What Devices Produce the Largest Power Draw in PCs? 21
bcboy asks: "Considering the energy situation in California, I borrowed a watt-hour meter and have been assessing my energy use, and waste. My daily use was almost exactly the California average. So far I've cut half of it, just by eliminating waste. Digital equipment was a full 20% of my usage. Idle hubs and DSL modems pull a surprising amount of power. So do idle HP scanners that conveniently lack 'off' buttons. The laptop was pulling 20 watts while 'off', 24 hours a day. The biggest lesson from this experiment has been that things that are 'off' frequently pull significant power. But my question is (now that my desktop PC is the biggest load after the electric drier and the fridge) has anyone assessed power usage inside the PC case? What's pulling the power? The fans? The hard drives? The CPU? The memory? The cards?" It's easy to determine the power draw of some devices in PCs by looking at the labels for the specs, but some devices, like memory, may not have such specs that are readily available that may draw quite a bit more of your wattage than you'd expect.
The power supply (Score:2)
Re:My contribution (Score:3)
The lighter the load is, the less time is spent charging the inductor from the mains and it is discharged slower by the load. So the technique is adaptive to varying loads. While the load has the inductor, very little power is drawn from the mains.
There are variations on the technique to try to minimize the RF problems with abruptly switching between the mains and the load repeatedly, and the unlocked frequency of the switching in the plan outlined above, but that's basically how they work.
Re:In short (Score:1)
For conventional CRT monitors, don't get to carried away with the powersaving settings. Having the interval before DPMS kicks in set too short will cause the monitor to cycle on and off *much* more frequently throughout the day if the computer gets used a lot.
This could eventually cause something to break (unlikely), but the real issue is that on power-up, a monitor will draw quite a bit more power than when it is running normally. Do this enough times in a day, and you'll end up using more power, even though the total time the CRT spends in "sleep" mode (or whatever you want to call it) is a bit more than it was before...
"Intelligence is the ability to avoid doing work, yet getting the work done".
My contribution (Score:5)
I'll add some more information.
When you feel those wall-warts, they are usually very hot. The reason for this is inefficiency. The cheaper a power supply is, the more inefficient it will be, in general. The reason for this requires some explanation of how power supplies work.
First, you'll usually have a transformer. They are notoriously inefficient because of how they work. They are basically two coils with differing numbers of turns next to each other. The input/output ratio is related to the ratio of turns. Basically, power is applied to the primary coil, which has a resistance. The secondary coil actually generates power from the EM field emitted by the primary. It doesn't take a genius to realise that most of the power applied to the primary is lost in the air.
So off the secondary you will have an AC voltage lower than what you put in (here in NA it's 120V RMS). You have to convert that to DC somehow. Usually it's a bridge recifier, which will drop the voltage by about 1.4V (from peak-to-peak, not RMS) for it's own operation IIRC. This isn't something to be worried about, however. Bridge rectification is the most efficient type of rectification I've ever seen. It basically takes the negative parts of the AC waves and makes them positive.
So now we almost have DC. There's one more piece. The voltage is still bumpy. Draw a sine wave and imagine taking the negative part and inverting it on the positive side. Our devices need clean, flat DC, especially if the circuit is frequency dependent, like radios, sound cards, hard drive buses, etc.
The last piece is the regulator. These are quite complex so I'll only briefly describe what they do. You want a voltage of, say, 5VDC. You get a 5V regulator, say an LM7805. Your input voltage must be, at it's lowest point, 2.3 volts higher than your intended output for it to work correctly. This means that you'll have to put in at least 8V to be somewhat comfortable. 9-12V is quite common and 40V is the maximum. The higher the voltage you put into it and the more current the load draws from it, the harder that little guy works on regulating it. Many regulators require heat sinks to dissipate all of that heat.
Oh, but wait, how complex is your average wall-wart. I've rarely seen the correct voltage come out of one. Try measuring the ouput of a Nintendo wall wart sometime. They rarely have "real" regulators in them, often a couple of transistors. At work there's a "3V" wall-wart that outputs 6V. It blew a set-top box that required 3V. We have some Elastic DSL modems that use real 5V regulated wall-warts (I was impressed). If your device says 5V or less on it, you'd better be damned sure that's what you put into it because it's highly likely that it's expecting real regulated DC on the input, or else it would have said 9V and had a regulator inside the device. See why that voltage is so popular, now? You could also put 40V into it, but I doubt the heatsink chosen for 9V-to-5V would provide proper temperature dissipation for 40V-to-5V. I don't recommend plugging a 16V into your 9v device, either.
