 Re: Some questions on overclocking
On 10 Sep 2005 06:19:16 -0700, "Random Person"
<nonexistent2032@yahoo.co.uk> wrote:
>Hi guys. A couple of questions on overclocking...
>
>If you had to pick an AMD64 processor to overclock, which one would you
>take, and why? I thought an AMD64 939-pin 3000+ Venice core might be
>good.
>
>For that matter, why does a smaller manufacturing die (e.g. 0.9nm)
>translate to lower heats?
It doesn't always, but generally it is because it allows
using lower voltage.
>
>If you had to pick a processor from any class to overclock, which one
>would you take? I read that overclocked AMD Duron 1600 processors beat
>some AMD XP 3000+ processors...
Sure, if you were able to o'c the Duron enough, but
apples-to-apples comparison would be to then o'c the XP3000,
too... and still it depends on the specific use. For many
uses, the XP3000 at stock speed may still be faster,
particular real-world uses instead of synthetic benchmarks.
>
>Also, supposing you fry your processor through overclocking,
Don't do that. Overclocking is not some kind of "reckless"
activity, it is simply using a CPU that produces more
heat... a bit like running a P4 Prescott instead of an
Athlon 64. You have to have sufficient cooling for any
particular level of heat.
>how will
>AMD know about it to void your warranty?
They will probably replace it anyway because they're looking
at keeping goodwill towards customers, BUT it might be
considered illegal to void your warranty and then seek
replacement. Generally though, if you let your CPU run too
hot, it matters not if you overclocked it, you voided your
warranty. The one exception might be if you had the
AMD-recommended chassis configuration with (expected)
typical ambient room temps and then the AMD retail fan
itself failed, stopped spinning.
> Perhaps through evidence of
>non-stock thermal grease on the CPU?
Why would you buy a retail CPU for overclocking? To get the
significant increases you're considering (similar to o'c a
Duron till it's faster than an Xp3000), you need a better
heatsink. Overclocking is not just about raising a clock
speed, it's about having entire system that can accomodate
that raised clock speed in every way.
Yes, if you left 3rd party grease all over the CPU they
might noticed and void your warranty, but if you cleaned it
all off (for example, using detergent and a toothbrush under
running water) they'd probably not say a thing. Even so,
that is not something you should do, you take the warranty
into your own hands and void it by deviating from any and
every stock part of use... including 3rd party
compound/grease. They can only realistically warranty what
they know for certain to be a valid configuration.
>
>I also read that you stand a chance of corrupting data on your hard
>drive through overclocking your CPU. However, if your CPU runs
>Primetest without any errors,
Prime 95's Torture Test.
That checks for CPU errors alone.
>is it safe to assume that your HD data
>will not be corrupted at all?
Yes and No, it only "greatly" minimizes the chances that the
CPU itself is instable, it does nothing to ensure a higher
front side bus, motherboard bridge bus, or memory itself is
still stable. You'd need to check any and all things that
have had their operational speed changed.
Generally speaking though, if your bus speeds are still in
spec, or rather close still, AND Prime95's Torture Test runs
without errors, there should be no data loss from these
factors alone.
However, if the data integrity is of utmost importance, do
not overclock. If the sytem is mission-critical, do not
overclock. However, any system with these requirements
should be just as extensively tested as an overclocked
system would to be qualified for such uses, so ironically
enough an overclocker may have a more stable system than a
non-overclocker because they've taken atypical steps to
check system stability, and rectified any such problems.
This is only considering moderate o'c though, not extremes.
With extremes, it depends a lot on exactly what you're
doing.
>Related to the above question, as I was planning on overclocking my AMD
>XP 2500+ Barton soon - the data on my HD and the accuracy of my CPU's
>computations are critical to my work. Do you recommend not overclocking
>at all then?
Yes, not overclocking it is certainly the best course. The
odds are high that you could o'c to at least 2.1GHz and not
have problems, but I will not presume you want to play odds
with critical calcs or data. However, if the calcs and data
are really critical, a system using ECC memory is then the
best solution yet most do not have this type except in
servers.
 | 
Linear Mode
