Everything you need to know about power supplies

The primary task of a power supply is to take the AC (alternated current) input and convert it to DC (direct current), and to do so at a variety of voltages required by the components of a PC. The specs of a PSU are defined by a number of factors, so how can you tell what to look for?

What’s watt?
The most important function of a PSU is its output power, measured in watts. In simple terms a watt is the rate at which energy is transmitted by a circuit, although other factors play a role (such as resistance). Generally, wattage is measured by multiplying amperage by voltage.

PSUs are commonly marketed by their total output power in watts. The value is a total of all the separate output rails combined (more on this below), but it can sometimes be misleading: firstly, as a combination of all the rails, you don’t necessarily get access to all that power where you need it most. Many PCs these days rely heavily on the +12v rail, and two identically rated 500W PSUs can provide two different output maximums on the +12v rail, for example.

Secondly the rated output power isn’t a hard limit. A 500W PSU can output more than this, but it’s not rated to do so efficiently or stably, and ultimately protection circuits kick in.

Finally, a PSU’s output power is also rated for a given temperature. Any good PSU for a PC will be rated for 50 degrees Celsius. The temperature rating is important – as a PSU heats up, its efficiency can decrease. This is known as the de-rating curve, and we’ll cover this more below.

You will also see some PSUs advertised with a ‘peak’ rating. Ideally, don’t use this as a consideration for purchasing, look for the ‘real’ rating that the PSU can consistently deliver. Any good PSU usually has a ceiling above its rating, but if you need that much power you’d be better off getting a more powerful PSU in the first place. That said, some cheaper PSUs are rated higher than the real output wattage they can reliably deliver. This is a dodgy practice but is usually relegated to low-end yum-cha units.

Each DC output is known as a ‘rail’, and PC PSUs are required to deliver a number of rails: +3.3v, +5v, +5VSB (standby – active as long as there is current from the wall, lets you turn your PC on from the front button), -5v, -12v, and the ever important +12v.

When it comes to buying a PSU the main stats you’ll see advertised (on the box and the PSU itself) are the total wattage and how this is divided among the various rails. The output of a particular rail is determined by multiplying the voltage of the rail by the amps – for example, +5v at 30A would be 150W.

The eagle eyed among you may notice that, depending on your PSU, the values don’t always add up – you might see the +5v rail at 30A and +3.3v rail at 24A given as a total of 170W. But if you do the math, it should be around 230W.

What’s going on? Even though the two rails have a maximum of 30A and 24A respectively, the total output power is 170W between them – meaning one or the other could be maxed out at a time, but not both at once. Usually this isn’t a problem; you’ll often find the +3.3v and +5v rails aggregated this way (because the +3.3v is actually drawn down from the +5v source), and they provide enough power for their given duties.

Note that once upon a time CPUs used to run from the +5v rail, but these days they use +12v. Any motors also draw from +12v – so fans, hard drives, optical drives, pumps and so on – as well as your GPUs. In fact, most of your PC’s load, as much as 90 per cent, comes from the +12v source. This is one reason it’s good to look at the distribution of power on a PSU’s rating – it doesn’t matter that it’s a 700W beast if only 400W is on the +12v rail and you need more than this to satiate your gear.

Efficiency is the second most important feature to look for after output power. A PSU’s efficiency rating is effectively a measure of how well it performs. In other words, how much energy is wasted in the conversion process. A PSU with 80 per cent efficiency will pull 500W from the wall to deliver 400W for your system, while a PSU with 70 per cent efficiency will need to pull 570W from the wall to do the same. On the surface this means a more efficienct PSU is slightly cheaper to run, but there’s a more important benefit: wasted energy is lost as heat, and heat is the number one enemy of a PSU.

As a PSU gets hotter it becomes less efficient and its maximum output power drops. In turn, its efficiency drops and the process spirals. Additionally, if it gets too hot, it becomes harder for the PSU to maintain the rails within their regulation (usually +/- five per cent) and since your gear is developed with these specs in mind, the result can be instability or crashes.

This loss in performance as a PSU heats up is known as the de-rating curve, and is the reason a PSU is rated for operation at up to a certain termperature. A 500W PSU rated for 50 degrees Celsius is rated to provide 500W at up to 50 degrees, past this its maximum output can drop. Considering the insides of most cases can usually get rather warm, there’s a good reason PC PSUs are held against the 50 degrees threshold.

Naturally as it gets hotter the PSU’s fan will spin faster in an effort to expel the heat, and it’s not a lost irony that if it’s getting hot due to a high load then it’s usually pulling in hot air from inside the case as well.

