An update: the error on my uncle's machine is indeed caused by the dead RAM. After replacing the 256MB module with a new 512MB module, the system is up and running, faster than before!
So if the system is doing random failures, check the memory!
Some FAQs about PCs you wish you knew before you bought it!
Friday, March 11, 2005
Wednesday, March 09, 2005
A bit of advice for people diagnosing a seemingly random problem: turn OFF the "quick power-on test". This makes the PC do the FULL memory test and POST (power-on system test) instead of the abbreviated version. Nowadays systems are fast enough that the time difference is negligible.
Why? Let me tell you this story.
I just spent a hour trying to fix my uncle's computer. He has a 1.1 GHz Duron with 256MB of RAM, loaded with Windows 2000. He said it started giving him a bluescreen. When I got there, it says something about INVALID DEVICE. I rebooted the system, picked "last known config", and got "cannot load SOFTWARE" registry. I rebooted and try "safe mode", same thing. Tried "safe mode with command prompt", same thing. Tried "Diagnostic mode", same thing.
At this point, I'm willing to concede the system is completely hosed, and he probably got a virus, but I know my uncle. He's a very safe computer user, never downloads anything except official sites, etc. etc. Chances of him catching a virus or trojan is nil. Yet the evidence is the setup is probably hosed. So what now?
I put in the Win2K CD and the 4 boot floppies. After the 4th disk, just as it tries to load the setup menu... The computer restarted itself. Yikes!
I put in my XP setup CD and set the BIOS to boot off the CD. The setup screen loaded all the drivers... Just as it got to "install or repair" screen... It blue-screened out with some error in a certain driver. That makes no sense, as we haven't even seen the GUI yet! So I rebooted... This time I got the "install or repair" screen. I hit REPAIR, just to see what would happen... the computer rebooted.
At this point, I'm completely stumped, so I'm changing things in the BIOS and started turning off all the options, so I turned OFF "Fast Power-On Test", and rebooted. Ahem! Memory Test Fail. The 256MB DIMM lost at least 1 bank. It stopped at roughly 236MB during the memory check.
The Fast test would have never picked it up, and I would have been still in the dark.
You may have wondered... didn't I turn on the BIOS halt errors? Yes I did. Originally it was on "halt except disk errors". I set it to "halt on all", but still the same. It's only AFTER I turned off "fast POST" that I saw the memory error.
So next time you trying to diagnose something random, think memory problem first.