Choosing the right PSU for your PC

Choosing the right PSU for your PC

Summary: So, just how big a power supply unit does a PC really need? In this post I'll show you how you can work out how big a PSU your system needs and some of the considerations that you need to keep in mind when buying a PSU.

TOPICS: Hardware

So, just how big a power supply unit does a PC really need?  In this post I'll show you how you can work out how big a PSU your system needs and some of the considerations that you need to keep in mind when buying a PSU.

It seems to me that one of the most misunderstood aspects of building or upgrading a PC is the PSU.  I've read more rubbish written about the PSU than almost any other hardware component in the PC (notice that I said hardware there, when it comes to the amount of rubbish written about software, all bets are off!).

Obligatory warning and disclaimer: Messing about with PSUs is potentially hazardous - they can deliver lethal levels of charge.  If you are in any way unsure about what you are doing or how to use your equipment, consult a qualified electrician.

Any PC has a specific amount of power that it needs to draw to work.  What you need to have in mind when choosing a PSU is the average running load that your PC consumes when running.  Peak load can come into play sometimes and this can be significantly higher than the average running load, but as a rule high loads are more likely at start up than any other time.

So, how do you work out your average load?  Well, thankfully the days of having to pull up specs on the individual components are over.  On the web there are a number of really good and up-to-date power supply calculators that you can use.  You just enter the spec of your PC and the calculator will work things out for you.  One of the best online power supply calculators that I've come across is made available by Antec.  By using the simple form you can quickly and easily find out how many watts your PC will need and use this information to choose a PSU that delivers this.

Next -->

What the power supply calculator on Antec's site is very good at is making you think not only about the standard components that are inside your PC (CPU, RAM, hard drives and so on) but also non-standard items such as water pumps for a cooling system and cold cathode lights.  Also, it allows you to factor in that you hook up to your PC via USB and Firewire.  The calculator generates some very good data that you can work with and even allows for you to figure in surge compensation and capacitor aging (as the electrolytic capacitors inside the PSU age they lose some of their initial capacity). 

For example, I entered in the information for the following system:

  • Intel Core 2 Duo E6700
  • Four sticks of DDR2 RAM
  • Twin NVIDIA GeForce 8800 GTX SLI
  • Two SATA hard drives
  • One DVD-RW/DVD+RW drive
  • One SoundBlaster sound card
  • Four USB devices drawing power from the system
  • Three 80mm fans

The recommended PSU wattage for the system was 487W.  Factoring in 15% surge compensation and 20% for electrolytic aging and this gave a total of 658W.  For a system with the above specification I would have recommended a 650W PSU, and it seems that I would have been right.

This calculator is also handy for finding out how far you can upgrade a PC that you bought without having to fit a beefier PSU.  I've found that as a rule manufacturers save money by fitting PSUs that are able to cope with the demands that the system places on it but not much more.  This means that if you want to take your Dell or HP PC beyond the spec much, you'll have to bin the PSU.

I know a lot of people who build PCs based on a small PSU.  The idea is that it is cheaper and the PC doesn't waste power.  While it's true that it might be cheaper (a good quality 400W PSU is cheaper than a 600W PSU of equivalent quality), the idea that a bigger PSU is less efficient is false.  The PC will only draw the power it needs and no more.  In fact, having a larger than required PSU can be an advantage since it doesn't need to work as hard to deliver power at the lower end of the scale and it can even allow your PC to ride a very brief power dip without a hiccup.  While I wouldn't recommend putting a 1KW PSU in a PC that only needed 500W, remember that unless you buy a high quality 1KW PSU, it's only really likely to give you, at best, 75-80% of the power they claim anyway (tests show that the failure rate of PSUs at full load is pretty high).  And if you factor in that cheaper PSUs lose efficiency much more rapidly than a quality one as they heat up, it becomes obvious that buying cheap can mean buying twice.

