherboard doesn't post.

Switch board back again, to make damn sure I haven't accidentally fried the CPU by cleaning it with vodka. (It seems nail varnish remover no longer tends to include acetone, and thirty year old vodka—produced before manufacturers started watering the stuff—is the nearest thing I could find to isopropyl alcohol.) The old board posts.

Resmear the thermal grease and stick stuff back into new motherboard... which continues to not fucking post.

Treble-check jumpers and everything else... still nothing.

Unscrew motherboard from case panel, leave on desk, power up... we have lift-off.

It seems that the brass mounts on the case panel were shorting everything, in spite of the pretty red plastic washers. So, stripped it down to four well-insulated mounts, having spent the evening and early morning punching walls.

CURRENT OUTCOME:

Despite the RAM drop, everything is much faster and more responsive. In particular, the DDR boost helps the onboard graphics provide decent DVD playback, something the previous board needed a separate card to achieve. Net access has also resumed a perkier rate; presumably because it isn't being funneled through a dying serial/PCI bus. I now have enough USB ports for all of my devices. The processor is running much cooler, although I consider airflow inside the case could be improved for the benefit of the board and hard drive.

I've also got two redundant sticks of PC133 SDRAM, the 256Mb one of which will be heading towards my sister's old box, and a sizeable array of now-redundant parts (I've decided to stick with the £7 serial modem I bought rather than putting any cards in the machine. It'll work well with Linux when I have time or inclination to dabble.)

Hopefully this machine will see me through until I want to build another from scratch; I can always shove more RAM or a faster AMD processor into it when prices are good, and I may look into sound-dampened fans at some stage.

💬 Comments are off, but you can use the mail form to contact or see the about page for social media links.