On Thu, 17 Aug 2000, you wrote:
> Hmmmm... now tell me - how can a program make a CPU hotter ? By
> accessing more registers or being L1 and L2 cache intensive or what ?
> Because until today I considered what you said being science-fiction.
>
> Roman
If you visit the numerous hardware and overclocking sites on the
Net, the first issue you'll seen again and again is the revelevance
of hardware load and temps to performance and stability. As I said
below, install some monitoring software (it's supported by all but
the junkiest of hardware), and see for yourself. It will be
immediately apparent, I promise. If your hardware won't support
monitoring, just read lm_sensors docs.
The relationship between system load and temps is very basic and
prob'ly gettin OT for this list. A detailed explanation certainly
is. Still tho, I believe it's important that those who might be
unaware of Netscape's ld-linux.so2 gone wild problems, should also
be acquainted with it's implications involving hardware
failure/file system corruption.
--
~~ Tom Brinkman [EMAIL PROTECTED]
> Hey Tom,
> >> > You're hardware is at risk when ld-linux.so.2 goes awry. It will
> >> > cause your cpu's internal core to increase significantly. Left
> >> > 'untreated', and if your system cooling is marginal or inadequate,
> >> > harware failure is risked. I wouldn't let the situation go till
> >> > you got home.
> >>
> >> I don't think many programs manage to change the physical measurements
> >> of a cpu's internal core. Of course, you probably mean that it heats
> >> the core up. If it would do that then you would be in a way right
> >> because of the greater oscillation diameter of molecules being exposed
> >> to more infrared radiation (heat). However, no program can heat the
> >> CPU more than it already is without the program. When a computer is up
> >> and running, it uses the CPU nonstop, no matter what program and for
> >> how long is running. Even if you would pause at the POST memory check,
> >> the CPU would keep executing the internal "check if he presses
> >> something so I can continue (and other things)" loop. The CPU is
> >> always doing something when it is powered. *What* it does doesn't
> >> matter. The power is on.
> >>
> >> Someone correct me if I am wrong, please.
> >> Roman
>
> > You're wrong about, if I read you right, the cpu being in more or
> > less constant state when the power's on. Install lm_sensor's or
> > Khealthcare and you'll instantly see what I mean. At idle to normal
> > load in an air cooled system, your cpu will (should) be at room
> > temp, eg, 22 to 26C. Under load, this will immediately rise to
> > (hopefully no more than) 30 to 35C. Take the load off, and the cpu
> > should immediately drop back to room temp. Under EXTREME load (like
> > with cpuburn), ideally the core temp should be kept under 45C. Intel
> > cpu's (>= Pentium) are the best to monitor because they have an
> > internal diode that reports to a (decent) motherboard's i2c/SMBus
> > the cpu core's temp. Using a probe (thermistor) is a within 10C
> > guess at best.
>
> > I have very adequate cooling for a p3-450 overclocked to 608mhz.
> > I have the cpu's core temp displayed constantly on Kpanel using
> > Khealthcare. At idle/normal load, it's room temp (~25C). During a
> > kernel compile it will reach 35 to 40C. Running cpuburn it will
> > quickly reach 43C and stay there till cpuburn is quit. When
> > Nutscrape goes wild (ld-linux.so.2) the temp will immediately jump
> > to ~40C. Using this temp gauge, I often know NS's fixin to crash,
> > before it does ;)
>
