Corsair HX1000W Power Supply Review

Load Tests

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

First we tested this power supply with five different load patterns, trying to pull around 20%, 40%, 60%, 80%, and 100% of its labeled maximum capacity (actual percentage used listed under “% Max Load”), watching how the reviewed unit behaved under each load. In the table below we list the load patterns we used and the results for each load. Then we tried to pull even more power from this unit and the results for this test are in the next page.

If you add all the power listed for each test, you may find a different value than what is posted under “Total” below. Since each output can vary slightly (e.g., the +5 V output working at +5.10 V), the actual total amount of power being delivered is slightly different than the calculated value. On the “Total” row we are using the real amount of power being delivered, as measured by our load tester.

Because HX1000W has two independent power supplies inside, we had to take extra care to not pull all current/power from only of them, what would make the power supply to shut down or have an unexpected behavior. So to the +12V1 input from our load tester we connected only cables that were connected to the +12V1 rail from the power supply (motherboard main connector, EPS12V connector and video card connector from the modular cabling system). We did the same thing with the +12V2 input: we connected the video card connector that comes from inside the power supply, which is connected to the power supply +12V2 rail. So on results below +12V1 and +12V2 really represent the +12V1 and +12V2 rails from the power supply.

Updated 06/25/2009: We re-tested this power supply using our new GWInsteak GPM-8212 power meter, which is a precision instrument and provides accuracy of 0.2% and thus presenting the correct readings for AC power and efficiency (results marked as "2" on the table above; results marked as "1" were measured with our previous power meter from Brand Electronics, which isn’t so precise as you can see). We also added the numbers for AC voltage during our tests, an important number as efficiency is directly proportional to AC voltage (the higher AC voltage is, the higher efficiency is). Also, manufacturers usually announce efficiency at 230 V, which usually inflates efficiency numbers. We added power factor (PF) numbers as well. These numbers measure the efficiency of the power supply active PFC circuit. This number sh
ould be as close to 1 as possible. Under light load (20% load, i.e., 200 W), the active PFC circuit from this unit isn’t as good as when operating under higher loads, but 0.989 is still an excellent number. At full load power factor was of 0.999, which is probably as high as you can get!

This power supply can really deliver its labeled power at 50° C (on test number five we collected data when the temperature inside our thermal chamber was at 47° C, but we let it working after that to see what would happen and the power supply worked fine at temperatures above 50° C).

Efficiency was always above 80%, however there are other 1,000 W power supplies that present higher efficiency, like OCZ EliteXStream 1000 W.

Pulling 1,000 W +5 V and +3.3 V voltages dropped to 4.76 V and 3.18 V, respectively. These values are still inside the 5% tolerance set by the ATX specification, but we wanted to see these values closer to their nominal voltages. On the other hand, we have an explanation for this behavior.

Our load tester is limited to 33 A (396 W) for each of its +12V inputs. To achieve 1,000 W we had to pull more current from +5 V and +3.3 V outputs than we wanted (we try to pull as much as we can from the +12 V outputs, since current PCs will pull more current/power from +12 V, which is the line that feeds the CPU and the video cards). Because the peripheral and SATA power connectors were connected to the +12V2 rail, we didn’t connect them to the load tester, because on our load tester peripheral connectors are physically connected to the +12V1 input, and we didn’t want to mix +12V1 and +12V2 rails. So the wires on the main motherboard cable were the only ones carrying +5 V and +3.3 V voltages and we believe that voltage dropped because we didn’t have more wires carrying these voltages connected to the load tester.

In summary, it is our opinion that you should not worry about these values we achieved.

For all other tests voltages were within 3% from their nominal values.

Ripple and noise were below the maximum allowed: less than a half from the maximum admissible, even when we were pulling 1,000 W from the reviewed unit.

Below you can see noise level when we were pulling 980 W (test number five) from this power supply. Just to remember, the maximum allowed for the +12 V outputs is 120 mV peak-to-peak and the maximum allowed for the +5 V and +3.3 V outputs is 50 mV peak-to-peak.

Figure 20: Noise level at +12V1 input from our load tester with the reviewed unit delivering 980 W (50.6 mV).

Figure 21: Noise level at +12V2 input from our load tester with the reviewed unit delivering 980 W (42.6 mV).

Figure 22: Noise level at +5 V input from our load tester with the reviewed unit delivering 980 W (27.8 mV).

Figure 23: Noise level at +3.3 V input from our load tester with the reviewed unit delivering 980 W (23 mV).

Now let’s see how much power we could pull from this unit keeping it working inside ATX specs.