<quote who="[EMAIL PROTECTED]"> > just tipped us over the edge into using virtual memory too much. How > is VM nowadays? I know it's been dodgy for quite some time now, with > much heated discussion amongst kernel developers.
No, VM is just dandy for me with 2.4.19-pre4 uptime 23:54:15 up 15 days, 16:23, 4 users, load average: 0.36, 0.37, 0.20 Last reboot was to move a video capture card in. > Jan Schmidt wrote: > > > The machine was perfectly happy with the 300W under low CPU conditions. As > > soon as I ran something like burncpu it would lock up though. > > Hmm, I guess I can write a while true do and see what happens. I'll > send this email first, though. :-) As I said, I was using the cpuburn package, which runs tight assembler loops to heat CPUs up. With the dodgy PSU and the standard CPU core voltage, it would just hard lock within 10 seconds. After the PSU upgrade, I ran it for 8 hours and got the CPU up to 55C Other symptoms were programs randomly segfaulting, particularly video encoding and other such CPU intensive apps. > Jan, did your crashes have any log message written to > /var/log/messages, or was it a crash without warning? Nothing - the machine hard locked at the physical level, I believe. > I'm also surprised that 2 of today's reboots (the one at 4am and the one > at 6pm), were completely unattended. If the system had just locked up, > it wouldn't do that, would it? The power would have to drop for that > to happen, surely? Modern CPU's can be quite sensitive to voltage supply fluctuations, particularly athlons. With el cheapo power supplies, the droop caused by the CPU requesting more power can be enough to drag the voltages down. J. -- Jan Schmidt [EMAIL PROTECTED] Pants Pants Pants Pants Pants Pants Pants Pants. Lovely Pants, wonderful Pa-ants. Lovely Pants, wonderful Pa-ants. (Shut up! Bloody Vikings.) -- SLUG - Sydney Linux User's Group - http://slug.org.au/ More Info: http://lists.slug.org.au/listinfo/slug
