A Matter Of Simple Addition
While there are many online calculators out there meant to help you figure out how much power your build will use, they all share one major weakness: they are all based on the build’s maximum power requirements. They then take into account the PSU’s maximum efficiency at a load of 50 to 55 percent and make a (generous) projection based on those numbers. The drawback is that this method does not consider power consumption when the computer is idle. As we pointed out on page three, that’s where things tend to get really ugly, efficiency-wise.
The following table is meant as a guideline, showing you how much a certain component is likely to use under a specific load. If you know the exact figures for your components, you can just plug them in instead and do the math.
| Component Type | Description | Idle (W) | Load (W) | Number Installed |
|---|---|---|---|---|
| CPU | Current dual-core CPU Current dual-core CPU, overclocked Current quad-core CPU (Mid-range) Current quad-core CPU (High-end) Current quad-core CPU, overclocked (High-end) Older dual-core CPU (AMD) Older dual-core CPU (Intel) Older single-core CPU | 20 25 35 40 45 35 55 35 | 65 90 95 125 140 90-125 125-140 60-90 | 1 |
| Motherboard | Current microATX board without graphics Current microATX board with graphics Normal mid-range board without graphics Normal mid-range board with graphics High-end board High-end board, overclocked | 15 30 20 25 35 40 | 25 40 35 50 45 55 | 1 |
| RAM | Current DDR2 or DDR3 RAM, per 2 GB module Current DDR2 or DDR3 RAM, per 4 GB module Overclocked RAM per module (estimated) | 2 3 4 | 4 5 6 | 1 - 4 |
| HDD | Conventional hard drive | 2 | 8 - 10 | Varies |
| SSD | Current solid-state drives | 1 | 4 | Varies |
| ODD | DVD-ROM only DVD burner Blu-ray reader/burner | 1 1 2 | 6 10 12 | 1 |
| Fan | Normal CPU fan, mandatory Silent case fan Performance case fan | 1 2 3 | 1 - 3 2 3 | 1 Varies Varies |
| Add-In Card | External sound card TV card Controller card | 2 1 1 | 8 2 - 5 2 | 0 - 1 0 - 1 Varies |
| Graphics Cards | Current office-oriented card Current mid-range gaming card Current high-end gaming card | 10 - 16 16 - 30 25 - 35 | 35 - 75 75 - 180 180 - 375 | 1 1 - 2 1 - 2 |
| Other | Cold cathode lighting, modding parts, etc. | See manufacturer info | Varies | |
Now that we have an idea of how much power each of these components draws, calculating the overall power consumption when idling and under load is simple. Based on this information, we will define a power window in the next chapter and choose the best PSU for the job based on their individual characteristics.
- Brought To You By Granny’s Radio
- How A Switching Power Supply Works
- Efficiency, Efficiency, Efficiency!
- Of Power Factors, Apparent Power, And Effective Power
- How To Spot An Efficient PSU?
- Don't Get Burned: Safety Before Stinginess
- How To Determine Your Power Requirements
- The Power Window Is Important
- Example 1: The Office PC
- Example 2: Mid-Range Gaming PC
- Example 3: The Enthusiast’s System
- If You Don't Like Our Advice, Buy A Fire Extinguisher

unfortunately that is not always the case, you could buy a resonably expensive Thermaltake TR2 RX series and end up with a piece of crap.
i want
Would like to see more articles written like this.
Well written, and well done !!!
i want
No one really listens to this.
unfortunately that is not always the case, you could buy a resonably expensive Thermaltake TR2 RX series and end up with a piece of crap.
Now I got a CM PSU and no problems since!
Helps keep the Phenom under control.
most good psus are able to out live a pc build under 24/7 use, some able to last 10 years.
where i live power costs 11 cents per kwatt. it should be 9 like my state average, but whatever, i wont go into that further here.
now with 11 watts difference, over the coarse of 1 year under 24/7 use, that comes out to about 10$ and out national average is also 11.2 cents, so rounding down to 11 is appropriate.
now here are some numbers, based on estamates.
11 watts - 10$ a year
22 watts - 20$ a year
33 watts - 30$ a year
and lets go with a pc thats built right lasts 4 years, and is used for 5 (waiting on parts to be released and such)
base/4year/5year
10/40/50
20/80/100
30/120/150
basically you have to look at the long run, and anticipate a mid cycle upgrade, such as new gpu, with 50-100 more watts headroom.
lets also assume that you turn the pc off at night, some people do, but leave it on all other times. even at half the cost, on the gameing, and mid range, its more cost effective to buy the better psu, the highend though, used a 750 watt Corsair, there should be a 550 or 600~ watt one that would be cheaper and better fit in line with the other psus tested.
more to the point. we needs a toms chart for psus, right now we have a efficiantcy chart, but what we need is a build chart.
one that takes the psus, and puts them in a standard pc enviorment, like the ones here, and measures the watts used.
than gives us a baseline like this one is 100% than how much more the ones above use in %, than in numbers next to it, and than figuring the annual power cost for the extra watts they use, and find out which ones really the best for your builds, assuming a 4 year build cycle with a 5 year use, like i did above.
i honestly think that could result in an interesting benchmark.
Yes but in most cases if you buy something cheap the vast majority of the time it's going to be cheaply made so therefor it will break a lot faster then something you buy at a preimuim that's has quality. Your odds of getting a lemon Bently is a lot lower then buying a Dodge that's a lemon. of coure you pay a hell of a lot more for the Bently then your typical dodge but that's because you are paying for the high preimum of quality and craftsmenship that goes into it vs something that is made mostly by a machine.
Would like to see more articles written like this.
Well written, and well done !!!
Beware of retailers that don't pass on full specs. Always check the manufacturer website for specs. Some retailers seem to forget to add complete descriptions for PSUs (and products in general too) which they have to order, but seem to add more info to the ones they have on stock. Probably to make them look better so they get rid of it faster. Living in Romania I already have "blacklisted" quite a few retailers from which I'll never EVER buy, based on this solely alone. It was actually very subtle, and it wasn't that they simply "forgot" either. As soon as I noticed this discrepancy I started to question their entire business model. Didn't even bother to tell them (didn't want to) because I'd rather have some other smart person catch on to their scheme instead of guiltying them into correcting it.
CM PSUs aren't that amazing either
http://database.ul.com/cgi-bin/XYV/template/LISEXT/1FRAME/index.htm
Look for the symbol on the PSU label that looks like a backward "R" next to a "U". Under that symbol, you will find the letter "E" followed by a 6-digit number. Put that in the UL Online Certification Directory on the line that says "UL File Number" and it will tell you who that file number belongs to. For example, you would find that some Silverstone PSU's are made by Silverstone, some are made by "Enhance Electronics Co. LTD." and some are made by FSP.
Fast forward another year sitting in a 100yr old dorm room with a power grid that was not meant to run the amount of draw that we were all pulling, and we were all loosing parts left and right. This was long before active power correction was normal, or even affordable. But I lost my video editing rig right in the middle of finals week, and it was entirely due to the power available (though not the power supply that lasted me several more years without problems when I moved out of the dorms).
On my last build I bought a very nice 500W power supply that has all of the modern niceties of voltage regulation and power protection (as well as the power connectors that the old supply did not have), but now that I am looking at doing a new build next year I am not sure if it will be big enough. I think it will be fine at first, but I am going to do SLI down the line, and according to the math here I am going to need a peak of ~900W available, meaning I need a ~1000+W PSU... and that is going to hurt the wallet