On Tue, 2004-02-24 at 17:11, Keith G. Robertson-Turner wrote: > On Tue, 24 Feb 2004 00:00:08 -0300, Alexandre Strube wrote: > > > Besides a bad memory module (as Keith already pointed out, and you can > > test with http://memtest86.com), there's a issue with rpm which > > sometimes does that. > > I've found memtest86 a less than reliable method for determining if memory > modules are bad. > > On several systems I've repaired, memtest86 passed every time, but a > kernel rebuild loop caused segfaults. Replace the ram, no more segfaults. I've also found this. I installed a new pair of 512MB sticks into my home server and began regularly crashing within 30 minutes. 24 hours of memtest later, showed no problems, but exchanging the ram for 2 new sticks fixed the problem. I was very disappointed to see that memtest86 did not exercise the ram as strenuously as I had believed it would. It's a great tool, but it is not infallible. Cheers, Ben
