Introduction

Reducing the power usage of computers has become a priority to manufacturers of both processors and graphics cards. Not strange, as the last few years the usage of PC's has risen to alarming levels. We checked out different systems and the actual true power usage of these, and wil give you a few tips on how to reduce this and save on your energy bill.

Not even that long ago computers had PSU's that could supply a maximum of 200 or 250 Watts to the different components, and it was unthinkable that the system would actually need this amount of power. However faster and more complex processors, graphics cards and other components are responsible for a change in this.

Nowadays a computer store will not sell a PSU that can deliver less than 300 Watt to a system, and a 400 Watt PSU is not an oddity anymore. High end machines require a minimum of 500 Watt, and there are plenty of PSU's on the market that can even supply more than this. Does your machine acually use this amount of power or is it all overkill? Sadly we have to conclude that the asnwer to that question is yes, it does. There are prcessors on the market that alone require 130 Watts and high end graphcis cards that use less than 70 Watts are very hard to find. Of the 10 machines that we have tested there was one energy efficient PC that uses 132 Watt at 100% capacity, but the heaviest PC does require more than 400 Watt when in full use.

Test

To show you the power usage of a PC we built ten configurations, 5 with an AMD prcessor and 5 with an Intel processor. The configurations differ from ultra high-end (with the fastest processor and graphics card available) to a simple budget PC.

We have used the following Intel processors: Celeron D 351, Pentium 4 530, Pentium D 805, Core 2 Duo E6700 and the Core 2 Extreme X6800. From AMD we used a Sempron 3400+, a Athlon 64 3000+, Athlon 64 X2 4200+, Athlon 64 X2 5000+ and the high-end Athlon 64 FX-62. To make the test as real-life as possible we combined budget processors with budget graphics and balanced all systems as well as possible. The cards used from budget to high-end are a GeForce 7300GS, 6600GT, 7600GT, 7900GTX and in the high-end system we used a SLI or CrossFire configuration of two GeForce7900GTX cards and two AI Radeon X1950 cards.

To measure the actual power usage we used a EMU power meter. The usage is measured in seven situations; when the PC is off, hibernates, stand by mode (S1 and S3), idle, high processor usage (PCMark 05 CPU test), and high processor and graphics use (3D Mark 06 3D benchmark).

Measure the usage yourself

It is fairly easy to keep track of the actual power usage of your machines yourself, any respectable DIY shop sells power meters, and they do not cost the earth. The usage is measured in Watts, and the total usage over a period of time usually is stated in KiloWatts per Hour (KWh). Your energy supplier will state the amount of KWh's you have used in a period in your statements as well.

A KWh is the amount of power a device of 1000W would use in an hour, so for example if your PC would use 250 Watt average, it would use 1 KWh in the space of four hours. The costs of a unit will differ per supplier, however you can calculate the average cost of an electrical appliance according to the following formula :

(Average Usage / 1000) x Average usage hours per day x 365 x Price per kWh

A small example : A PC runs 24/7 and uses 150 Watt average, one KWh costs 11 pence, this would come to a total cost of : (150 / 1000) x 24 x 365 x 0,11 = £ 144.54 per year.

Our tests show that the cheaper power meters can be quite inaccurate, especially when measuring low values This makes it very difficult to measure the actual usage in "standby" or "off" mode. Higher values (when an applicance is "on" or running) are measured accurately by all meters, also the cheaper ones. The professional meter we have used in our tests does give accurate values for both low and high values.

Idle

We started with measuring the power usage of PC's when they are "idle", so on the windows desktop but no other applications being run. As expected the low budget Intel system with a Celeron D and low power graphics uses the least, only 97W in idle mode. The Pentium 4, Pentium D and Core 2 Duo systems all use similar power when idling, between 124 and 136W. The high-end system surprises a bit, with the two X1950 graphics cards in CrossFire the systems uses no less than 186W, so even when not gaming a high-end system with two graphics cards uses almost twice the power a low budget system uses.

