That is a nasty hit. absolutely. It comes from drawing massive current at only the peak of the incoming sine-wave. Don't take my word for it, do you know who actually came up with the ATX 12V 2.x spec?
Google.
The primary difference with the 2.x spec for computer power-supply, is that is requires harmonic filtering on the input, which incidentally provides an internal network that reduces the harmonic energy transmitted to the source, and brings the power factor closer to unity (never unity on a switching supply until we discover alternate physics). A great deal of the losses in switching supplies operating at non-capacity loads, is that the front-end must continuously keep the driver section pumped to deliver the load-sections nearly instantaneous demands. If the load section is lightly loaded, the mid section is still responsible to deliver instantaneous current to maintin output voltage. Hence, the mid section ( actually, the high-voltage up-converter) must dump it's extra capacity somewhere. Sort of like chopper's on an AC Drive's DC buss.
Even as I recall, it wasn't more than 10 years ago, that switching supplies weren't used in industry, simply because they absolutely, positively required a guaranteed minimum load on the primary output. Without that minimum load, the supplies would oscillate wildly in voltage output, and would destroy the connected equipment. Today, switching supplies are commonplace, but they still suffer the problem of regulating the output... and they still require a minimum load. The minimum load requirement has dropped significantly, but at the price of switching the minimum output load to an internal section of the power supply.
Really, a modern switching power supply is an awful lot like an AC Vector drive, there is an input section, which generally rectifies the incoming voltage and stores the energy in a very (very) small capacitor bank. Then there is an inverter section, which takes the available DC 'buss' voltage and turns it into a relatively high frequency, high voltage AC (square) waveform. Then the converter section (not a part of an AC drive) rectifies the (now very high voltage, low current) mid sections waveform into very high, but short duration pulses, that are filtered before sending to the attached load. There is a finite frequency that can be filtered and regulated, and transformed (via L-C filters) to present a suitable voltage to the load. Now, where does the excess power go? The regulators charge-pump (to the filter) has a limited bandwidth (as do the filters). In an ideal world, the demand on the incoming supply would reduce... but, that would result in the inverter section starving to maintain it's available charge (for a peak demand) These aren't very big capacitors after all. The inverter must always maintain enough charge to keep the converter satisfied at maximum demand (Again, think of a PC, which might go from an idle state of 5VDC (or nowadays less, 3.3VDC, which is regulated down even further to the core voltage of 1.1 to 2.8VDC)) that is only using, oh, say, 30 watts (6 amps).... You wake the PC up from sleep, the CPU doesn't gradually wake up, it fires full force, so now, instead of 30 watts, it demands (in one cycle) 160 watts. If the mid section isn't sufficiently charged to deliver that power, you PC faults out. So, switching supplies will internally shunt off power (at the mid (high-Frequency) stage. Where does the reserve power go? burned off in those big resistors mounted on the heat sinks.
Switching supplies are inherently more efficient than linear supplies, but at a cost. The linear supply is burning almost no energy at no load, but extremely high energy at high load (because a linear supply must essentially shunt current to ground to maintain it's output). A switching supply is almost exactly the inverse. It must shunt (near full load) current when lightly loaded, and virtually none when operating at capacity.
--- Edit ---
Somehow, I expect flames on this one.
kamenges said:
While I can't directly with argue the previous two posts, largely because they are right, this would mean that the harmonic content goes up so high that the input power actually triples as the load goes to zero.
Let's momentarily assume that as rdrast said the power supply is rated at full load, which makes perfect sense. At that point, as Leadfoot calculated, we would expect about 1.2A in at 115 Vac to meet the published specs on the sheet. But as Cryogen posted initially, he has a line on the power supply itself that states the input current at up to 120 Vac is 3.3A. Based on the above posts, this would mean that the harmonic content increases to such a level at light loads that we burn up twice the peak real work-producing power in losses. That's a nasy hit, if that's the case. If that really isn't the case then the 3.3A value listed on the power supply is wrong or the peak efficiency published on the sheet is a bit high.
Keith