No power, no dice: it's really that simple. All components in a PC system need electricity, and ever-increasing amounts of it, so the power supply has to come up with the goods. Manufacturers can claim what they like, but the claims don't always translate into reality, and very few users have a good idea about the power drawn by the individual components.
This is why we test power supplies ourselves. To help you understand the procedure we use, we will now shed some light on the issues involved, and explain our testing procedure.
Performance
Performance is always at the top of the list of a power supply's features, but the trick is to find out the device's actual power output, and whether voltages lie within the specified values. Voltage tolerances at all load levels are described in the ATX12V Power Supply Design Guide. This document is the technical basis for all power supplies, and the standards it contains are used for all our measurements.

THG's power supply test platform in action
We employ a special platform to test loads on power supplies. At its core are four electronic high-performance loads, each of which can handle a maximum current of 50 amperes. This setup enables us to record measurements extremely accurately. Conventional adjustable resistors are used for the standby voltage and the rarely used -12V and -5V paths. A network filter ensures that any interference pulses from the network do not distort our measurements.
The Design Guide specifies three power classes for every power supply. The lowest range is "light load", a very low level defined as being 20% of full capacity; this might be representative of the idle condition when a PC displays the Windows desktop. At this load, the output voltages are measured immediately upon power up, and also 30 minutes later. This second measurement is needed because when the components inside heat up, the voltages often vary in tight ranges.
The second power class is when half of the total operating performance is requested; this is called "medium load". Here too, the voltages are measured immediately after the new power setting, and then again after a short time.
The third level is the "full load" test, which demands the total promised performance with all voltage rails loaded to maximum. This is where we separate the wheat from the chaff and force the power supply to show what it's really made of. The voltages are again measured immediately and after around 30 minutes.
Ripple voltage, defined as the ratio of alternating voltage to output voltage, is also measured, and an oscilloscope displays the voltage graphically. The fixed ranges are also listed in the Power Supply Design Guide.

The heart of the THG test platform: Electronic load from Statron

Each load can handle up to 50 amperes
While performance is the primary issue with a power supply, efficiency is also very important. When demand for power is rising, the efficiency of the test devices becomes particularly significant, so we rate efficiency even more highly on high-capacity units.
Why is it so important? First, because it represents energy cost savings. At an output of 600 watts, a power supply running at 80% efficiency would sap a whopping 750 watts from the network. At an efficiency rating of just 60%, the same output would eat up a kilowatt from the grid!
Besides the savings in electricity, there are other aspects to consider. The most notable is that the power that is not used to drive components is converted to heat by the circuitry. Thus, the lower the efficiency, the more waste heat the device produces, which in turn, translates into a raised need for cooling. Good cooling is only achieved with larger and faster fans, so a power supply with low efficiency is generally louder than one with high efficiency.
Efficiency can also affect the supply's useful life, too: the warmer the components get, the shorter their lifespan. This point is not necessarily a determining reason for buying a highly efficient power supply, but all things considered, it makes it a far more attractive purchase.
Leads And Plugs
Even when you've got performance and efficiency, that's not everything. There are many power-hungry components in a PC, and whether they actually get their energy when they need it depends on the number and length of plugs and leads available. After all, a power supply with 600 efficient watts is worth little if there are only four 5.25" connectors to distribute the power!
A 24-pin plug for ATX output is used in the new ATX12V 2.01 specification. Users with older mainboards using 20-pin ports will be interested in the matter of divisible connectors that can be cut down to 20 pins, and the availability of adapters. Thus, we also assess these issues in our evaluation.