On Sunday 17 June 2001 01:09 am, David E. Fox wrote:
>
>�> that doesn't work for you, it's a hardware problem. Still, AMD
>�> isn't going to fully support temperature monitoring till the Athlon
>�> 4's come out
>
>�Isn't that more of a mainboard issue than a CPU one?
� �Not practically. �While decent motherboards socket A boards have
monitoring chips/buses, the most important value .. the cpu core temp
is bogus. Temps garnered from a thermistor (probe), no matter how well
placed on the die, are a guess at best. Like tryin to measure the
temperature of the wiring inside a wall, by pressing a thermometer
against the sheetrock.
� �With the cpu's core temp jumping 10 to 20�C in a heartbeat under
instant load, a probe temp of 50�C from a Tbird or Duron is useless. �
Even AMD docs now caution to add between 10 to 20�C to the reported
probe temp under load to best _approximate_ the actual core temp. Even
with this fudge factor, spikes in the core temp will not be seen.
Overclockers have long known / used this rule of thumb.
� �Current AMD cpu motherboards are ready, have been for some time, and
the Athlon4's will have an internal diode (like Pentium processors
have had all along) to accurately and usefully monitor the core temp. �
The only way to sort'a kind'a accurately tell now if an AMD cpu is
experiencing intolerable core temp spikes, is if the system randomly
freezes or reboots (assuming everything else is in order).
� �Until AMD implements accurate core monitoring, the best fix if
you're concerned about cpu overheating is to underclock. Reduce the FSB
about 5% or reduce the multiplier by .5 � Of course, case temp should
be kept at or near room temp, and a better than AMD 'retail' heatsink
and fan with thermal grease (not pad) should be used. Another trick is
to try an' reduce Vcore a little, no more than ~5% and increase IO
voltage up to ~12%. 'Course now this is a motherboard issue, you havt'a
have a decent enough one to support doin' all this ;>
--
Tom Brinkman [EMAIL PROTECTED] Galveston Bay
