HardwareLogic’s PSU Testing Methodology, Version 2.00.

This is a living document. Over time it will change as our methods are improved or change with "the times." Changes will be dated either inline with the text body or within the Addendum at the end (or both).

Last edit: 09/22/07

As of September 2007, HardwareLogic upgraded its lab to include a FAST AUTO ATE 828/4500 system capable of load testing PSU units into the kilowatt range. Thus we join a handful of other review sites that have similar equipment. Our goal is to continue to bring quality PSU reviews to our readers, but with more quantitative data backing the reviews. This means more accurate information gets to the reader to allow them to make a more informed purchase.

Our motivation and philosophy remain the same. We aim to give you, the reader, an indication as to whether or not a particular PSU will fit your needs, if that PSU works as advertised, and how that PSU may stack up to ones of the same class (mainly as a measurement of value). As always, we encourage you to research any purchase beyond our own review and will provide links to additional reviews or information when available.

Power Supplies: How we review

Feature Set, 10 of 100 points

The Feature set we typically look at includes anything above and beyond what one might expect in a generic PSU. This includes "look and feel." While our more practical side tells us the color of a PSU is not important, in a day and age of truly personalizing the personal computer through activities like modding, our aesthetic side tells us different. Looks alone won't score a 10 in this section though. Additional accessories such as splitters and adapters are also considered.

One thing that is not considered in this section is the packaging. In fact, the only time you should hear us talk about packaging is if something on the packaging is misleading or unusual. Although, we may mention the packaging if it happens to be the perfect size to use as a cookie transport box.

Cables, Connections, and Dimensions, 15 of 100 points.

These are all very important aspects of a PSU. We consider a number of questions here. Are the cables long enough? Are they too long? Are there enough and an adequate variety of connectors? Does the size of the PSU conform to the ATX specification that most cases allow for? Are there adapters included for any short comings? Does the wiring adhere to the ATX specification?

All of these question and more will determine if a PSU gets the 15 points available in this section.

Performance - 45 of 100 points

This is the most important section of our review and the section we put the most effort and time into. If a PSU will not perform as advertised (or at all), it is of little use regardless of the features, price, or color. To that end, we run our PSUs through a pretty strong test suite.

Our testing includes putting real loads on the PSU rails while monitoring voltage, room temperature, exhaust temperature, DC out voltage ripple, and power efficiency. The heart of our system is a FAST AUTO ATE 828/4500 tester capable of load testing PSU units into the kilowatt range. We currently run six tests partially based on the Intel ATX12V Power Supply Design Guide Version 2.2 Specification. It is important to note that the tests we run put a PSU in some unrealistic load configurations for the purpose of testing and can be a bit harsh on a PSU. The logic here is two fold. First, the Intel specification recommends that a PSU pass similar tests. Second, if a PSU can pass these test, we are quiet confident it will operate in a range meeting or exceeding the class it is designed for.

Also worth noting is that the ATX12V Guide only covers power supplies up to 450 watts and is a bit out of date. Thus, our tests (especially on larger supplies) are "based" on the testing philosophy gleaned from the guide.

Load Tests 1 and 2:

These tests are based on the Intel ATX12V Guide percent amperage loads of "Light" and "Typical". These are non-negotiable, semi-proportional, load tests. "Non-negotiable" means we will not adjust the tests to meet qualifiers that might be listed on the PSU's specification panel (see tests 3 and 4 for a better explanation). We base these loads on the stated maximum currents of each rail as printed on the specification panel of the PSU. The 5VSB rail is held at 1 amp in both tests. These test are run until the PSU reaches thermal stabilization (10-30 minutes).

Load Tests 3 and 4:

Our tests 3 and 4 are designed to emulate the ATX12V specification of "Full" and "Max". While the specification bases these tests on maximum amps per rail, modern PSUs often "qualify" the maximum amperage ratings with a maximum combined wattage rating (also allowed in the ATX12V specification).

For example; a PSU may state that the 12V1 and 12V2 rails can each take 20 Amps maximum load. That's a whopping 480 watts at full load! However the PSU may also state that the combined wattage on the 12V rails cannot exceed 300 watts. So, while you may be able to load either of the rails to 20 Amps, you cannot do so to both of them at the same time. This can be a bit misleading when you look at the specification panel on a PSU, but it is done for purposes of clarity (and marketing). The end result of this is that we base tests 3 and 4 on maximum wattages rather than amps.

