Now to the infamous load tests. Load tests are performed on an industry ATX SMPS ATE, or ATX Switch Mode Power Supply Automated Testing Equipment. Specifically, a SunMoon SM-8800 which can put as many as 10 different loads on eight different connectors at a time.

     Each load is held for 15 minutes before results are taken. "Watts" is actually the DC output of the power supply. Efficiency is calculated by dividing the "Watts" reading from what is being pulled from the wall.

     Temperatures are taken using a Type-K Thermometer with a probe in the intake and a probe inside the exhaust vent of the power supply.

     We can see that the 12V rails did not drop much at all. Our average 12V voltage was 12.255V during load test 1. During load test 6, the average 12V voltage was 12.1475. That means there was only a drop of .1075V. That is EXCEPTIONAL considering the load on each 12V rail went from 2A each (total of 8A) to 11.5A average (total of 46A.)

     Efficiency was also very good and was at or above 80% at what would likely be the most common load. We can see active power factor correction working with a near perfect .99 reading, and the exhaust air stayed fairly cool exhibiting only a 14 degree delta even while putting out 696.5W.

     Now we're going to run the power supply in the "hot box." All the "hot box" is is a case with the side panel on it. The CPU vent in the side panel has a vent hose coming from the exhaust fan of the load tester. Since all of the power the power supply puts out is converted to heat, the load tester has a lot of hot air to exhaust. I simply pump this air back into the case to increase ambient temperatures.

     Things were getting a little warm during test six, but nothing alarming. The delta was only 8 degrees, so obviously the power supply's fan was doing an excellent job of moving air, and the noise from the fan was never really noticeable.

     There was only a slight drop in voltage from the heat, 3 degrees at the most on one of the 12V rails, but everything seemed to hold it's ground just fine.

     After all of the tests were done, I then maintained the 11.5A average load (11, 11, 12, 12, as shown in the load test tables) on the 12V rails while dropping the 3.3V down to only 3A and the 5V to only 4A. Despite this, the 12V rails still maintained with a respectable 11.79, 11.76, 11.81 and 11.77V.

     Next, we take a look at the oscilloscope readings.

     While the power supply is subjected to the load tests, tiny fluctuations in voltage (called ripple) are output by the load tester. This output is then fed into an oscilloscope where they are converted into a legible waveform. Let's see how the OCZ GameXstream 700W did here...

OCZ GameXstream 700W
Test 1	+3.3V	+5V	+12V1
	+12V2	+12V3	+12V4
Test 2	+3.3V	+5V	+12V1
	+12V2	+12V3	+12V4
Test 3	+3.3V	+5V	+12V1
	+12V2	+12V3	+12V4
Test 4	+3.3V	+5V	+12V1
	+12V2	+12V3	+12V4
Test 5	+3.3V	+5V	+12V1
	+12V2	+12V3	+12V4
Test 6	+3.3V	+5V	+12V1
	+12V2	+12V3	+12V4

     Things were satisfactory until test 3. During test 3, I started to see ripple on 12V1 and 12V2 that was nearly 100mV. By test 5, the ripple was 100mV on 12V3 and 12V4 and had exceeded 100mV on 12V1 and 12V2. By test 6, the ripple on 12V1 and 12V2 had exceeded 150mV.

     ATX specification has an "allowance" for 120mV. Mind you, test 6 was very stressful, pushing the power supply to it's limits, but it doesn't change the fact that the ripple exceeded spec and this is the first power supply to date I have had do this.

    What's wrong with too much ripple? Well, if the voltage fluctuates too erratically, it can over work other regulators used to provide Vcore to the CPU, GPU, etc.

    A representative of OCZ did find my findings concerning and asked that I return the unit for an exchange. I did so, and in the interim I obtained an FSP Epsilon 600W. The 600W version of the power supply the OCZ GameXstream is based on. In testing that unit, I found a good deal of ripple as well. When a brand new GameXstream was finally returned to me, I immediately plugged it in and found the same results as the first sample.

    Fortunately, it is very unlikely any of us will have one of these power supplies at or above 400W, at least for any considerable period of time, but this high ripple measurement will be taken into consideration when calculating the performance score.