We compare 46 power supplies in total, of which 20 are specially tested for this article. We tried to gather power supplies from all segments to give a complete as possible overview of the market.
For the tests we use professional Stratron load generators which we used to stress the power supplies in our lab to up to 1600W. Each power supply unit was tested in increments of 100W starting at 200W up to the maximum capacity. At each increment the voltage on various rails was measured. The closer to the official values of 3.3, 5 and 12V the better, while a deviation of more than 5% is unacceptable. We used a professional Zes Zimmer power analyser to measure the power consumed from the power outlet. Based on that we calculated the efficiency. We used an oscilloscope to measure the ripple or the fluctuations of the output direct current.
It is important to know that in all cases we get about 50 watts out of the 3.3 and 5V rails and the rest comes from the 12V rail. This is not the way the 80 Plus organisation and most manufacturers test the power supply units but it is closer to the reality. In a PC the 3.3 and the 5V rails have very limited usage as the power thirsty components (such as the CPU and GPU) only use the 12V rail.
Along with the standard tests we have also tested the units at low loads of 22.5, 50 and 100 watts. These values and the distribution on the 3.3, 5 and 12V rails are based on a real system set up of one Intel Core i7 4770K Haswell-processor, one Gigabyte GA-Z87X-UD3H motherboard, two Corsair Vengeance Pro memory modules, one OCZ Vertex 4 SSD and one CPU-cooler. The 22.5W test comes very close to the idle usage values of the above-mentioned system while the 50W test is based on the situation when one CPU core is functioning fully and the 100W test is based on both the CPU cores being used to the limit.
Furthermore we measured the current drawn by the power supply from the power outlet without any output. It is important to know if the PSU uses up too much power in standby mode.
Our sound tests were conducted in a sound-proof box in order to be able to accurately measure values of as low as 18dB(A). We placed the meter at the distance of 10cm from the unit and measured the noise at 100W, 300W and 500W after the unit has been running each wattage for 20 minutes.
Alongside the tests we look at the practical aspects: the amount of connectors and the length of the cables. Due the large number of power supplies in this test, we’re not able to discuss them individually. What we can do is highlight the test results that stand out and subsequently name the best choices.
Because the large amount of test results, we have made a selection for this article. You can find the complete test results here. In the charts on the following pages you can recognize the (semi-)modular power supplies by the blue bars and the non-modular variants by the red bars.
A controversial subject with power supplies is the origin of the used capacitors. Many do not believe that all non-Japanese capacitors are necessarily bad. So far we have not found any convincing proof for this. There is no reason to doubt that a higher positioned premium brand power supply has good capacitors, whatever its origin is. The long warranty period of some Chinese capacitors however, is a strong indicator that the manufacturers certainly have trust in the quality. Of course this does not mean that there are no cheap power supplies with bad capacitors, but the quality is not determined by its origin.