The AMD machines do not differ hugely from the Intels, the Sempron machine is slightly better than the Celeron D and uses 89 Watts when idling, the Athlon 64 3000+ stops at 112 Watt and the two X2 machines are virtually the same as the Intel ones 134 and 136 Watts. Also a high-end AMD machine with two cards in SLI gobbles a lot of power when in idle, we measured 213W in the system with two nVidia GeForce 7900 GTX cards!

A lot of users do not switch their machines off when they are not using them, however our test already shows that this is not a very good idea. Even the most energy efficient machine (89W) uses 780 KWh a year when never switched off. A high end gaming machine like we have in our example with a FX-62 and two nVidia cards uses no less than 1865 KWh a year if you leave it on all the time.

100% usage

Next in line is the power usage when stressing the systems.We tested both 100% CPU usage as well as 100% CPU and graphics usage. It will not come as a surprise that the Celeron D machine does pretty well when 100% stressed, we measured 140 Watt. The Pentium 4 and Pentium D machines use around 180 Watt when we stress the CPU to the maximum and the Core 2 Duo does a bit better with 170 Watt. It is evident that Intels latest core does what it promises, the performance is a lot better than the Pentium 4 and Pentium D, but the system uses less power. The Core 2 Extreme system clearly has the highest usage, we measured 201 Watt at 100% CPU usage.

The AMD budget system did slightly better, it uses 125 Watt, the AMD 64 3000+ uses 138 Watt but we saw a signifcant increase with the X2 machines, they used 185 and 205 Watts. The FX-62 is the clear "winner" in this test, it used a whopping 282 Watt when stressed.

This test shows that even though it is fun to participate in Folding@Home, UD-Grid or RC5-72 distributing computer projects, it does mean that you PC will be running at 100% CPU capacity all the time and therefore use more power. This can cost a lot of money, as the most energy efficient machine uses 1095KWh a year and the high end gaming machine no less than 2470KWh a year!

Distribited Computing migh be fun, but it will cost you a lot of money when doing it all the time!

Graphics cards

High-end graphics cards these days use as much if not more power than the average processor. Die-hard gamers have actually resorted to using two cards in a SLI or CrossFire setup, which does give a huge performance boost, but is not very kind on the overall usage of the system. To find out how much power a graphics card actually consumes we ran the 3Dmark06 benchmark on our test machines, and the resukts are quite scary.

Our two budget systems have a GeForce 7300GS card, and the power consumption during the benchmark is nothing special, 132 Watt. The mid range systems with a 7600GT card use considerably more, a total between 190 and 200 Watt. The ultra high-end systems are the clear winners though, the system with two nVidia cards used no less than 331 Watt during the 3DMark06 benchmark, but the system with the two ATI X1950 XTX cards topped that easily with 407 watt!!

It is important to choose a very decent PSU for your system, but seeing these values it is no surprise that both ATI and nVidia recommend a minimum of 550 Watt PSU's for an SLI or CrossFire system.

a CrossFire-setup uses more than 400 Watt according to our test!

Monitor

Not only the PC consumes power, a monitor does as well, even when in stand by mode. In our tests it shows that a regular 17 inch TFT monitor uses around 30 Watts, a 19 inch model slightly more with around 40 Watt. In our product database it will state the avaerage power usage of the monitors when in use and when on stand-by. We think this is very usefull information when shopping for a monitor.

Standby: S1, S3 and S4

The most simple way to save on energy bills is putting your PC on stand-by, in stead of completely switching the system off. There are different types of stand-by however, all of them different and therfore all them them will affect the power usage when in stand-by. The different types of stand by are S1, S3 and S4 on a ACPI computer, and if you haven't changed anything in your BIOS when setting up your machine the PC will allways go in S1 mode when on standby. S1 is also know as "Power on Suspend", meaning that some parts in the PC are powered down, like the hard drive and optical drives. In the S1 mode however most other parts will just act as when in idle mode, for instance the graphics card will stay on, but wont show an image n the monitor, the processor will consume the normal amount of power and the moemory will maintain its state. S1 standby will not do much for energy saving, as the system will not consume much less power than when in idle mode.

