Enermax Galaxy: A KiloWatt power supply

TESTING

For a fuller understanding of ATX power supplies, please read the reference article Power Supply Fundamentals & Recommended Units. Those who seek source materials can find Intel's various PSU design guides at Form Factors. A complete ssi EPS12V PSU standards can be found on the ssi forum.

You can find out everything you want to know about SPCR's test equipment at SPCR's PSU Test Platform V.4. The testing system is a close simulation of a moderate airflow mid-tower PC optimized for low noise.

In the test rig, the ambient temperature of the PSU varies proportionately with its output load, which is exactly the way it is in a real PC environment. But there is the added benefit of a high power load tester which allows incremental load testing all the way to full power for any non-industrial PC power supply. Both fan noise and voltage are measured at various standard loads. It is, in general, a very demanding test, as the operating ambient temperature of the PSU often reaches >40°C at full power. This is impossible to achieve with an open test bench setup.

The 120mm fan responsible for "case airflow" is deliberately run at a steady low level (~6-7V) when the system is run at "low" loads. When the test loads become greater, the 120mm fan is turned up to a higher speed, but one that doesn't affect the noise level of the overall system. Anyone who is running a system that draws 400W or more would definitely want more than 20CFM of airflow through their case, and at this point, the noise level of the exhaust fan is typically not the greatest concern.

Great effort has been made to devise as realistic an operating environment for the PSU as possible, but the thermal and noise results obtained here still cannot be considered absolute. There are too many variables in PCs and too many possible combinations of components for any single test environment to provide infallible results. And there is always the bugaboo of sample variance. These results are akin to a resume, a few detailed photographs, and some short sound bites of someone you've never met. You'll probably get a pretty good overall representation, but it is not quite the same as an extended meeting in person.

REAL SYSTEM POWER NEEDS: While our testing loads the PSU to full output in order to verify the manufacturer's claims, real desktop PCs simply do not require anywhere near this level of power. The most pertinent range of DC output power is between about 65W and 250W, because it is the power range where most systems will be working most of the time. To illustrate this point, we conducted system tests to measure the maximum power draw that an actual system can draw under worst-case conditions. Our most power-hungry Intel 670 (P4-3.8) processor rig with nVidia 6800GT video card drew ~214W DC from the power supply under full load — well within the capabilities of any modern power supply. Please follow the link provided above to see the details. It is true that very elaborate systems with the most power hungry video card today could draw as much as another 60~100W, but the total still remains well under 400W in extrapolations of our real world measurements. As for high end dual video card gaming rigs... well, to be realistic, they have no place in silent computing today.

SPCR's high fidelity sound recording system was used to create MP3 sound files of this PSU. As with the setup for recording fans, the position of the mic was 3" from the exhaust vent at a 45° angle, outside the airflow turbulence area. The photo below shows the setup (a different PSU is being recorded). All other noise sources in the room were turned off while making the sound recordings.

INTERPRETING TEMPERATURE DATA

It important to keep in mind that fan speed varies with temperature, not output load. A power supply generates more heat as output increases, but is not the only the only factor that affects fan speed. Ambient temperature and case airflow have almost as much effect. Our test rig represents a challenging thermal situation for a power supply: A large portion of the heat generated inside the case must be exhausted through the power supply, which causes a corresponding increase in fan speed.

When examining thermal data, the most important indicator of cooling efficiency is the difference between intake and exhaust. Because the heat generated in the PSU loader by the output of the PSU is always the same for a given power level, the intake temperature should be roughly the same between different tests. The only external variable is the ambient room temperature. The temperature of the exhaust air from the PSU is affected by several factors:

• Intake temperature (determined by ambient temperature and power output level)
• Efficiency of the PSU (how much heat it generates while producing the required output)
• The effectiveness of the PSU's cooling system, which is comprised of:
  • Overall mechanical and airflow design
  • Size, shape and overall surface area of heatsinks
  • Fan(s) and fan speed control circuit

The thermal rise in the power supply is really the only indicator we have about all of the above. This is why the intake temperature is important: It represents the ambient temperature around the power supply itself. Subtracting the intake temperature from the exhaust temperature gives a reasonable gauge of the effectiveness of the power supply's cooling system. This is the only temperature number that is comparable between different reviews, as it is unaffected by the ambient temperature.

TEST RESULTS

Opening conditions for this test were measured at 21°C and 20 dBA. The input voltage was 122VAC.

It took some effort to get this big boy into place, but we came out with a sense of accomplishment. The connectors were duly plugged in, and the testing was begun!

Our PSU tester is beginning to look like a Frankenstein creation.

ANALYSIS

1. EFFICIENCY
The efficiency was a bit of a letdown. It seems the marketing department might have been overzealous with their "80-85% efficiency from 20% to 100% load" claim. On our test bed the PSU barely peaked 80%. Higher than normal measurement error might have sprung from some tinkering we had to do in order to produce the 1kW power load needed. However, those changes don't apply until 500W and higher tests loads (where the 12V3 line was engaged), and the numbers below that are still unimpressive. Even if your system is hardcore gaming, at typical idle to low loads (less than ~200W), it will likely be running at a not too thrilling 65-78% efficiency. Since heat was up fairly dramatically in the higher power ranges, temperature probably took its toll on the PSU efficiency.

