f13.net

My M3A mobo based computer is having odd problems. Original system configuration had WD740 (old 74G Raptor) plus storage HDD. One day C:, but not subdirectories, were gone. You could not read or write to C:, but C:\DIRECTORY NAME for example would still work. Surprisingly in this condition system would still boot and work normally...

Without investing too much time into investigating this problem I decided that this was caused by faulty registry read or write probably due to end-of-life HDD (expected). I got VelociRaptor 150G as a replacement HDD, connected it SATA1 as the only HDD and started WinXP64 install.

This is where weirdness started again. Install from CD would go fine, but after that system would crap out at random points of patching/installing drivers. It would only boot in safe mode and otherwise experience extreme slow downs, (15+ mins to boot). It would not quite hang, just slow down to an absolute crawl.

What could be causing this highly unusual behavior?

a) My guess is that SATA controller on mobo is going in a very unusual way, new faster HDD stressing it more and magnifying problems
b) ? They found me again, its time to go into hiding ?

I tomorrow I plan to hook up new HDD to my computer and run stress test, also closely examine contents on the old HDD to see if I can determine (or even duplicate) the problem.

Please let me know if you encountered anything similar.

A very untechy suggestion would be to check your fans. Overheating can do some really wacky thins to a system. Its also easer to fix than a dying motherboard.

I've seen a similar thing (not the c: missing but the rest) on a machine with a dying psu.

Check your memory. Use another machine to burn a copy of Memtest86 or just go for http://www.ultimatebootcd.com/ and burn a copy of that, then run the memtest86 available on that one.

Problems just stopped. After another attempt OS install just took and it fully patched and stable. Run 10 consecutive HDD stress tests without any problems. Something is dying, I have no idea what.

I don't think it is memory, I don't have any issues with system up and running. It won't crash or BSOD under load. I will tun more tests tonight. I think it is more likely PSU, since it is at least 5 years old.

Just to reinforce this, one of my old Linux servers once started crapping all over its root filesystem.  I did everything I could think of, including replacing the IDE (old Linux server) cable and the drive itself, but the corruption continued unabated.

It turned out that I needed to reseat the RAM.  Over years of use thermal expansion and contraction had caused the SIMMs (old Linux server) to work their way just enough out of their sockets that the connection between SIMM and socket had become intermittently flaky.  The machine would POST consistently, but as it warmed up, the connection to the RAM became less reliable and all kinds of crap would flake out, including filesystem metadata writes.  Pulling out the SIMMs, blowing out the sockets with some compressed air, and reinstalling the SIMMs recreated a solid connection, and the problems disappeared.

On modern hardware, with component temperatures inching toward 100C under load, there's more thermal expansion and contraction, and the effects are felt more quickly.  Whenever one of my systems starts having bizarre intermittent problems in multiple subsystems (disk, video, RAM, etc.), the first thing I do is reseat the RAM.  Doing so doesn't always fix everything, but I'm always amazed by how many completely strange problems it does fix.

When the RAM starts flaking out, the symptoms can be just about anything, because nearly all of the data your system deals with is in RAM at one point or another.  Reseating your RAM doesn't take much time, and it can clear up problems that will otherwise stump you.

I reseated memory, no problems so far. Whatever problem was, hopefully it went away.

Wow, I honestly never knew that actually worked. Back when I was younger, that was always the running joke my friends and I had about tech support: reseat the RAM was always their answer and it never seemed to fix anything for us.

Back in the days when RAM and other chips were socketed this was a real problem. The chips would sometimes work their ways up out of the socket because of the constant thermal expansion and cooling. It really was the official suggestion by Apple to pick up and drop the Apple /// to reseat its chips when that happened.

I'm surprised this happens to SIMMs since those are "latched" (clipped into place) so there should be no way for the boards to work their way out of the sockets unless the latches are broken or they weren't installed properly in the first place.

Actually it can still happen.  Most DIMM slots use a metal wiper mechanism to scratch a clean copper connection point as you tip the memory module upwards to lock it into place.  Reseating the module will wipe a new spot for a better electrical connection if the original was flakey to begin with.

That's more a dust and corrosion problem then, rather than thermal expansion pushing the boards out of their slots.