Originally posted by: GeneralOreo
Originally posted by: BonzaiDuck
So like our colleague above said -- " . . . . it saddens me. . . . "
I'm not sure I understand you or your colleague.
__________________________
A matter of "over-clocking philosophy." Intel has already tested and binned these processors, for the prevailing technology and revision (B3 vs G0).
As I've said in other posts, "random variation IS a form of Natural Order." Take for instance, Jack London's adventure stories. Ostensibly written for adolescents, they were written by a socialist who was intrigued by the concept of evolution at the time Darwin was just getting published. A pack of wolves are on an ice-floe, fighting over a dead seal. By accident, the Alpha Male, strongest and fittest in the pack, slips by accident on the ice, and gets torn apart by the other dogs. so in the history of evolution, you will find the path strewn with the corpses of the fittest, but in the long run, as the rule of averages prevails, the fittest as a statistical population prevail.
The CPUs, I said, were tested and binned. Intel states on the box "Maximum Voltage = 1.35V." Intel is a business, and they didnt' get to be dominant firm by making cavalier choices. They set that voltage spec because they've tested and determined that the number of RMA returns under warranty -- if run within that spec -- will be near 0%.
So there is a statistical distribution of failure as you increase that voltage. somewhere above that voltage is a voltage setting where CPUs ON AVERAGE -- WILL FAIL after so many months.
There are two aspects to failure. Heat -- a function of voltage(squared) and bus-speed -- degrades the circuits, especially at the interface between metal parts and silicon. Thus, the "thermal specification" -- and there is also another statistical probability distribution from the testing labs as you increase temperature, with various rates of failure.
The second aspect is the voltage and the phenomenon of electron migration over time, regardless -- nearly regardless -- of prevailing operating temperature. If you lower temperature as much as possible, the desire would be to over-clock an extra hundred MHz or two at the existing voltage, because you've lowered some aspects of resistance in the circuitry. But if you increase voltage anyway with the false comfort that temperatures may even be below room ambient (with TEC-chillers and phase-change), you STILL have electron migration.
So as to the "philosophies" -- we're either trying to achieve a Bonneville-Salt-Flats speed record, knowing that you're going to burn out or blow up your rocket-car getting there, or you're going to make a modest gamble toward a trade-off between component longevity and a healthy increase in speed and performance, knowing that this is about random variation in an industrial production process.
so I was being facetious -- even condescending -- about "sadness."
B3 stepping Q6600 -- expect a reasonable over-clock without undue voltage stress or heat stress at around 3.2 Ghz
G0 stepping Q6600 -- expect a reasonable over-clock without . . . . . . . . at around 3.4 Ghz.
G0 stepping of the C2D E6850 (!!! ) -- expect an easy over-clock to 3.6 Ghz, and probable OC as far as 4 Ghz.
The more cores you pack into the CPU, the more severe the limitations.