PSUs

Modern PSUs have their work cut out. They have to convert 240V AC from the mains to the various DC voltages (3.3V, 5V, 12V, -12V) required by each circuit in the PC, while making sure they don't consume too much power themselves. An inefficient PSU will not only drive up your electricity bill, but also require more powerful (noisier) cooling to dissipate the wasted energy.

Whether you're planning on upgrading from an older, inefficient PSU, or you simply need a more powerful PSU for your next PC, we've tested 29 different PSUs, encompassing the complete range, from bargain-basement 550W models to 1.5kW monstrosities.

What a PSU does

A PSU is simply an AC to DC converter, converting the 240V AC (alternating current) supplied by the mains to the DC (direct current) voltages required by all the components inside a PC. A typical PSU for a desktop PC has to output five different voltages, known as rails, which are 3.3V, 5V, 12V, -12V and 5V standby (5VSB).

What to look for

For most mid-range PCs with a couple of hard disk drives and optical drives, a single graphics card and a mid-range CPU, a PSU of around 500W will suffice. However, if you're building a high-end, dual-graphics system with a multicore CPU then you need to think about buying a more powerful PSU.

Aside from power, you also need to consider which connectors your PC will need. For example, all new motherboards use a 24-pin ATX connector, but older motherboards use 20-pin ATX connectors. Therefore, if you're upgrading from an old system (a couple of years old or more) then it's worth checking the feature table carefully, as most high-end PSUs don't support 20-pin motherboards. It's possible to buy adaptor cables that convert the 24-pin connection to the older 20-pin connection, but you'd have to buy this separately.

In addition, many high-end motherboards also have 8-pin EPS12V connectors instead of the more common 4-pin ATX12V connectors, so you need to make sure that the PSU you buy supports your motherboard. Some dual-processor motherboards, such as the Intel D5400XS (Skulltrail), even have two 8-pin EPS12V connectors, which very few PSUs support.

Unfortunately, the trouble with selecting an appropriate PSU is that the limited information provided by many manufacturers doesn't actually tell you how good the PSU is. Each PSU is rated as being capable of outputting a certain wattage (600W, for example), but this doesn't tell you much, as many manufacturers list the wattage in a confusing manner.

Many people simply add up the wattage over the primary rails (3.3V, 5V and 12V), and list this as the total. However, no PSU is capable of simultaneously producing the maximum wattage from each rail, so this method gives an overly generous impression of the PSU's capabilities. Make sure that you note how many watts can be produced in parallel by each rail when reading the label on the box. More importantly, check the current rating on the 12V rails, as these rails power the CPU(s) and graphics card(s) - the most power-hungry components in a modern PC.

Unfortunately, even closely scrutinising the label won't tell you anything about voltage stability, which is probably the most important factor in a PSU. Producing a stable voltage on each rail is a pretty tough task, and some PSUs simply aren't up to it.

Voltage stability is important, as some components may not start up if the voltage is too low, or could burn out if the voltage is too high. You certainly don't want to risk damaging your new CPU or graphics card, just because you bought a nefariously labelled PSU.

Fortunately, the Intel ATX spec, which you'll find at www.formfactors.org, lays down the physical and electrical characteristics necessary to adhere to the spec in black and white. If you read last year's PSU Labs test, in which we discovered plenty of PSUs that failed to make the grade, including several that blew up, then you'll be pleased to hear that our rigorous testing procedures have compelled manufacturers to improve the quality and safety of their latest PSUs. That said, a small proportion of the models tested this year failed some of the voltage stability tests, so there's clearly still work to be done.

The ATX spec cites that there can be a maximum 5 per cent variance (above or below) the voltage on the 3.3V, 5V and 5V standby rails, and up to a 10 per cent variance on the -12V rail. For the 12V rails, the voltage can vary by up to 5 per cent at typical load levels, but at full load, it can vary by as much as 10 per cent. To save you from having to work this out for yourself, the table below shows what the minimum and maximum voltages are for each rail.

How we tested

As PSU test equipment is highly specialised, we used a third-party lab - in this case, the UK branch office of Enermax. This enables us to use the latest generation of PSU testing equipment. By using a combination of a Chroma 6430 programmable power source and a Zentech 2100 digital power meter, we were able to measure the voltage stability, overall efficiency and PFC efficiency of each PSU.

To test each PSU, we programmed the load testers to drain the amount of power that each manufacturer claims its PSU can deliver.

The voltage of each rail was measured at 50 and 100 per cent loads to determine if it was within the ATX spec. We then left each PSU running at 100 per cent load for 15 minutes to test if it could produce stable voltages over an extended period.

Let's take the Cooler Master Silent Pro M700W as an example. Cooler Master claims that it can output 28A on the 3.3V rail and 25A on the 5V rail, but together these rails can output no more than 165W. It also has a single 12V rail rated at 50A, a 0.5A -12V rail and 2.5A 5VSB rail. This meant that we had to carry out 18 individual measurements to test whether it meets the ATX spec.

An area of increasing interest is the overall efficiency of the PSU or, in other words, how much power it requires from the mains to produce the requested load. Efficiency is very important because, in a manner similar to that of an internal combustion engine, any wasted energy is dissipated as heat, which the cooling system has to expel, otherwise the effectiveness of the electronic components inside the PSU will be reduced, or the PSU will burn out. For example, many PSUs are only rated as being capable of providing their claimed output at 25ûC to 40ûC, and their output will drop above this level. This means that a less efficient PSU will require a more powerful and noisier cooling system. An efficient PSU, on the other hand, will run cooler and draw less power from the mains, which will help to reduce your electricity bill. The ATX spec cites that, at full load, a PSU has to be at least 70 per cent efficient. As you can see from the Test Sheet, the Cooler Master Silent Pro M700W is quite efficient, averaging 83 per cent efficient at full load, as it drew 840W from the mains to produce 700W.

It's also important to test the efficiency of a PSU's Power Factor Correction (PFC) circuitry . The PFC helps to reduce uneven harmonics in the incoming current, which then helps to reduce the PSU's power consumption. We measured the PFC efficiency during each of the load tests, and the results for the Cooler Master Silent Pro M700W are shown in the Test Sheet.

The scores

Although there were several hundred test measurements to analyse, the process was made slightly easier by using the ATX spec as a guide on how to judge the PSU.

The voltage stability tests take pride of place in the scoring box and account for 50 per cent of each PSU's overall score. The scores shown are the number of test measurements that are within the limits of the ATX spec, converted into a percentage to allow PSUs with a different number of rails to be compared.

Obviously, we recommend that you only buy a PSU that has perfect (100 per cent) stability. Unfortunately, we don't have space to print all the voltage stability measurements, but the graphs on p88 show the voltage output from each rail during the 100 per cent load stress test.

The Features score includes the overall efficiency and PFC efficiency as measured during the 50 and 100 per cent load test scenarios, the number, type and quality of the cables, plus how noisy the cooling fan(s) are at 100 per cent load. The Value score is a combination of the Stability and Features scores, and the total output (in watts) divided by the price. These three scores were then added to the Overall score of each PSU.