Next -->

So how do you spot a good quality PSU?  There are a number of ways.  First off, buy a decent brand name.  No-name PSUs are generally inferior and if you plan on making your PC work for its keep (pushing it hard, leaving it on 24/7) then it's false economy. Another good way to tell a good PSU from a cheap one is weight.  Cheaper PSUs will be lighter and feel flimsier than quality PSUs.  Also, try to get a PSU that has a 12cm (4.8") fan which doesn't need to rotate as fast to keep the PC cool, making it a lot quieter.

Also, remember to check the label on the PSU.  Most good PSUs will list the ratings for the +3.3V, +5V and +12V rails.  You can see an example of this kind of label here on the Newegg site.  These ratings will be optimistic, but provide a starting point.  Make sure that the power rails (especially the 12V rail) gives you ample power.  Also, read the label carefully.  For example, two 12V rails rated at 22A doesn't mean that you've got 44A to play with.  You have to take the total Watts on the +12V rail and divide that by 12 (so on the label I linked to above the 12V rails have a total rating of 384W, so 384W/12V = 32A).

Also, don't be tempted to test the rails using a multimeter - most multimeters can't handle above 10A.  If you want to test the power output either rely on a software monitor which can read the values directly off the board such as Motherboard Monitor (anyone know what's happened to the official Motherboard Monitor site?) or read the values from the BIOS "PC Health" screen, if available (however, be aware that these values aren't always that accurate - I've seen BIOS updates which change them significantly). 

If you know what you are doing with resistors and don't mind doing a little math you can use a multimeter, but be careful not to blow it up!  Unless you know 100% what you are doing, don't try this at home!  It's far safer to test the voltages at the rails - if this fluctuates, you can be sure that current does too.

Also, while talking about PSUs I should probably mention that having a good quality UPS that conditions the power going into your PC is a good idea if you want a stable, reliable system.  Most PCs are happy with input voltages which vary by ±10%, but beyond that things get hairy.  A good quality UPS gives you the piece of mind that under and overvoltages aren't going to trash your system.

Thoughts?  How do you choose your PSU?  Do you buy cheap or do you spend a little extra?

<< Home >>

Topic: Hardware

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • I uae a table

    This one
    Component Wattage Required
    Motherboard 15-30
    Low-End CPU 20-50
    Mid To High-End CPU 40-100
    RAM 7 per 128MB
    PCI Add-In Card 5
    Low To Mid-Range Graphics 20-60
    High-End Graphics 60-100
    IDE Hard Drive 10-30
    Optical Drives 10-25

    This gives me an idea of possible wattage. But I also add about 100 watts to the range to just make sure anything I add can be handled.
    Actually if your pc only uses(say 250watts) but you have a 500watt PSU. This PSU will actually be more effiecent and what this means is that there is less heat generated in the power supply.
    Michael L Hereid Sr
    • I like to use a power meter

      Most of my Core 2 systems use no more than 190 watts peak.
      • So does this power meter give instantanious

        peak power readings of when turned on. Sure it may show 190 watts when running but when turning on it can be as high as 4x that value.
        Also you may have peaks of power usage that are of such short cycle that the power meter does not have enough time to react and those can be 2x that value.
        Besides that at 190 watts not all the hardware is using peak power even harddrives may only use as you say 10 watts but at idle not when being acessed.
        I have over 40 years of experiance in Electronics.
        Michael L Hereid Sr
        • 4x? What are you smoking?

          "peak power readings of when turned on. Sure it may show 190 watts when running but when turning on it can be as high as 4x that value."

          I've put up to 300 watts of peak load on the Seasonic 330 watt PSU on two separate machine for more than a year with no problem. You're going to tell me it's under powered for my 190 watt peak load?

          FYI, the system has a temporary 140 watt spike during bootup and I can assure you it's not "4x". The system idles at 125 watts and peaks at 190 watts during game play. My own meter doesn't have graphing but I've used one of those expensive meters with graphing and PCs just don't have that kind of spike you're talking about.