2. VOLTAGE REGULATION
The voltage regulation was excellent. No matter what the load combination, values stayed pretty much where they were supposed to. Perhaps this was a benefit dual 12V transformers. Voltages varied only 3% from target, and only on the 3.3V rail. Other rails barely breached 1% at their worst.

3. RIPPLE
The AC ripple was vanishingly low! It's about on par with the random electrical noise you find in most cities. We can only imagine what kind of hardware was necessary to make this happen, or maybe they had some spiffy new trick up their sleeves. Either way the numbers (+12V 6.4mV @ 500W???) reflected a well built power supply you could trust to supply darn well anything.

4. POWER FACTOR was excellent throughout the tests. At very low load, it varied a tiny bit, but once we started demanding a little more from the PSU it hit the ideal 1.0 and stayed there.

5. LOW LOAD / CROSSLOAD PERFORMANCE
Our usual low / no load testing could not be done. That little Power Guard guy does its job quite well. We were prevented from testing at very low power or using no load starting conditions. Later on, we were held back from overdriving the system, too. Best of all, when we were in proper operating range Power Guard left us well enough alone. Included was that handy-dandy light and beep setup to explain the source of an error. A clear table in the manual explained the meaning of each color and light combination. The Power Guard shows us how it's done.

Neither could the crossload test be performed. Our usual procedure is to obtain ~75% load, almost entirely on the 12V lines, with only one amp each on the +5V and +3.3V lines. With a 1000W PSU, this is a pretty extreme load, some 750W in total. The Power Guard simply refused to let the PSU be turned on with this load, even if we built up to it by adding loading incrementally while the unit was running; it would shut down when the +5V and +3.3V lines got too low or when the 12V load got too high. If such conditions are not safe for the PSU, then the Power Guard's protective action is a good thing. Besides, it's hard to imagine how any system would demand over 740W on the 12V lines while pulling under 10W on the +5V and +3.3V lines.

One remote downside of the Power Guard: It is possible that in some unusual imbalanced loads, the Power Guard's protective action could cause a problem. In the context of the kind of system the Galaxy is desgined to be used, it's an unlikely scenario.

6. LOW AC VOLTAGE PERFORMANCE

Low VAC Test: Enermax Galaxy at 720W output
VAC	AC Current	AC Power	Efficiency
120V	7.62A	912W	79.0%
110V	8.41A	921W	78.1%
100V	9.34A	932W	77.2%
90V	10.5A	942W	76.4%

The Galaxy was unfazed by low AC voltage inputs at ~75% of rated power. Ripple stayed well within limits and voltages stayed stable at all times. As expected, the efficiency dropped a bit as the input voltage was lowered.

7. TEMPERATURE & COOLING
This PSU was very well cooled at the lower levels. The fan controller (as you'll read in the next section) made sure of that. As we got close to maximum power, things got really hot. 55°C at 718W is not bad at all, judging by previous test results. But at full load, we hit 71°C and it would have risen higher had we run the test longer. The high rise in temperature had to do with the sheer amount of heat represented by nearly 1300W. That's the amount of power drawn by the PSU into the test box. Even at close to 80% efficiency, there was still over 270W of heat to remove from within the PSU — not to mention the role of the PSU fans in pulling the rest of the heat (1000W) out of the test box.

We also mentioned that there's a possible the mismatch in airflow between the two fans: The 80mm can only pull 44-46 CFM while the bigger one does more than double that. In any two fan push-pull setup such as this, the combined airflow through the two fans cannot be higher than the slower of the fans. As a result, the slot vents on the output cable side could act as a pressure valve for the higher airflow/pressure of the larger fan. This could mean some of the PSU heat was just being recycled back into the test box.

With the quiet ~45 CFM 120mm fan used to cool the text box, there was simply too much heat in the test box for the fans to cope. If you have a system that can really pull this kind of power, we'd suggest the equivalent of at least a hundred CFM airflow exhaust for the case.

8. FAN, FAN CONTROLLER and NOISE
Okay, here's the make or break when it comes to our reviews. PSUs can be a lot of things and still pass with us, but this is where we get tough.

The Galaxy simply wasn't aimed at the quiet market. We mentioned earlier that the unit's efficiency may peak around 500W. At the same time, fan voltage (along with speed and noise) also hit its max. The Enermax Galaxy started at a relatively high 29 dBA@1m. The noise quality was a mix of buzz, hum, rattling and resonance. It would be perhaps borderline acceptable, but it's basically an unpleasant sound. The noise level climbed steadily and linearly with load and temperature, which can only mean one thing: The priority here was to keep the PSU cool enough at all times. The noise reached an alarming 50 dBA@1m at 500W load.

Contributors to the noise:

• buzzy, rattling ball bearings
• resonant clear plastic of the fan blades
• linear temperature / voltage fan speed controller instead of stepped
• turbulence caused by the dual fan setup

Suffice it to say that the acoustic performance is not charming. You can hear it for yourself in our recordings on the next page.