Solving the above is easy, change the stand-by mode to S3 or Suspend to RAM, this way all parts of the PC (including case fans) will be powered down apart from the memory that is used to remember the state the PC was in before going on stand-by. The PSU is powered off as well, and only a tiny bit of power is switched through the 5V lines to keep the memory "alive". To use this mode all you haveto do is change the power option in the BIOS to S3 or STR. Not all hardware (especially older types) is compatible with this mode, which is the reason most BIOS'es have the S1 type as standard. If you have a newish PC though (up to around three years old) it should not be a problem using S3. As you can see in our table the power condumption when in S3 mode is only slighltly higher than when completely switched off.

S3 is the best mode for stand-by, and when you own a media centre PC that is switched on through a remote or by software or that powers itself up when recording TV it is definitely worth changing the stand-by mode to S3.

Hibernate

The final option, S4 or ‘Suspend to Disk', is not stand-by as in the true meaning of the word, but does allow you to start from where ever you left off when powering down. It basically means that the entire state of the memory and PC is written to a part on the hard drive before the PC is switched off completely. This means that power consumption in this mode is exactly the same as when powering down completely, between 5 and 10 Watt.

One of the biggest disdavantages of S4 is that it takes a lot longer to restore the power and that the C will not come on on a pre determined time, as you can do with S3. For media centres this mode is not preferrable, but for other types of machines it is.

Laptop users will benefit the most from hibernate, as the power is completely off and the battery will last forever in the hibernate mode. S3 will use some power and the battery will drain after a period of time, but S4 will not and you will still be able to continue where you left off once you are powered up again.

An easy way to set the PC to hibernate is pressing the "shift" button in the power down menu of Windows XP, the stand-by buttin will then automatically change to "Hibernate"

Pressing the "shift" button in the power down menu will change stand-by to hibernate

Power consumption Stand-by and Hibernate

In the table you will find the comsumption of the PC's in both stand-by (S1 and S3) and hibernation mode. S1 mode is pretty simple, the average consumption is only a fraction less than when the PC is idling, so it will not make a massive impact on your energy bill come the end of the year.

Systems that are in the correct stand-by mode (S3) are usually pretty efficient, we measured and average consumption of around 8 Watt. So when you PC is on standby this will cost you around £ 6 a year, having a media centre PC and keeping it on S3 stand-by is not a lot worse than sitching it off completely.

Hibernation means that teh PC is powered down, but this doesn't mean it does not use any power. Our tests show that machines still use between 4 and 8 Watt when switched off, so if you want to ensure that the system is really off you will have to switch the PSU off or use a on/off switch on the power socket.

Important: efficient PSU!

One of the most important items in a PC is an efficient PSU, as this changes the 220-220V form the outlet into the 12V, 5V and 3,3V that the PC uses. As in all processes the PSU will lose some of the power, so the actual usage from the socket is always more than what the PC actually needs. The smaller this difference, the more efficient the PSU is. According to the official ATX specification a PSU should be at least 72% efficient, however this is not a very good score. If your system uses 200 Watt and your PSU is 72% efficient this means your PSU actualy needs 277 Watt to power it all.

All the PSU's that are taken up in our product database have the efficiency figures in the comparison tables. The recent tests we did showed efficiencies between 66.7% and 93%. The worst PSU therfore uses 398 Watt to power a 266 Watt system, the best one only 286 Watt. Quite a large difference and well worth considering when shopping for a PSU.

The most efficient and the worst PSU in our database at the time of writing this article.

A PSU will perfrom most efficiently when the system uses around half what the SU can deliver. This means that when you purchase a OSU you will need one that is double what your system actually needs. For instance when you are building a Core 2 Duo E6700 system as we have in this test that uses 220 Watt you should buy a PSU that can deliver a minium of 450 Watt.

