Petz, Re: With 68w Athlon 64 vs. Prescott power difference in idle mode x 24 hours, that's 1.6KWh/day or $10/month per PC in Southern California at homeowner marginal electric rates
You shouldn't trust any results that claim that Prescott's idle power is 68W higher than Athlon 64. That is just so beyond credible that it's silly. With the presence of a simple sleep state (one that shuts off all internal clocks after the OS passes the HALT instruction and there are no snoops present on the FSB), the idle power will drop to the size of the leakage power. Dothan's sleep power at 2.0GHz is rated at a max of 10.5W*, which assumes the leakiest chips in the batch (most individual chips will be smaller than this, but as far as spec sheets go, you always have to list the worst case). Prescott has a lot more logic transistors, but it's on the same process, so the sleep state power should scale with the ratio of logic transistors that Prescott has over Dothan.
Does anyone have any idea what this is? It should be pretty easy to estimate Prescott's idle power dissipation if this information is known. Somehow, I don't think it's in the neighborhood of 10x, which seems the amount necessary to get a 68W difference between Prescott and Athlon 64.
* By the way, I looked up the sleep power for Banias, and it's 7W, which suggests that Dothan is a leakier chip. Unfortunately, I don't have any information on how much power is leaked by the additional SRAM cells (Dothan has 2x Banias). Even though SRAM leaks less than logic transistors, an extra megabyte worth is non-trivial.
** Another last minute thought: the sleep power tells us that about 50% of Dothan's power comes from leakage. If there were a process that could eliminate leakage totally, then Dothan would be a about a 10W chip.
