When it comes to testing a power supply, there are two courses to travel. One takes you down a path using a device to stress out the PSU and provide data regarding the power levels across all three rails. The second, and the one I make use of, utilizes an actual test system to give a more real-world account of what the power supply is capable of. While both methods have their merits, I prefer to use an actual computer to more closely resemble the manner of use that you, the potential customer, will put the product through.
That said, let's take a quick look at the test system. I have continued to beef up the system to put a more realistic strain on the power supply.
MSI X48C Platinum motherboard (Supplied by MSI)
Intel Core 2 Duo E6600 processor
2x 1024MB Corsair XMS2-8500-C5 memory (Supplied by Corsair)
2x Sapphire X2900XT graphics boards in CrossfireX configuration (Supplied by Sapphire)
Western Digital 250GB SATA hard drive
2x Western Digital 160GB SATA hard drives
Western Digital 80GB hard drive
Sony 52x CD-ROM optical drive
Samsung 16x DVD-R optical drive
1x 200mm fan
2x 120mm fans
While this isn't a Quad-GPU setup, we are certainly in the realm of having a system that is going to put a significant power drain on any power supply. Each of the X2900XT boards is capable of pulling close to 250 watts of power. Testing will consist of checking the power levels across all three rails at idle and again while the system is under stress. This should give us a good look at the capabilities of the power supply being tested.
While testing under a load, I expected to see the 12v rail drop off a bit. Considering that the test box is running a pair of very hungry boards, I was surprised to see only a small dip in power across this rail. The small rise in the 5v rail isn't totally unexpected since this system primary uses a 12v rail and has limited need for the 5v.
It should also be noted that all tests were conducted while the power supply was set to "Normal" mode, which utilizes six independent 12v rails. Even while trying to overload the system with multiple processes being run in the background and a hefty free-for-all in UT3 taking place in the foreground, I never once noted any problems with stability. For most users, the "Normal" mode will work just fine. For those who may want to run something along the lines of a Peltier based cooling setup without a separate power supply in place, the "Turbo" mode may be more useful.