On 13/03/13 02:35 PM, Nick Lidakis wrote:
On Wed, Mar 13, 2013 at 08:11:36AM -0700, Kelly Clowers wrote:
On Tue, Mar 12, 2013 at 11:55 PM, Long Wind<longwind2...@gmail.com> wrote:
linux is stable, or is it?
Sure, but there is difference between stable and "never crashes,
ever". Namely one is a real thing, and the other does not and cannot
exist, it is purely theoretical.
Even with a provably correct program (and we cannot prove anything so
complex as a full featured OS kernel, much less desktop, etc),
hardware could fail in one way or another... a cosmic ray could hit a
memory cell or a cpu transistor, and it could crash. You can add error
handling to try to deal with all sorts of errors but you can't catch
everything, and the error handling code has bugs of its own.
Not to hijack the thread, but I was thinking about this and desktop systems.
That is, whether folks use ECC RAM for their desktops. Also, how many crashes
might be attributed to RAM errors. I can't find the link, but I once came a
across a Google study that claimed it happens a lot more often than not.
Though, one would need to buy or currently be using a motherboard that
supports ECC. I generally use Tyan boards and most have supported ECC RAM.
Personally, I don't see the cost as prohibitive as I only upgrade maybe once
every 5-8 years.
Nick
No,, your points are valid. However ECC RAM is rarely used because RAM
can be very reliable.
Unfortunately I've noticed a lot of boards lately that don't work well
under stress. While memest86 may not find errors, in real use using the
disk and accessing memory at the same time can cause errors that won't
be detected by simple tests.
The problem could be cured in the BIOS by manually setting the RAM to a
slower speed but the real fix should be for motherboard manufacturers to
stop assuming that they can push RAM to a limit that their boards can't
handle.
Again, this is NOT a RAM problem - it works fine according to memtest86.
It is a system problem where the motherboard can't do two things at once
fast enough to meet its own hardware settings.
ECC RAM probably won't fix this issue.
--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/5140d184.70...@rogers.com