Today any good PSU will be branded with the ‘80 plus’ sticker that shows it can maintain 80 per cent efficiency for its typical load (usually 50-75 per cent of its rated power). It’s true that some PSUs lose efficiency at low power loads (say, 20 per cent of power) as well as at very high loads (especially when reaching 100 per cent or more). Naturally, if efficiency drops, more heat is generated, compounding any heat issues. These days, however, more and more PSUs are increasingly able to deliver 80 per cent efficiency across all loads – but check when you buy to see if this is the case.

It’s for this reason that it’s recommended to get a PSU with at least 30-40 per cent more power than you think you’ll need under load. This isn’t about headroom for future growth (you should’ve taken that into account already), it’s about ensuring you don’t approach the PSUs maximum output power to ensure maximum efficiency and minimal heat. The secret to a stable and quiet PSU is a higher output power and a high efficiency.

But even though 80 per cent efficiency is rather good (PC PSUs used to be 70 per cent and less only a few years ago) wouldn’t it be better if it could be 90 per cent? Indeed, a push is already under way in the industry to see a new 90 per cent standard introduced, and we may well see ‘90 plus’ PSUs in the next few years.

Ripple, noise, power factor correction
Other factors you’ll read about but which are less important are ripple, noise, and PFC. Noise and ripple (or regulation) are used to describe the cleanliness of a rail’s output, how much the voltage fluctuates and how tightly the rails adhere to their rated voltages. Some say overclocking can be limited by PSUs with poor results here, but as long as they stay within limits when under load you’re fine. PFC is a little complicated given the space to elaborate here, but lets just say it’s good if it’s as close to ‘one’ as possible, and most PSUs do this just fine.

Modular versus not
An old wives tale used to put modular PSUs at a disadvantage, stating that the extra resistance induced as a result of the modular connections was a cause for instability. Poorly made modular PSUs could also be problematic by creating loose connections, adding another level of complexity to diagnosing power problems.

In truth, when modular PSUs first appeared, these tales were based in fact, with some badly made modular PSUs giving rise to the wisdom that modular wasn’t the way to go.

But this isn’t the case anymore. Any good PSU with modular outputs usually ensures both tight fitting and snug connections, and doesn’t suffer any sort of issues related to increased resistance – after all it’s just another connection, same as plugs on the end of the cables.

So don’t be afraid to go modular. Additionally, modulars provide the benefit of being able to hook up only the cables you need, in turn reducing cable clutter and ultimately improving airflow within a case.

How much power do you need?
There are a number of power calculators you can use online, but keep in mind they’re only a guide. To give you a rough idea, however, an overclocked dual-core SLI system with a couple of drives might use around 450-500W at peak, and 250-300W at idle. If this sounds like you, and allowing for a minimum 30-40 per cent headroom, a 700-750W would be sufficient.

Is bigger better?
In our culture the tendency is to view bigger as better and PSUs are no different – we now have 1200W+ PSUs on the market. In truth, however, most enthusiasts – let alone normal people – would have a pretty hard time trying to load these babies. In other words, these PSUs are more for e-peen value than actual functionality, with one exception: efficiency on load. A 1000W PSU should, for example, be able to deliver 600W without skipping a beat, ensuring that its efficiency remains high under load and doesn’t heat up too much. Overall, however, you don’t need the beefiest PSU you can find. Even a quad-core three-way-SLI system can get away with an 800W PSU nicely.

Single or multiple +12v?
The ATX specification states that no more than 240W can pass through a single wire, as this much power could cause enough heat to melt the insulation and possibly start a fire.

Contrary to what you might think, bar very high end (1000W+) PSUs that actually do have two or more seperate +12v sources, most PSUs with multiple rails draw from the same +12v source. That is, when you see ‘+12v1, +12v2, +12v3’ on your 850W PSU it’s actually a single +12v rail split into ‘virtual’ rails, all in the name of safety. These ‘rails’ are simply limited connections that cap the maximum draw through any one rail.

So how does a single rail +12v PSU with an output more than 240W on the +12v rail get away with it? Herein lies much of the debate: if you think about it, both single and virtual rail PSUs all still draw from the same single +12v source, and all use multiple leads with various connectors. In other words, the distribution of power is inherently limited by the cabling – a single PCI-E 6-pin cable for your GPU, for example, isn’t going to be supplying more than the GPU demands (which is far less than 240W), and loading up a SATA cable will use even less. Across your whole machine you may use 500W or more, but no set of cable individually carries this much. It’s not surprising, then, that many high end PSUs, even with multiple +12v rails, sometimes list output power above 240W for a given rail.

Additionally, there’s one argument in favour of a single +12v rail: the maximum output power is available to use, whereas with a multiple rail configuration you can ‘lose’ output power from the headroom of each rail that doesn’t get used (you’re limited by what connectors are attached to each ‘virtual’ rail).

So the argument basically comes down to flexibility; a single +12v rail is more flexible for whatever your system demands, though technically a multiple ‘virtual’ +12v rail PSU is supposed to be safer.