          "I have over 40 years of experiance in Electronics."
          Here you are questioning my methods when you have not done a single actual measurement on a PC. What does ?experience? have to do with doing the scientific experiment? Absolutely nothing because you?re just guessing.
        • You're just guessing without a scientific method...

          Sorry (and no offense) but 40 years of working with electronics and with no scientific method and conclusion doesn't make sense when working with power supplies...

          You've just lost me.

          For me, understanding and conducting measurements for adding up all the components' power requirements doesn't mean I'm an IT Technician/Electrician as I've never had a work experience before.

          Conducting a scientific method on measuring all the components' power consumption isn't guessing; it's about experienmenting/testing every aspect of all the computer components' power consumption during the idle and load session (or testing if you prefer). You have to do it like a few times to ensure accurate results.
          Grayson Peddie
  • Watch out, George Ou will tell you you don't need that much ;)

    Watch out for fellow blogger George Ou - he'll tell you you probably don't need more than 300 watts as long as it's a continuous power supply.

    As for myself, I have a 500 watt continuous power supply by Ultra. He may call it overkill, I call it future proofing. Who knows? I may SLI together dual GeForce 8800s in the future :).
    • And oh, yeah . . .

      And oh, I forgot - I do indeed use a UPS. Needed to, I was getting brownouts in my neighborhood, and that killed my first power supply (collapsing magnetic fields cause spikes in transformers).

      So now I use a UPS - provides clean power, plus a few minutes to shut down my computer in case of a complete power outage.

      My multimeter is rated at 20 Amps, BTW. Dunno if that's enough. Not that I need to test it anyways, as my power supply is pretty overkill for my current configuration.

      But better to be overkill than to find out it's not enough, I say.
      • 20 Amps is more than enough

        P=E*I or 115v * 20= 2,300 watts also my power supplies are a bit of overkill but as you get the power supply closer to full load-the efficency goes down and and the power supply generates more heat-heat is a power supply killer.
        Tell you what if you can measure the temp in the case and then the psu, I bet it is much higher in the psu.
        Michael L Hereid Sr
    • Well seeing I have been into Electronics since

      1960-AA Degree in Electronics since 1966. Designing and building my own electronic projects since then. Amatuer Radio for 30+ years. So I think I may have just a little more experiance than him.
      Michael L Hereid Sr
  • I think the Antec calculator is seriously faulty

    I entered the peripherals into the calculator and did it with my method. Antec is practicaly double of my calculations. Why I don't know. But I trust my method more than Antec.
    Michael L Hereid Sr
    • Here's an oddity

      I got a much higher number with your table than the Antec calculator. I have seen things similar to your table before and tend to believe it.
    • Ok I goofed I just the calculation again

      Belive it or not it calculated only 1 watt higher. So It may be pretty good-I must have put wrong facts in. Sorry
      Michael L Hereid Sr
    • Just another scam to lighten your wallet.

      Antec is in the business of selling you more expensive power supplies. All these calculations are nonsense and Antec is pulling theoretical numbers out of thin air. I'm not going to spend $150 on a power supply when a $60 silent and efficient PSU works just fine with plenty of safety. All these numbers that are being quoted doesn't say a thing about actually doing the scientific experiment and actually testing the peak power consumption with a power meter.

      My Core 2 Duo system even when it's overclocked with a jacked up voltage peaks at 190 watts when I'm gaming. If I had an 8800 GTX instead my peak power might get raised as high as 280 watts and my quality 330 watt PSU has been loaded higher than that though it's not something I would recommend. For the kind of system you're talking about in this blog, I'd probably stick in a quality 400 watt PSU just to have plenty of room to put in more hard drives and DVD burners. Each HDD uses no more than an extra 10 watts. A good quality 400 watt PSU like the silent Seasonic S12 is overkill for the system you quoted and it simply makes no sense to waste an extra $100 on a 600 watt PSU. I know there will be some people that will argue to the death on this point but not a single one of them have ever tried to measure their actual power consumption. It?s this typical elitist attitude that my power supply is bigger than your power supply which makes absolutely no sense.

