How Much Power Can a Generic 500 W Power Supply Really Deliver?

Load Tests

We conducted several tests with this power supply with the equipment described in the article Hardware Secrets Power Supply Test Methodology.

Since we didn’t know beforehand what the real wattage of this power supply was we did something different from what we usually do when reviewing power supplies. We loaded this power supply with 50 W and then started increasing the load pattern in 25 W increments until we reached the maximum this power supply could deliver – i.e., until we burned it. We knew that we would burn this power supply for sure, we only didn’t know when.

Our generic 500 W power supply died when we tried pulling 275 W from it, so the maximum amount of power we could extract was 250 W – half the labeled amount! This value matches the components that were used. In the table below we summarize how tests with this power supply delivering 250 W. The value listed under “total” was the total amount of power the unit was actually pulling, as measured by our load tester

The power supply died silently, no explosion happened. After disassembling the power supply we measured all main components and what burned was the +5 V rectifier.

Room temperature was below what we usually use because we couldn’t increase temperature inside our “hot box” since the power supply wasn’t pulling too much power and thus not heating enough.

Voltage regulation during our tests was o.k., with all outputs within 3% of their nominal voltages – ATX specification defines that all outputs must be within 5% of their nominal voltages – except +5 V, which was at 4.81 V when we were pulling 250 W from the power supply. This value, however, is still inside the 5% tolerance set by the ATX standard. Of course we want to see all voltages as close as their nominal values as possible.

Efficiency was surprisingly high for a generic unit; we were expecting something below 70%. The best value was when we were pulling 100 W (78.7%) and the worst value was when we were pulling 50 W (73.2%).

But the main problem with this generic unit was noise and ripple. This is something regular users don’t even think about: most users choose a power supply solely based on its wattage, paying no attention on how clean the outputs are.

Outputs from the power supply are continuous voltages and when watching them on an oscilloscope they should be a straight line on the screen. This, however, doesn’t happen; outputs aren’t perfectly continuous. They can have a little oscillation (called ripple) and, on top of that oscillation, some little spikes (called noise). If the value of this oscillation and spikes are low enough they won’t offer any risk of damage to your equipment.

ATX specification says that ripple and noise should be within 120 mV for the 12 V outputs and 50 mV for the +5 V and +3.3 V outputs for the outputs to be considered safe for the electronic components used inside the PC.

The problem with this generic power supply is that its noise level was above those values all the time! When we started at 50 W noise level at +5 V output was already at 105 mV! When delivering 250 W noise level at +5 V output was at 220 mV and at +12 V output was at 180 mV!

So even if your equipment isn’t pulling a lot of power – for example, you have a very basic PC with a low-end video card or even on-board video –a generic power supply can cause you trouble because of the amazingly high noise level (caused by the removal of the coils and capacitors from the filtering stage in order to cut costs). Have you ever heard of instability problems solved by replacing a generic power supply with a “branded” one, even when the computer wasn’t pulling a lot of power? Well, this explains it. The bottom line is: wattage is not everything.

This also explains why we say that 99% of power supply reviews around the web are wrong: since the majority of websites don’t have an oscilloscope, they simply can’t see something horrible like this happening. A power supply being able to deliver voltages on their correct levels means nothing; we need to know how clean these voltages are.

Just a clarification before anyone asks. We said that noise at +5 V output was at 220 mV but on the chart below you are seeing a 160 mV pattern. What happens is that the power supply was producing a lot of fast high spikes that aren’t being shown on this captured screen.

Figure 10: Noise level at +5 V when the power supply was delivering 250 W.

For you to visualize how bad this is, we posted below the noise level measured on the +5 V output from a true 500 W power supply, Antec Earthwatts 500 W, with the exact same load pattern described on the table above (noise level here was below 20 mV). Both charts are in the same scale (2 ms T/div and 0.02 V/div).

Figure 11: Noise level at +5 V on Antec Earthwatts 500 W delivering the same 250 W.