Aaron Kulkis wrote: > Richard Bos wrote: >> Op Friday 11 May 2007 20:24:49 schreef Petr Klíma: >>> Aaron Kulkis wrote: >>>> I think it's more of an observation effect. >>>> >>>> Windows is so buggy, that when hardware errors do occur, >>>> it's just background noise in the all-too-typical crashing >>>> and failing... >>> Well, I don't agree at all. What I experienced was correct Windows >>> behaviour (no errors, at least none reported) while in LInux programs >>> crashed now and then without apparent reason. >>> >>> I agree with Carlos that LInux most probably uses hardware more >>> aggressively, something like leaving less time between successive >>> actions therefore leaving less time for the things (signal levels etc.) >>> to settle down. That wouldn't be a problem for perfectly stable >>> hardware >>> whose critical operating frequencies are quite higher than the real >>> operating ones (e.g. all transients finished soon enough). Once you got >>> hardware which is operating at (or behind) the edge, you may get >>> anything. >> >> As example: we once obtained a computer that had been running windows >> fine for ages. The moment we started installing linux on it, it >> failed. Indeed it already failed during the installation! Running a >> memory check tool showed that memory was bad => computer to the IT >> department, they stated that there was nothing wrong with the system >> using their tools! After talking a bit longer the faulty memory got >> replaced and the machine started to behave correctly. > > And upon re-installing windows, it MAGICALLY worked properly > with bad memory??? > > > I call bullshit. > > NO software works properly with bad memory. >
That would depend on whether parity checking is enabled. If not, you could easily have errors and not know about them. If the bad memory contains executable code, you could have a crash. If data, you might have an error and not realize it. Incidentally, parity memory introduces another failure mode, in that you don't know if the parity bit or parity check circuitry is bad. You only know you're getting errors from somewhere. There's a better system called CRC, which can not only check for errors, but can also correct many. That system should also generate an interupt, so that the OS knows there's bad memory and can notify the admin. -- Use OpenOffice.org <http://www.openoffice.org> -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
