TEST BED

TESTING METHODOLOGY

     Using a dedicated measurement instrument such as a Chroma or a SunMoon to test power supply units is without doubt the most ideal and accurate way (not to mention the fastest) to do that currently. However, it's certainly not the only way there is and so pretty much anyone can test a power supply unit just by using a test rig. Certainly, limitations do apply and so you can't really push a 1000W power supply to its limits if your system only uses 500W at peak loads and that's why over the years we saved certain hardware components for the purpose of building a dedicated PSU test rig. True it may not be as accurate as the above mentioned solutions, but it comes really close and is in fact much closer to real world usage. So as always, we ran several games with maximum graphic options enabled at a resolution of 2560x1600 in order to stress every hardware component and increase the overall power demands of the system. The Passmark BurnIn Test was also used to overstress the components in an effort to provide the most accurate results possible. As a final test we also used the latest OCCT 4.4 software and its dedicated PSU testing suite since it can really bring a power supply to its knees after inside a few minutes.

   Rail stability was checked/measured with the CPUID Hardware monitor and a Metex multimeter which also recorded the system load in idle and in load. As always try to remember that the power consumption numbers listed in the graph are the highest (Peak) ones recorded during the entire duration of the tests and not the average ones. Noise levels coming from the fan were recorded using the high precision HD600 ExTech Sound dBA Meter from the rear of the unit and at a range of no more than 5-10cm. Readings under load are recorded the exact moment we manually switch the fans of all graphics cards from full speed to almost zero, that way the fan of the power supply does not have enough time to slow its RPM and so by doing this we get very accurate noise level readings. Needless to say, in order to get 100% accurate readings, you need to have a noise isolated room for that exact purpose, something which is quite impossible unless you are working inside a real lab (some people use very small noise insulated boxes but due to their size both heat and noise exceed normal levels and so the results can't really be considered to be 100% accurate, nor realistic for that matter). Also do take into account that since all noise measurements take place from just 5-10cm away the final noise levels to reach your ears will be considerably less.

* After well over 10 years of testing PSUs the Intel Core i7-920 CPU of this rig failed and so we replaced it with a Xeon X5660 (we also swapped the GA-X58A-UD7 for the G1. Assassin).