The statement about a "quality" 600w power supply outperforming the Mac Pro's 1000w power supply is wayyyy baseless.
Actually, it's not baseless.
I worked backwards some time ago on it, and the numbers didn't lie. It's efficiency is 70%, not the 80% or better you can buy, albeit not inexpensively.
This translates into how much current (voltage droop) is found on the rails at full load. So in reality, the 980W is a peak value, and it's nominal value is only ~700W.
If you look at some of the posts on graphics cards, we had a member try 2x 4890's (IIRC, not 4870's), and it killed the PSU.
This shouldn't have happened on 980W nominal. So this supports the fact it's a peak value. (4890's use ~300W at idle each, according to online test results). Granted, how the rails are divided and loaded, have bearing on the specifics. But the PSU in the MP's are a single rail (according to a post on the unit's label), so that eliminates any load splitting as the possible source of the failure (too much current drawn off any single rail).
Unfortunately, PSU's are one of the areas the biggest shortcuts (i.e. cost cutting) takes place. Too small of heatsinks, not enough capacitance are the primary areas, but even protection circuits may be reduced or eliminated. The end result is more often than not, the power ratings are peak, not nominal (sustained). As a general rule of thumb, if the power rating isn't listed as nominal, it's the peak value. But peak values look better on the box, and are a marketing trick.
Peak values BTW, are only good for a very short period of time (well under a second), and meant to be applicable for startup loads that only last such a brief period of time (cold boot).
You are assuming that Apple uses a crap quality PSU to compare against. For the price you pay and how much work Apple has put into developing this machine, you better believe their power supplies are quality.
This is an assumption as well, and it's a bad one IMO. I've been involved in too many design projects (electronics, specifically computers), and have witnessed the managers set the directives as lowest possible cost over quality many many times. Hence the knowledge of what happens in PSU's. They get bids from a very basic set of specs, and go with the cheapest.
What's worse, is if you actually load test what gets used, they often don't even meet specs.
Particularly the basic ATX PSU standards (exceeding 10% regulation on the voltages under full load). 10% is sloppy BTW, and was set to allow for lower production costs to begin with. But that isn't stopping component vendors from making further shortcuts to increase their margins, even if it doesn't truly makes the grade.
The lack of proper Quality Control has a great deal to do with why/how this happens time and time again. (No testing of incoming parts prior to use in the assembly line).
With dies getting smaller and smaller, Voltage droop is becoming a bigger problem. It may end up pushing the regulation to 1% in the not too distant future, and then we'll see a spike in PSU costs. Perhaps double current prices on units that actually deliver on the specs. That is, 1kW = nominal, not peak.