      As for the fans, I prefer 120 mm fans only because they don't make as much noise and they're roughly double the size in area allowing them to move a lot more air even at lower RPM and lower noise levels. I built this system shown in this guide and it only consumed a peak of 110 watts yet I had no shortage of people criticizing me that my triple-overkill 330 watt PSU was too small and that I needed at least 500 watts. This myth has gotten to the point of absurdity and it?s just another one of those enthusiast urban legends. It?s the same kind of rip-off like the high RPM hard drive myth. There?s absolutely nothing to be gained by going to an even bigger power supply other than lightening your wallet.

      This kind of reminds me of another situation I've run in to. I've know a few guys in facilities that insisted on powering a data center based on the maximum theoretical UL ratings on the back of the power supplies which was simply ludicrous. This meant he could only power 2 servers per 15 AMP circuit which was absolutely nonsense since I could power 10 servers on a 15 AMP circuit while consuming a worst case peak of 10 AMPS with plenty of safety room. The actual typical load was somewhere around 8 amps for 10 1U single socket servers.
      • George that isjust good engineering practice

        You listed typical loads but a good engineer or technician prepares for atypical situations. I would have done the same.
        Michael L Hereid Sr
        • I listed PEAK loads

          I listed PEAK loads under gaming conditions which are WAY less than 330 watts. But I said for expansion room, I'd use the quality conservatively-rated 400 watt PSU.

          The funny thing is that everyone who argues with me on power supplies never actually cite any measurements and then argue my methods are flawed.
          • Well funny thing is

            I checked on the spec's of the power supply your using. It isa Dual rail +12v supply. one rail is +12v1 8 amps other +12v2 14amps.
            Guess what V1 supplies up to 96 watts to. I would seriously consider not overclocking too much.
            You are only considering total power used, but thing is you should also consider what each of the differant voltages can supply continously.
            I wrote in a review on power supplies at epinions dot com, that a high quality 300 watt psu can beat a cheap 400 watt supply any time.
            Michael L Hereid Sr
          • That's what concerned me.

            8 amperes up to 96 watts...? I took it that George didn't have any problems overclocking his Core 2 Duo processor 25% above the stock speed...

            To be in the safe side, I'd consider going with a PC Power and Cooling Silencer 310w power supply, which has up to about 19 amperes for a +12v rail, which provides up to 228 watts of power.

            By the way, don't combine the number of amperes for +12v rails. You have to divide the number of watts for multiple +12v rails by 12 to get the total amount of amperes provided by +12v rails.
            Grayson Peddie
  • PSU

    Weight and Brand Name are my first considerations. I am running a Dual core P4 3.4 with 4GB of DDR667, 4 SATA 500gb HDD RAID and USBs galore and a top of the line power supply of only 450W. No Problems.
  • PC Power & Cooling's PSU Myth (to George Ou)

    Hey George, can you comment in [url=]PC Power and Cooling's PSU Myth[/url]? I tended to side with PC Power and Cooling.

    For my budget/low-end system with AMD Ahtlon64 3800+ 65w and an ATI X1600, I went with PC Power & Cooling's Silencer 310w Power Supply. I didn't go cheap or expensive, though.
    Grayson Peddie
    • A silent 310 is fine for your system

      Your PC shouldn't use more than 190 watts under peak loads and probably idles at less than 100 watts. Since I can't measure your system, I'm quoting high numbers. I have some similar AMD systems that use less power than that.

      The webpage you linked to has some good info, but it doesn't really give you measurements. I do know that many mainstream PC makers will typcially bundle even smaller power supplies in the 250 watt range for PCs with embedded graphics.