[conspire] looking for good FAQ Websites for CPU heatsink/Fan hardware
rick at linuxmafia.com
Sat Apr 5 09:25:50 PDT 2008
Quoting Nick Moffitt (nick at zork.net):
> The real reason why BIOS memory checks will often pass when other things
> succeed is that they are most often run *at cold boot* time.
> This is one of the reasons why people insist on running memtest86 (or
> whatever it's called these days) for at *least* an hour before
> trusting a pass result.
I've seen moderately bad RAM slide by 1-2 hours of memtest86 / memtest86+
testing, but fail an _overnight_ run. People used to come to CABAL
meetings and attempt to triage their questionable RAM _during_ the
meetings, that way, and it just doesn't work well enough. On
reflection, as you say, this may be just because of heat loading having
not yet reached equilibrium -- but, regardless, I always advise that the
software needs to run overnight, before a "pass" result can be deemed
reasonably dependable. (Some subtle RAM flaws can still get past that
-- which is where you need "make -j" kernel compile loops, or Cerberus
CTCS, to flush them out.)
> Also the DL385 G2 systems have that great "ALL POWER TO THE FANS!" mode
> on boot. You'll never have dust problems with those babies!
Indeed, and ditto with some of the Dell 1Us. You wonder who slammed on
the airbrakes, and worry if we're almost out of runway or something.
More information about the conspire