So I went to the shop yesterday and got a replacement for the defective RAM module. The salesman took the old one, disappeared for ten minutes (to check it with an analyzer) and came back growling "yes, it is _really_ broken" and gave me a different one. I went home, plugged it in, and, at last: memtest86 does not find any bit errors on this one. Instead it just crashes with an invalid operand error after 20 minutes. WTF?! I tried it again, this time it crashes after 6 minutes. Weird. Okay, maybe memtest86 doesn't work on that board properly. Booted up Linux, copyied some pictures around, and the machine crashes. Well, something is wrong here. Let me check the RAM module again, with one of the original 256 MB modules in slot one and the 1 GB module in slot two, just to find where the module is failing. What, the problem is gone? After memtest86 running without any problems for five rounds I decided that the system is stable now. No errors even after running the bit rot, um, fade test over night.
I think it is a thermal problem: the first RAM slot is quite close to the CPU heat sink, probably too close (I need to check the specs...) and the new module with RAM chips mounted on both sides generates more heat than the old single sided modules. After a while it gets too hot in slot one and starts to fail.
I mean, yes, PCs are quite cheap in comparison (though my FSC desktop system wasn't exactly cheap, and Kingston memory isn't either), but this is ridiculous. And don't get me started on PC based server systems. To be fair, the FSC server systems don't seem to have too many instabilities, contrary to many other well-known brands in the PC server market...
Comments
lynard-.livejournal.com 18 years, 1 month ago
Hm... prüf mal die beiden Riegel einzeln. Evtl. war der Riegel auch nicht 100%ig in der Fassung drinnen - auch wenn es beim reinstecken klickte (habe das selber schon erlebt, das produziert Zufallscrashs ohne Ende).
Link | Replywoelfisch.livejournal.com 18 years, 1 month ago
Die Riegel hatte ich in den Kombinationen 1 GB / 256 MB, 1 GB / -- und 256 MB / 1 GB drin, und das 1 GB-Modul hatte ich auch mehrfach neu in Slot 1 gesteckt. Das zeitliche Verhalten bei identischer Last (kaltes System: 20 Minuten bis Crash, warmes System: 6 Minuten, reproduzierbar) ist dabei aber schon eindeutiger Hinweis auf ein thermisches Problem. In der aktuellen Konfiguration ( 256 MB / 1 GB) läuft das System seit 36 Stunden ohne Probleme durch.
Die gute Nachricht ist: die Ersatz-Festplatte ist unterwegs (wohl im Schneechaos steckengeblieben, sonst wäre sie heute schon gekommen.)
Link | Replylynard-.livejournal.com 18 years, 1 month ago
Hm... okeee
Hehe, die Platte wird wohl festgefroren sein, oder? ;)
Ah... Hab heute eine Mail vom CaCert bekommen, Punkte wurden gutgeschrieben ;)
Link | ReplyNew Comment