Saving..

How to save money on your energy bill is easy, firstly, do not use distributed computing programs that keep your system at 100% all the time, as this seriouly impacts your energy bill at the end of the year. Also downloading 24/7 is not as cheap as you think it is, as a PC will still use around £ 70 worth of power when you leave it on all the time. Leaving a PC on without using it is even worse, and if you wish to have quick access to your PC setting in on S3 stand by is a lot more sensible. If you really want to save power you can either switch the PC off or keep in in S4 stand by mode. If you think 8 Watt is still to much than you will have to buy a extension lead that you can switch off completely. This will not only enable you to switch the PC off but also swtch all the other equipment around the PC as well.

Conclusion

It is a good thing that both AMD and Intel are making more energy efficinet processors lately. Our test do show that the average power consumption of a system has increased massively over the last few years, and with power becoming more and more expensive people will need to start looking at what their system actually uses. The first place to look is off course the components that you purchase, like the PSU. Also not leaving your PC on unnecessarily or not participating in distributed computing projects helps a lot, and if you are not using your PC for a short period of time put it on stand by or hibernation.

Testresults

The follwing tables show you the results of our power consumption tests.

CPU	Intel Core 2 Extreme X6800	Intel Core 2 Duo E6700	Intel Pentium D 805	Intel Pentium 4 530	Intel Celeron D 351
Chipset	Intel 975X	Intel P965	Intel 945P	Intel 945P	Intel 945P
Memory	2 GB DDR2-800	1 GB DDR2-800	1 GB DDR2-667	1 GB DDR2-533	512 MB DDR2-533
graphics	ATI Radeon X1950 CrossFire	nVidia GeForce 7900 GTX	nVidia GeForce 7600 GT	nVidia GeForce 6600 GT	nVidia GeForce 7300 GS
Hard disk	WD Raptor 74 GB	Samsung HD160JJ	Samsung HD160JJ	Samsung HD160JJ	Samsung HD160JJ
.	.	.	.	.	.
3DMark06	407 W	220 W	196 W	192 W	132 W
PCMark05 CPU	201 W	170 W	180 W	184 W	140 W
Idle	186 W	136 W	124 W	133 W	97 W
Standby (S1)	.	144 W	110 W	120 W	78 W
Standby (S3)	12 W	9 W	6 W	6 W	6 W
Hibernate	10 W	7 W	5 W	5 W	5 W
Switched Off	10 W	7 W	5 W	5 W	5 W

CPU	AMD Athlon 64 FX-62	AMD Athlon 64 X2 5000+	AMD Athlon 64 X2 4200+	AMD Athlon 64 3000+	AMD Sempron 3400+
Chipset	nVidia nForce 590 SLI	nVidia nForce 570 SLI	nVidia nForce 570 SLI	nVidia nForce 4	nVidia nForce 4
Memory	2 GB DDR2-800	1 GB DDR2-800	1 GB DDR2-667	1 GB DDR400	512 MB DDR400
Graphics	nVidia GeForce 7900 GTX SLI	nVidia GeForce 7900 GTX	nVidia GeForce 7600 GT	nVidia GeForce 6600 GT	nVidia GeForce 7300 GS
Hard disk	WD Raptor 74 GB	Samsung HD160JJ	Samsung HD160JJ	Samsung HD160JJ	Samsung HD160JJ
.	.	.	.	.	.
3DMark06	331 W	248 W	194 W	155 W	132 W
PCMark05 CPU	282 W	185 W	205 W	138 W	125 W
Idle	213 W	136 W	134 W	112 W	89 W
Standby (S1)	.	121 W	95 W	82 W	84 W
Standby (S3)	8 W	8 W	6 W	11 W	7 W
Hibernate	4 W	6 W	5 W	8 W	6 W
Switched Off	4 W	6 W	5 W	8 W	6 W