Verger wrote:
> Hi,
>
> I was wondering if anyone can give me some advice on the maximum in
> overclocking a P4 at 1700 mhz? I run it at over 1800 mhz now but I
> wonder if there is some generally recognised limit before you damage
> the thing?
>
You can look up processor overclocks here.
http://www.cpudatabase.com/index.cfm?Action=search
Some processor families have had observable failure conditions.
Note that the entries in the cpudatabase are not always good for
the long term health of the processor. (And some entries in the
database are pure nonsense.) Each processor has a max Vcore listed
in the processor datasheet (which you can download), and some are
more sensitive than others to abusing that limit. For example,
a Tualatin (nominal 1.5V) would fry after about a month at 1.8V.
Some people have lost AthlonFX processors on S939, and the cause
is believed to be related to the difference between Vcore and Vdimm.
(Also, some Athlon64 processors display what seems to be electromigration
failure, a gradual loss of overclockability.) So when you start an
overclocking project, it is a good idea to review the available
results from other enthusiasts.
"Sudden Overclocked Northwood Death Syndrome"
http://www.xbitlabs.com/news/cpu/display/news6375.html
For the really hot processors, sometimes the limits can be
thermal (you can no longer cool the processor well enough
or you fry the Vcore regulator). For example, the Pentium D 805
draws a tremendous amount of power if you run it at 4GHz (a
couple hundred watts). Rather than the processor failing, it
might be the motherboard that dies first. Some D 805 projects
got so hot, they melted foam placed underneath the processor
board, in the Vcore regulator area, while doing overclocks
on a table.
For my overclock on the computer I'm typing this on, my results
were rather obvious. Part of the frequency versus voltage
curve was rather gradual, and then I reached a point where
I could raise Vcore right to the datasheet limit (a "wall" so to speak),
and it wouldn't go a single hertz faster. That made the decision
as to what the operating point should be, rather obvious. No
point in cooking the thing, using more Vcore than really helps.
Other processors will tempt you with a more gradual curve, making
it harder to decide how much is too much. To make your own curve,
increase the CPU clock in steps of 5MHz, and only increase Vcore
if you cannot reach some stability criterion (some amount of Prime95
torture test time). Plot the points on a piece of paper, to see
the trend line.
Paul
>> Stay informed about: P4 1700 Max OC