I had none of the "possible" problems listed here. It was a 300 W Antec PC power supply on an Abit motherboard. The input voltage at the CPU power supply was NOT fluctuating. The output voltage was. For a CPU dissapating 70 watts with a 1.65 V core we are talking about over 40 amps. Modern CPUs can have large (more than 10 amps) current fluctuations that occur on the order of several CPU clock cycles. Even at "only" 900 MHz that is much faster than any switching power supply can compensate for. And at those currents, even a very small resistance can cause 50-100 mv drops. On Pentium IIs and I belive Pentium III's too, Intel had output pins that gave the CPU power supply a "heads up" when it needed to prepare for such a change. Large capacitors (usually 3 or more 1500 uF caps) on the power supply output can supply the current (only for a short time), but the larger the capacitance, the longer it takes (this may be why these fluctuations are "visible"). Hugh numbers of small bypass caps right on the CPU power pins help, but they simply can't supply enough current. I'm just telling you what I actually saw by actually putting a digital multimeter directly on the output of a CPU power supply. Try it for yourself.
|