The Big Power Supply Guide - Ratings
(Or, why 2 + 2 = 3)
Herein we will discuss the way all of those magical power supply ratings are forumlated, what they really mean, and why you shouldn't put too much faith in them. This is, by necessity, more technical than the other sections. There might even be some math. You've been warned.
Time for a quick lesson on the guts of a power supply. One of the core elements of a typical PC power supply is a transformer. Transformers take in one AC voltage and output one or more other AC voltages. In a PC power supply, those output voltages are then rectified into DC voltages.
A computer needs 3 main voltages: 3.3 V, 5 V, and 12 V. Now, it would be possible to use 3 separate transformers and devote one to each voltage. But transformers aren't cheap. And 3 cost twice as much as 1. So many power supplies just have one somewhat beefier transformer to do all the work.
Now this is cheaper, and lighter, and saves space. But now instead of spreading the heat out across 3 toroids (the metal ring that is the soul of a good transformer), it's all focused in one. And all that magnetic flux is going through one core, so your chances of saturating it (saturating it is not good) are increased. The bottom line is, using fewer transformers limits power output.
You see, if two voltages are drawn off the same transformer, there are three limits to consider. There's the power rating for each of those "rails", of course. And then there's a combined rating for how much total power they can deliver. So if you start pushing the limits of one of the rails, you're going to be limiting the output of the other one.
For this reason, many power supplies have a "combined 3.3 V and 5 V" rating. So suppose for a moment that the 3.3 V rail is rated for 100 W, and the 5 V rail is rated for 100 W, with a combined rating of 150 W. If you're drawing 100 W on the 3.3 V rail, you can only draw 50 W from the 5 V rail before you start to go out of spec.
If you've ever figured out the wattage rating of each rail of a supply and added them up, you probably noticed it was vastly more than the total power rating the manufacturer advertises. And this is why. The supply simply cannot deliver full power to all the rails at once.
This is really the same concept as above. Many power supplies have 2 (or more) 12 V rails. And when people see that, their natural instinct is to add them up to get the total deliverable 12 V power.
Oh, if only it were so simple.
It's the same thing we saw above. Multiple rails, yes. Multiple transformers, usually no. While some power supplies have as many as 4 rails delivering 12 V power, these often come off of the same transformer but have separate Over Current Protection devices. So as above, the power draw on one can limit the power another one can deliver.
A lot of people will say you shouldn't buy a supply rated for a much higher wattage than you need, because it will waste power. They have no idea what they are talking about.
A power supply will only deliver as much power as its load demands. Power supplies are rated in terms of how much power they are capable of delivering, not what they always will.
Efficiency is a hard number to properly interpret. If you pay attention, you'll notice efficiency ratings are always given in the form "> 70%". That is, the manufacturer guarantees that the supply will operate at at least 70% efficiency as long as it's used in spec.
Efficiency is not a constant. It varies as the load varies. For a typical power supply, efficiency is poor for very small loads, then improves up to something like 80% load, and then tapers off again.
Honest manufacturers figure out efficiency ratings by putting a supply under fairly heavy load, pushing them toward their minimum operating efficiency. Dishonest manufacturers use lighter loads, which means higher efficiency ratings.
For this reason, it can be a fool's errand to compare efficiency between two substantially different supplies. If you put the same load on a good 350 W supply and a good 500 W supply, and they have the same rated efficiency, the 500 W supply may be more efficient.
Have you ever heard the phrase, "numbers don't lie"? It's not true.
Heat is the bugaboo of many electrical circuits. Power supplies are especially vulnerable. When they get hot, they get less effective. They can't deliver as much power, they aren't as efficient, and in general they simply don't work as well. And this matters to you for two major reasons.
The trick here is that many manufacturers are choosy about the conditions they use to test their supplies. They test at, say, 25 C. Even though the realistic temperature inside a supply is substantially higher. So the power supply looks a lot better on paper than it really is. PC Power & Cooling has made a big deal about testing at realistic temperatures, and some other companies do as well.
Let's face facts. Not everyone is honest all the time. Power supply makers are, by and large, rarely if ever honest.
500 W doesn't always mean 500 W. 500 is a popular number with the trash sellers that want you to believe their product isn't garbage when it really is. 500 is less popular with the people that don't lie out of habit.
Strictly speaking, a power supply only has to be able to meet a particular specification through an hour of continuous use to qualify as "passed". Now that doesn't mean your, individual supply. That means something like one out of each batch, sometimes hand-chosen to ensure it will pass, and sometimes provided with some rather exotic cooling for good measure (see the section above).