Sorry I've been a bit absent lately, as well as the usual 'life' problems I've been trying to hunt down an obscure group of intermittent computer problems that were very slowly getting worse.
The first problem was that compositions that I had previously run without issues were just occasionally causing jack to kick out some of the applications. I thought that maybe I'd made some incompatible upgrades, so used my spare partition to do a dual boot with an older version - no dice.
The next problem was again, only on rare occasions the entire system would lock up for no apparent reason.
I carried on like this for a while, occasionally trying various changes, but over the last couple of weeks things started to get much worse, and this is when I began to suspect hardware. To prove the point I got the very oldest version of 64studio and installed that. This behaved exactly the same way so software could positively be eliminated.
Next I tried running memtest. This ran for about 5 hours without showing a single problem, which threw me off the scent for a while. But it eventually dawned on me that memtest won't stress the graphics card at all, and won't put to much pressure on the CPU either, so it was back to the drawing board. By now the computer had actually become unusable, also I noticed quite dramatic CPU temperature rises being reported.
Eventually, I took the computer out of its rack, and opened it up to have a good look.
The first thing I noticed was that the graphics card wasn't completely seated. This rather surprised me as I'm normally quite careful about that sort of thing. While working on it, I noticed there was quite a lot of flex in the case and that this made the card 'walk' out of its socket. Now my house is quite old and all the floors move when you walk across them. If you remember, a few months ago I put the computer in a new workstation that holds it quite rigidly, so if the floor moves, and flexes the workstation this forcibly flexes the computer as well. The temporary answer has been to remove the securing screw on the card and let the board socket hold it in place.
Further examination revealed that one of the memory cards wasn't completely seated either (although I don't know how this survived a memtest). It looks as if the clip on the side had been snagged by one of the internal cable looms, but I'm not certain of that.
Finally, while I had everything apart, I took off the CPU heatsink cleaned the surfaces and applied fresh heatsink compound, just in case I'd managed to break the seal while messing about with all the other stuff - bearing in mind that over the last 18 months I've doubled the RAM, changed the graphics card and upgrade the CPU.
Initial tests look promising. No more lock-ups (so far), jack hasn't booted anything out in a couple of hours continuous running, and temperature peaks at 60deg C fairly heavily loaded, whereas before it was getting above 80