Finally, after all of that, your device is powered, but you've wasted a lot of energy getting there. Is there a better way?
Of course, the expensive option! The good news: you're already using it. That power supply in your computer is a switching power supply. These are much more efficient but a lot more complex. Generally, they often have much more intelligence.
There's a good book at Radio Shack on power supplies that explains all this, including how to build them. There's a section on switching power supplies as well.
I've posted here before, showing my disdain for wall warts. What came out of it was a project [funktronics.ca]. Basically, the plan is to eliminate wall warts altogether and distribute power to various devices from one "power server" similarly to how it's done inside your computer. The main difference is that that devices actually request the voltage they want. You will also be able to use it on almost any existing device.
Ok, that went on a little long, but, as you can see, I'm sort of passionate about the topic.
In short (Score:5)
- the screen : those things run from 70 to 150 Watts/hour depending of size, which is really huge. A flat panel will go 20 to 30 Watts.
- the CPU - if you have one of the latest beast (Athlon 1400/Pentium 4 1,7 Ghz) it can go above 50 Watts during intensive computations
- the graphic card can be a power drain - when the GeForce or Voodoo 5 arrived some motherboards had to be redesigned to provide enough power to the board.
- DVD-ROM, hard drives or sometimes fans (if you have 4 or 5 of them, it can stack up to over 10 Watts )
Measure the load with the watt hour meter (Score:1)
That will give you interesting numbers that aren't very usefull. Running with less RAM and/or disk isn't progress. but if you do it, post the numbers!
More usefully, go to an overclocking website and reverse the directions. Underclock and see if the difference is significant. Likewise, see if you can spin down idle disks. Report any results, positive or negative!
I've got a machine that's sole purpose in life is to do disk to disk backups of my other systems. Would it be worthwhile to clock it down until it can just barely saturate the ethernet? I've got too many other projects to find out today, but if there's a big win, I'll try.
Re:My contribution (Score:2)
Mark Duell
Unused cycles (Score:2)
I use a HLT program on my Dell PII laptop (that runs HOT!), and I have noticed that when I send HLTs to the processor while idle, it does cool down quite a bit. Is that really a significant drop in power usage? Is thermal output in direct relation with power usage on processors?
Look at what gets hot (Score:3)
Really- if you want to save power through your computer- turn it off! Yes, places like SETI@Home can take advantage of your "unused cycles" but don't ever forget that somebody still has to pay for it.
I haven't seen any numbers recently, but I wonder what a chart of watts/instruction would look like as we progress through the processors of history. Are we getting more efficient, less, or just holding even?
Re:Unused cycles (Score:3)
Absolutely! The power has to go somewhere. The power can be stored, but usually that's too inefficient to be worthwhile, so we just radiate it into the environment, mostly as heat. The less heat a computer makes, the less power it is using. Yes, some of that power is going into making other forms of lower frequency EM radiation, but that's insignificant. Eventually all the energy we expend ends up as heat. The laws of thermodynamics are kind of depressing. Stolen from the web somewhere:
The simplified Laws of Thermodynamics: 1. You can't win; 2. You always lose; and 3. You can't quit.
Re:My contribution (Score:1)
In a word - no, I'm afraid your wrong on this one, your basic coil transformer is incredibly efficient, they have been around for years and have been refined to a point of near perfect efficiency... I found the following on a website about transformers:
"But the transformer is a passive device -- it cannot add power. Power is equal to the product of the voltage and the current. If the voltage increases the current drops. A high voltage and low current exits the transformer carrying almost the same amount of power along the transmission lines that the initial low voltage and high current did. Most transformers operate at high efficiency, under normal conditions, transmitting about 99% of the power that enters them. (About 1% of the power is lost in heating the transformer.) "Smoothing and regulating the power can produce a lot of extra heat / inefficieny but but the coil section of the transformer is just great :)
Although I do agree that the best way to find power consumption is to find the things that emmit the most heat though...
Re:Unused cycles (Score:2)
hlt stops the cpu (*IF* I recall my assembley it stops it until a IRQ comes in) thus there should be (very few) gate switches, thus not much power consumed :)
Good on you (Score:3)
Re:Damn screensavers. (Score:1)
if a processing-capacity-challenged (most of them, depending on where you are) student/employee/other user sees some computers with login screens and other computers with no screens, he/she/it will assume the computers with no screens are broken, and ignore them.
