Tag Archives: power efficiency

Power efficiency of ATX PSU, power adapters and UPS

This article goes hand in hand with my previous one, since the findings I report in here are those that led me to update the power source of my server.

When you buy a PC, usually the PSU is the last of the worries: the more Watts it’s got, the better, as it can handle all the power hungry hardware you plug to it. Also, usually, this is “just right”, as we use our main PC for just a few hours per day (as long as you make sure you either use suspend for short pauses, or hibernation for longer ones, so you don’t waste energy while you’re not actually using it).

When it comes to servers tho, that are made to run 24/7, power efficiency is of paramount importance, because, make your own calculations, a single watt, year-wise, will be going to cost you something. Using my estimates, and living in Italy, for each watt of consumption of any always-on electric appliance, I will pay, after a year, about 1.75€. Not much in itself, but try and multiply it by 100.

Also, the more power you use, the more you end up polluting the environment, and raising the earth temperature.

So, of the 100, 150 or 200W-whatever that your PC drains, how many do you think are really needed to run the PC, and how many go wasted in the form of heat? For each fan you need inside you PC, you are raising one notch the waste-index of the computer.

Giving for granted the waste of power that goes into the hardware itself, I am going to talk about what regards power supply units, the PSU boxes that nawadays are very nonchalantly sold in the range of 500W-1000W. Do yourself a favour and buy a wattmeter (kill-a-watt or whatever you call it) and measure the power absorbed by your home computer, it will surely be way lower than the maximum rating of the PSU it’s using; the more the real load is distant from half the nominal power of the PSU, the more you’re wasting in heat, since the efficiency of a PSU is a gaussian curve that has its peak (be it 70% in the old fashioned PSU’s, or up to 86% in the newest ones) at 50% the maximum load. In other words, if the hardware in your PC drains 100W, you should get a 200W PSU, even if nowadays it’s hard to find one, so it’s still better a recent 86+ 5ooW one.

For a simple comparison: my current server, before I did the complete switch described in my previous post, was equipped with a 350W PSU, the only one I had available, and it drained little more than 42W in idle. One day, I had too much free time in my hands and went scavenging for other PSU’s, so I had the chance to test it with a 250W PSU, and the drain lowered to 39W, while, with a very old, and supposedly very inefficient, 120W PSU, the drain went even lower at 37W. As soon as I bought and installed a PicoPSU, and connected it to a PSU brick rated 12V@3A that I had from an old external disk, I got a surprising 29W drain while in idle, a total saving of 13W off the 350W PSU. Translates to roughly 23€ savings in a year… the amount it costed me to buy the PicoPSU from the US. A break-even after a year is a good break-even, I say.

Uninterruptible power supplies are another source of waste, even if you would never suspect it: my car-battery modded UPS drained an additional 10W in idle; thinking about it, though, it makes sense, as with a UPS you’re doing an additional conversion, from 220V to 12V, or from 12V again to 220V (in the PC it goes even one more time from 220V back to 12V), which is intrinsically inefficient. In a year, 10W would mean 17€, more than what I paid for a 120W 12V fanless power adapter from China; I already had the car battery, so again a one-year break-even by exchanging the UPS with a plain AC/DC 12V adapter, good!