50% Load Myth

So apparently some people read my efficiency article. And they read the part where I said most PSUs get their highest efficiency between 40% and 60% load. Then they told that to some other people. Who took that to mean that you should never use a PSU over 50% load, or that you should size your PSU that way, or whatever garbled version of the message they got.

This idea fails for one simple reason: the amount of power a computer uses is not constant, and will never, ever pull the "max" wattage figures myself or another expert may state except under extreme stress testing conditions.



Let's take my computer. An overclocked Q9550, a Radeon 4870, couple drives, sound card, accessories, etc.

My absolute peak DC wattage pulled from the power supply is approximately 325W. This is the figure one would see if they put a few DC clamp ammeters around the cables leaving my PSU and measured all the current leaving it WHILE I was running LinX, Furmark, and HDTune all at the same time.

My power usage while the system is idle on the desktop will be approximately 100W.

My usage under heavy multitasking (playing a movie, a game minimized in the background, multiple firefox windows, moving files around, etc.) might be 150W.

My usage under heavy gaming conditions (Crysis or another demanding game running maxed, or running CPU+GPU F@H) would like be approximately 250W.


Now, let's take two power supplies. One is a 650W, one is a 400W. Let's assume they have the same efficiency curve over percentage load. At 10% load they get 78% efficiency, at 20% load they get 80% efficiency, at 50% load they get 82% efficiency, at 100% load they get 80% efficiency.


On the 650W my peak loading condition will be at exactly 50% load. I will be getting about 82% efficiency. At idle I will be at about 15% load--call it 79% efficiency. Multitasking will be 23% load--80% efficiency. Heavy gaming will be 38% load or about 81% efficiency.

On the 400W my peak loading condition will be at about 80% load. I will be getting about 81% efficiency. At idle I will be at about 25% load--call it 80% efficiency. Multitasking will be 38% load--81% efficiency. Heavy gaming will be 63% load, or about 81% efficiency.


Let's average those figures.

650W average efficiency = 80.5% efficiency
400W average efficiency = 80.75% efficiency

Hey, a tiny difference in efficiency--in favor of the smaller PSU. How about that. Now consider that a quality 400W power supply with that kind of efficiency curve can be had for about $40, while a 650W PSU with that kind of efficiency might run $80. A $40 difference. So that's $40, plus slightly (slightly) higher electricity bill, by going with the 650W instead of the 400W.




Then you ask, "But what about upgradability?" Great point! Except that nine times out of ten some gof comes along and says, "No, you want to run your PSU at 50% load for the best efficiency!"

... Thus starting the cycle over.

Besides, from what I've seen the vast majority of people who talk about their upgrades, never really get around to it. Unless you know you're going to upgrade, and you have it already budgeted out and know where the money is going to come from... Don't plan around it. We'd all like to upgrade to dual GTX580s some day, but only a few of us really will. Be realistic. Be smart.

And if you do pick your PSU to match future upgrades, then at the very least don't waste money needlessly upgrading your overkill PSU again as soon as you get a new graphics card because someone who doesn't know what they're talking about says "efficiency". Ok?




Further Reading
Edited by Phaedrus2129 - 12/3/10 at 10:05am

You have a dual GPU graphics card. Which consumes a lot more power than a single 4870, even at idle. Plus more drives, more of everything, and a higher volted CPU. Plus that's at the wall power usage, meaning after efficiency losses. I don't know exactly what Antec PSU you have (probably the TPQ-1000?) so I'll just guess 80% efficiency. That means you're using 260*.8 = approx 208W from the PSU, and you have a lot more hardware than I do.



So sounds about right.

A whole lot really do... And of course this will vary from PSU to PSU. The efficiency curve varies with design and components used. I just used a generalized, average curve. And I even gave it a higher than average 10% load efficiency to boost the 650W. And it still comes out worse.

More just a general idea that got spread into a bunch of peoples' heads and is getting very annoying.