There is no such thing as "butting in" on Usenet B^)

Overclocking depends on tradeoffs. Intel builds CPU's with a very large
performance margin to insure reliability. The official Intel limiting
operating temperature for most Intel CPU's is between 70 and 80 degrees C.
Now, if we are willing to monitor our systems carefully, and accept the
possibility of a few glitches, then we can overclock.

When a CPU is run at higher speeds, the transitions between high and low
logic levels get a bit uglier, and the possibility that a high may be
interpreted as a low (or vice versa) increases. There generally is a lot of
extra separation built in, so we can usually overclock 20% or more with no
other changes. Once instability begins to occur, then additional margin can
be gained by increasing the CPU core voltage by a small percentage
(hopefully, less than 10%.) This increase means the separation between the
logic levels will be higher, and that more current will be drawn by the
transistor gates as they switch. The higher current makes the switch faster
and more reliable. Increasing the speed increases the average current draw
(more switchings per second) and thus produces more heat. Increasing the
core voltage increases the base line current draw AND the current draw for
each gate as it switches, and thus ALSO produces more heat. More heat
increases the noise in the chip, which adds to the ugliness of the
transitions between the logic levels. Thus keeping the chip cool has a dual
rule... keeping the transitions as clean as possible, and keeping the CPU
temperature below the maximum rating. The heatsink/fan supplied with boxed,
retail Intel CPU's is quite good, and usually enough for a nice overclock.
Extreme overclocking requires better cooling, and heroic cooling (chilled
water forced convection or even cryogenic) can get even higher overclocks
(of course, by that time, it is cheaper just to get a faster rated CPU in
the first place.

Some individual CPU's are better overclockers than others.

Some types of CPU's are better overclockers than others.

Some steppings of CPU's (modification to the circuitry on the die at the
factory) are better overclockers than others.

The Intel Celeron 300a could easily do a 50% overclock, sometimes without
either increased core voltage or increased cooling.

The Intel Pentium III 1 GHz would barely do a 33% overclock, even with the
best of efforts (it was near the limit for that CPU core type.)

The Intel Pentium 4 1.6A GHz Northwood would easily do a 50% overclock
(albeit with a bit of voltage and cooling help.)

The Intel Pentium 4 1.6 non-Northwood would barely break 1.8 GHz (it was
near the limit for that CPU core type.

Also, the operating conditions of the other components have an effect on
stability. One reason the Celeron 300a is so easy to overclock is that ALL
the other components could be operated at stock speed. This is not
necessarily true for the Pentium 4. Parts of the chipset must be
overclocked. The memory may need to be overclocked. Higher power demands
put more stress on the motherboard and especially the on-motherboard DC/DC
convertor/regulators that power the CPU, chipset, and memory.

A short answer... your CPU temperature of 66 or 68 degrees C seems too high,
and an indication that something may be wrong. More diagnostic information
is necessary to give a helpful answer:
ambient room air temperture
motherboard temperature
CPU temperature at idle (if Windows 2000 or later is installed)
the operating system
the memory rated speed
....

It may not be the heat alone, but just that the higher temperature is the
final straw.

Hope this helps. Overclocking is just basically screwing around until
something works. The only killer is raising the core voltage
unreasonably... that can kill a CPU immediately. Heat death does not really
happen to Intel CPU's... they are too well protected by the internal
thermtrip diode that shuts the CPU down when the temperature gets somewhere
around 150 degrees C (and the CPU will have stopped cycling and cooled down
way before that point is reached.) The Prescott will throttle down even
before the rated maximum operating temperature is reached, willy-nilly.




--
Phil Weldon, pweldonatmindjumpdotcom
For communication,
replace "at" with the 'at sign'
replace "mindjump" with "mindspring."
replace "dot" with "."

"Mr Jones" wrote in message
...
Hate to butt in here, but I was just reading this thread and the original
poster has my same overclock. I upped mine from 133 to 140. Anything
higher
than 142 causes a rounding error in Prime 95. I notice at 144 with voltage
upped by about 3-4% it crashed in 30 minutes twice. Then I doubled the
increase in the voltage and it crashed in about 19 minutes three times. Of
course the second test got the chip hotter, but my readings never show it
going over 68c at the hottest point. And the first test temps were only up
to 66c.

What I'm getting to is, does anyone here think at that temp I could be
throwing an error in Prime95 due to heat alone? And in the same vein,
maybe
that's pretty hot anyway for that small of an amount of oc and a better
heat
sink is in order. I'll drop the dough if it means I can oc higher and stay
stable.

Tim



"Phil Weldon" wrote in message
ink.net...
I think you have your monitoring program set incorrectly. Though with
an
operating system like Windows 2000 or newer, the CPU temperature is
likely
to be lower than the system temperature when idle, this is NOT the case
when
the CPU is under heavy load. If you have good CPU cooling, 46 degrees C
is
not an unreasonable CPU temperature while running SETI (SETI, even
though
the CPU usage shows 99%, does not use as much power as programs written
especially to max out heat generation.) A temperature of 42 degrees C
is
a
bit high for a system temperature (actually the temperature at a spot on
the
motherboard), better air circulation within the case might bring that
down,
but if ambient room air is blown directly into the fan of the CPU
heatsink
(and the CPU heatsink fan blow into the heatsink), the motherboard
temperature has little effect on the CPU temperature. At any rate, your
overclock is VERY mild, and since you don't mention increasing the CPU
core
voltage, I doubt the CPU heat production is up more than (140-133)/133 =
5
%. Really, extra cooling for overclocking is for stability, not safety,
when using Intel CPU's since the Pentium 60/66. Your system will show
instability long before the CPU temperature exceeds specifications for
very
long, and the Pentium 4 will automatically shut down long before damage
from
heat occurs. Raising the core voltage is a different matter. If you
raise
the CPU core voltage by an unreasonable amount, it will kill the CPU.
Raising the CPU core voltage by a reasonable amount will generate more
heat,
but that will won't damage the CPU, though additional CPU cooling may be
necessary.

--
Phil Weldon, pweldonatmindjumpdotcom
For communication,
replace "at" with the 'at sign'
replace "mindjump" with "mindspring."
replace "dot" with "."

"David Hansen" wrote in message
...
Hi,

I have a P4 2.4GHz, with a 533fsb. I have upped the clock-speed from
133
to
140 in the BIOS, and the system seem to be running nicely.

I was wandering at what sort for temperatures (CPU & System) should I
start
to get nervous?

I currently have 46°C for the system and 42°C for the CPU, while
running
the
SETI program (99% CPU shown in Task Master).

I am using Easy Tune 4 from Gigabyte, to monitor the temperatures,
though
I
don't know where the system temperature sensor is (the Motherboard is
a
Gigabyte 8ST800).

Thanks,

David