The statements below are from this link:
http://www.audioholics.com/techtips/audioprinciples/amplifiers/ACDTEST5.php
Let’s take the following Scenario when comparing two similarly priced receivers.
Receiver #1: rated at 110wpc (# of channels driven not specified, but FTC mandates it must apply to at least two channels)
Receiver #2: rated at 70wpc ACD
If we bench test both receivers with just two channels driven, we more often than not find that Receiver #1 was able to comfortably exceed its rating into 8 ohms and deliver respectable power delivery into 4 ohms (usually higher than the 8 ohm rating). Receiver #2 will usually exceed its 8 ohm rating but typically at a smaller margin than receiver #1 and, more often than not, maintains a similar power rating for 8 and 4 ohm loads. Of course an ideal amplifier should act like a constant voltage source and double power delivery when speaker impedance is halved. Very few receivers, and only the better separates amps, have a robust enough amp topologies and correspondingly robust power supplies to achieve such a feat.
So, in reality, Receiver #1 has the potential to deliver more power than Receiver #2, especially into 4 ohm loads, when driving one or up to three channels simultaneously with a correlated audio source. But because of the thermal limitations of the budget sized receiver, which aims at designing a dynamic amplifier over one that can provide sustained power into all channels simultaneously, it has to limit the power delivery when the amplifier is taxed with a correlated audio source to more than three channels. During loud transients Receiver #1 will likely be able to deliver cleaner output because it has more available headroom than Receiver #2.
On the other hand, the ACD test can very easily boast over inflated power figures for amplifiers that don’t employ this form of current limiting. This is true in particular when the audio publications test in the 3 conditions we previously mentioned — instantaneous 1kHz power vs THD test, VARIAC regulated line, and bypassing the internal amplifier fuses. Thus, the consumer is misled into believing the inflated power measurements are continuous and achievable in typical household installations.
Since this test isn’t representative of music or movie program material, and it is usually conducted in conditions not representative of normal product usage, its validity is questionable at best. It is certainly not a clear delineation of true amplifier performance in a real world environment.