For testing power supplies we utilise professional Stratron load generators, which allows us to put a load of up to 1600 watts on PSUs in our test lab. Each PSU we test in increments of 100 watts up to the maximum load. At each step we measure voltages on the different lines. The closer those measurements are to the official values of 3.3, 5 and 12V, the better. More than 5 percent deviation means not good. At each step we also measure the current from the socket with a professional Zes Zimmer ammeter. Based on this we calculate the efficiency. An oscilloscope allows us to measure the ripple, which are fluctuations in the direct current (dc) output of a power supply which has been derived from an alternating current (ac) source.
It should be noted that in all cases we draw 50 watts of current from the 3.3 and 5V lines, and the rest from the 12V lines. This is different than how the 80 Plus initiative or most manufacturers themselves test their power supplies, but is closer to real-world performance. In a PC all energy-consuming components (mainly the CPU and GPU) only utilise the 12V lines, and the 3.3 and 5V lines have a very limited role.
In addition to measurements with load, we also measure the current drawn from the socket without any load. It is important to know how much leakage current a power supply has.
We have revised the way we measure noise levels. We test power supplies in a sound-proof box, so we can accurately register levels as low as 18 dB(A). We measure the sound level from a distance of 10 cm cm from the PSU, and we put 100W, 300W and 500W loads on it.
Good news! During our previous big round-up of budget power supplies a pretty significant number didn't make it to the end. This time around all 39 PSUs survived. We should note that the manufacturers that were responsible for the failing PSUs last time did not send us any this time, or they stopped making them.