CPU scaling vs Temperature

[John Moser]

On 03/06/2011 12:10 PM, Phillip Susi wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 03/06/2011 11:24 AM, John Moser wrote:
> You've got a hardware problem.  The system should never overheat under
> load if the fans are working properly.
>

Cool, I'll switch to ext2 which is faster, and delete e2fsck; anyway, 
proper data integrity and continuous mains power are up to my kernel and 
power company, and these things should never fail if the the utilities 
and the Linux kernel developers aren't incompetent.

I should inform my employer of how much money we can save getting rid of 
those silly things called "Back-ups."  There's no reason for fail-safes 
anyway, since all things that could be protected are fine and the 
problem is really somewhere else.


For those whom sarcasm is lost on:  I'm fully aware that this is a 
hardware problem.  It's nice to detect problems and somehow prevent data 
loss, rather than gleefully deciding that systems should not fill up 
with dust or experience fan controller failures (this happens quite a 
lot, actually), or just decide that the BIOS should save us from 
whatever happens (it does, by shutting the computer COMPLETELY off 
without warning; it does not nicely tell the OS to suspend to RAM or 
otherwise take immediate throttling steps to reduce system load and cool 
things off).

There was talk of using SMART tools to monitor hard drives a few years 
back, the idea being to pop up a balloon that tells the user, "Your 
drive has been reporting problems lately.  It may be about to physically 
fail; you should back up your data and be ready to replace your hard 
drive with a new one."  I had always assumed that SMART was snake-oil 
and gave more false positives and false negatives than actually helpful 
information, so the whole concept of using SMART to determine that the 
drive will "probably fail soon" was infeasible.  Now I wonder if people 
just decided, "Well you should replace your hard drive every 2 years 
anyway, and have back-ups, so this is a wasted effort."  Or maybe SMART 
doesn't exist in SATA or something.