Hot doesn't equal power hog (Score:2)
The best way is to go to the mfg sites for the parts, or look in any owner manuals, and get the power data from the source. Most 3-1/5" HDD's are in the 5-15w range. Most fans are in the 0.5-1w range, CPU chips are typically in the 15-65W range, the support chips for them are typically another 3-15W. Older motherboards with a lot of TTL and PLD's typically have another 3-10w of bus drivers, PAL's, terminating resistors, and the like. Switching power supplies are only about 80% efficient - so what ever you have in current power draw, you need to add an additional 25% for power supply losses.
Newer "Green" motherboards normally have just a few large low power CMOS VLSI parts on them, have BIOS support for cycling down HDD's when idle and idle power control for monitors under windows drivers. Linux isn't quite there on the power friendly side of the equation. Blazing fast is always power hungry - current generation mid-performance system will generally eat less power than either older systems or new fast systems.
Notebooks are a great alternative - mid performance, with max power savings. LCD's have power hungry back lights - turn the back light down to the lowest usable brightness and conserver power while getting the best life out of it.
Screen savers are nice, but make sure they don't stop the energy saving features from shutting down an idle display.
Re:Damn screensavers. (Score:1)
Re:Hot doesn't equal power hog (Score:1)
My two linux boxes both spin down their hard drives when not in use, and the one with a monitor powers down the monitor after a few minutes of dis-use. Linux is at least as there as windows when it comes to power savings.
Anything with an induction motor (Score:1)
Please don't contribute to misinformation (Score:2)
- Transformers are not "notoriously inefficient". They can have efficiencies close to 100%.
- The inefficiency in wall warts is due to the design tradeoffs, not the underlying physics.
- Not all transformers have differing numbers of turns on the primary and secondary windings.
Isolation transformers are a case in point.
- The salient feature isn't that the coils are next to each other (they may be on opposite sides
of the transformer core, or they may be concentric). What matters is that they share almost
all of their magnetic field (a feature known as "mutual inductance").
- It's not "related", the voltage is directly proportional to the number of turns.
- There are superconducting transformers, which have no resistance. They work under the same principles.
- That's a gross and misleading oversimplification.
- It doesn't take a genius to avoid making silly mistakes from oversimplifying things, but you've
not managed to steer clear of those pitfalls.
- Air is an essentially lossless medium for magnetic fields. The
excess losses in a transformer are almost entirely due to:
- Undersize transformer cores which require greater magnetizing current, also leading to
- Hysteresis and saturation losses in the core, plus
- Eddy-current losses from core laminations which are too thick (but they're cheaper), and
- Under-sized winding wire which has excessive resistance, causing IR losses (compounded
by the extra magnetizing current required by under-sized cores).
You can see that if you use a properly-sized and properly built transformer you can minimize the losses you hate. Unfortunately wall warts are built to be cheap, not efficient. (I never intended to write this much on the subject...)- The combination of the primary winding and the core material (if any) have inductance.
Inductance resists the change of current; for a coil of zero resistance, V = L dI/dt.
- If you apply a sinusoidal waveform to an inductor, you'll get a current through it which
is inversely proportional to the frequency and proportional to the voltage, and lagging by 90 degrees (for
the lossless case): if the applied voltage
is Vmax cos(t), the current is I = Vmax sin(t)/L. The magnetic field uses no
power, but losses in the windings (IR voltage drop) and in the transformer core (hysteresis and eddy
current losses) do use power.
- There's also a time-varying magnetic field associated with this current. This field induces
a time-varying voltage in any conductor exposed to it (including the primary itself).
- If you allow a current to flow through one of these other ("secondary") conductors in the
direction of the induced voltage, it will create another magnetic field. More to
the point, the induced voltage will always be in the same direction as the voltage in the
primary coil, but the field created by the secondary (driven) coil will be opposite from the
field created by the primary (driving) coil.
- Since the field in the secondary coil opposes the field created by the primary coil, the
reverse voltage (the voltage resisting the change in current) on the primary coil is reduced. Ergo,
more current flows in the primary coil until the two voltages balance again.
(That was more than I really wanted to write.)One last thing about wall-warts. They may have a nominal voltage printed on the nameplate, but as they are likely to be designed for the specific appliance they are running they may not produce their rated voltage under a different load (think of resistance in under-sized, cheap windings). The wart which makes 3 volts running a Nintendo may be 6 volts open-circuit. Caveat substitutor!
--
Desktop Power (Score:2)
Monster CRT's, AMD chipsets, DVD/CD drives, scanners, laser printers, all use a ton of power.
Here's my idea - everyone use laptops!
90% of users don't need Desktops. Just using more power, desk space, etc. Not to mention they suck to carry around.
hmmm.... (Score:1)