For test 3, we load the 5V and 3.3V rails to 80% of the maximum stated wattage for those rails. We then ramp up the combined 12V rail loads to 90% of the PSU's total wattage rating (or 90% of the combined 12V rail amperage - which ever wall we hit first). The -12V and 5VSB rails are loaded to 100% load and 1 amp respectively. We expect any reasonable power supply to be able to hold these levels without shutting down for a minimum of 1 minute, although we will gladly run the test until thermal stabilization is achieved if the PSU can take it.

For test 4, we deviate somewhat from the ATX12V guide. We increase the 5V and 3.3V loads to the maximum stated combined wattage for those rails. We also increase the load on the combined 12V rails until we reach close to 100% of the PSU's overall maximum rated wattage output. Again, the -12V and 5VSB rails are loaded to 100% load and 1 amp respectively. This is an exceptionally difficult test as it introduces an unusual cross load at maximum output wattage. No PSU manufacturer is expecting you to max out the 5V and 3.3V rails at maximum PSU wattage output. That said, the PSU should still be able to do it - for at least 17 seconds. At 17 seconds, if the voltages were maintained and the PSU did not shut off - it gets an unconditional pass. Bravo!

In the "real world", these tests are "hardcore". It is unlikely any computer will ever put these types of loads on a PSU for more than a few seconds - if ever at all. It is far more likely that if PSU is too small or weak for a system, one of the rails will be maxed out long before any of the others. So, a PSU that passes these tests is one we consider to be of very high quality. PSUs that "almost" pass test 3 & 4 may still be commendable depending on exactly what the short comings are.

Load Tests 5 and 6:

These two tests are designed to more fully test the PSU through extreme (but within Intel specification) cross loading. Cross loading occurs when one rail (or a combined set of rails) of a PSU is loaded heavily while another rail carries an unusually light load. This condition can be especially difficult on a PSU, but occurs "in the real world" as more the rule than the exception. These are not high wattage tests.

For test 5, we keep the combined 12V rail load we used in test 4 and drop the 5V and 3.3V rails to 10%. This is can be rather extreme cross loading, but a good PSU should be able to handle it and it does reflect a load a PSU may expect to see in real, if not extreme, operating conditions.

For test 6, we put a moderate load on the 5V and 3.3V rails (45% of maximum stated amps) while putting a lower load on the 12V rails combined (30% load).

*See preceeding text

Of these two tests, test 5 is the most important because it represent a condition most likely to occur in a real PC. Test 6 loads the opposite cross load on the PSU. If the PSU passes both these tests, it should pass any cross loading in between.

Of all of our tests, we weigh test 1, 2, and 5 respectively, the most - they make up 30 of the 45 possible points in this section.

Efficiency - 10 of 100 points

Throughout our testing we monitor efficiency. This is basically a measure of how much power goes in to the unit verses how much power comes out of it. This is important for a few reasons. First, PSU makers advertise this efficiency as a marketing point. We want to test to see if they are on the up and up. Second, efficiency is part of Specification 2.2. Third, lower efficiency means more heat - which means more cooling - which can mean more noise and even more power consumption. Lastly, in a time when we argue over energy resources, are watching our ice caps melt, seeing the Amazon River (largest catchment basin in the world) dry up, and, maybe more import to some, watching energy costs rise substantially, efficiency is important. After all, a little green can save a little green, and multiplied over millions of PSUs, maybe we actually can help the planet a bit (I plan on moving to planet Beta Ceti 4 actually, but those of you that missed out on Mothership tickets might want to consider a more eco friendly style of life).

Warranty and Support - 10 of 100 points

We judge this category passed on the length of the warranty and what is covered, along with any additional information on service we can find. We do not intentionally jam up service lines to check customer service (that's a subject for a different type of article). If we do need service help for some reason, we will include the experience as part of the score. Support information from the manufacturer's website is also considered.

Price and Value - 10 of 100 points

Price per watt matters here - we can use it to judge if a PSU is priced reasonably for its class. Value, however, requires comparing the price per watt to the results of our review and making a judgment. For example, a highly priced PSU may be worth the extra money because of a particular set of features, while a low price per watt PSU may turn out to be a bad deal at any price.

Addendum

09/15/07 - Changed tests 3 and 4 from simple % load tests to more complicated % wattage tests to attempt to deal with the "Load to Power" discrepancies on PSUs.

09/22/07 - Minor Edits