Just to play Devil's advocate to my own post-- let's say I did want to burn 450w under load. How could I do it?

The CPU and video are by far the biggest power consumers in a modern PC. Nothing else even comes close. So let's start there.

First, I'd definitely need a SLI (two video card) gaming rig with two high end video cards in it (say, 6800 Ultra or 7800gtx). That's +60w per card (when gaming) for a total of 120w on video.

http://www.behardware.com/articles/574-6/nvidia-geforce-7800-gtx.html

Second, I want a Pentium 4 "Hot n' Prescott". AMD chips use about half the power of a P4. And to make it as bad as possible I want the Pentium 4 dual core chip. For a high end dual core P4, you can easily burn 100w on the processor alone!

http://www.tomshardware.com/cpu/20050509/cual_core_athlon-19.html

That puts us at around 300w for the entire system, with a generous fudge factor for motherboard and other misc. peripherals. Adding dual 10,000RPM raptor hard drives might add another 50w at best, although I'm doubtful it would be that high.

So the best I can do is 350w. Bring your wallet.

Two dual-core chips would *definitely* be enough to push us over the 450w barrier, but I am not aware of any desktop motherboards or even CPUs that would make this feasible. It's an entirely different chip family (Opteron, Xeon) and I don't think those mobos support dual video card SLI, either.

We're more likely to see quad-core CPUs on the same die before that happens. A quad-core CPU would be hot, but far more efficient than two physical dual-core CPUs.

"..stop estimating and start measuring."

Or researching (google, this post) :)

I had an Athlon XP 2500+, 5 harddrives, two optical drives, and a 300watt power supply.

Windows would randomly blue screen every day.

I put a 350watt power supply in it, and the blue screens stopped.

Why would that have been, I doubt my machine would have been drawing more than 150watt total

> I put a 350watt power supply in it, and the blue screens stopped.

This is likely an issue of power supply *QUALITY* rather than absolute watt rating.

But yes, PSU quality is tremendously important!

You could generalize and say that "generic 500w PSU will probably be higher quality than generic 300w PSU". But that's not really getting to the actual problem..

Power supplies have multiple outputs with different voltage ratings and different maximum loads. The nominal maximum load is *not* spread evenly or dynamically across all outputs -- each voltage group has its own transformation circuit that determines its output characteristics.

Your system will crash if even a single output comes close enough to its limit to cause fluctuations in the voltage level. Short of having detailed data on a PSU's individual outputs (and who has that?), buying a bigger one is the simplest way to ensure that *all* outputs get sufficient power.

>>GeForce 5200 PCI video (for 3rd display)

Oh man, I've barely forgotten that you use Athlon X2's, now you have to rub my nose with the third display?! ;))) [I have only two 18" LCDs]
OK, to complete the picture (pun intended ;)) tell us - what are the screens: kind (LCD/CRT) and size?

Id like to get one (several) of those meters to see where my power $ going. I had a couple of machines (one win adv serv 2003 and one linux (also hosting vmware with win adv server 2003)) which eventually evolved into hyper threading P4's. Underload (i.e. either of them doing 'heavy' sql processing) or rendering 3d they' both lockup, shutdown or reset as they saw fit. I had to upgrade the powersupplies in both to over 400 Watts to solve respective problems. I tried various powersupplies, I'm not convinced the quality had anyting to do with it. Perhaps there are peaks/spikes of power required that the meters sample rate is to slow to 'see'?

> buying a bigger one is the simplest way to ensure that *all* outputs get sufficient power.

Not necessarily; generic ultra-cheap "500w" power supply may not be any more reliable than a 300w from a quality brand (say, Zalman or Fortron).

> what are the screens: kind (LCD/CRT) and size?

It's 3 Samsung 19" 1280x1024 panels. Nothing that special nowadays.

> I tried various powersupplies, I'm not convinced the quality had anyting to do with it

How many did you try? Quality is definitely a big factor. There are a lot of crappy no-name PSUs out there:

http://www20.tomshardware.com/howto/20050228/power_supply-39.html

"However, not every model delivered the performance that was promised. Some devices also showed voltages falling outside the specified ranges, which can result in a system crash. Usually these crashes cannot be reproduced, and everything seems to run normally again after restart. Only in the rarest of cases is the power supply diagnosed as the source of system problem. In our lab studies, however, we have often had the experience of seeing inexplicable crashing that resolved itself after exchanging the power supply."

I wonder if the kill-a-watt device properly accounts for reactive loads, like power supplies. For non-reactive loads like incandescent lamps, it's probably pretty accurate. But reactive loads need more analysis.

I went to the P3 international website, but their user-manual is not online. Their site doesn't mention the difference between resistive loads and reactive loads.

This is also known as power factor correction.

What does kill-a-watt say about that?

Typical computer power supplies list their capacities in VA (volt-amps), not watts. And because these are reactive loads, I do not think there is a direct conversion from VA to watts due to the phase difference...

Power supplies are rated in exaggerated terms (like cheap stereos). The wattage is usually advertised as peak, not continuous.

Interestingly, 500W Peak to peak = 357W RMS (true V*A power - divide by square root of 2 or about 1.41).
So get a no-name advertised as 500W may well be the same as a 350W quality branded PSU. lol

You wouldn't happen to live in the midwest area and want to let me bum that Kill-A-Watt for a day or two, would you?

You can find out where I live easily enough by visiting the "About Me" link on the main page.

I bought a killAWatt shortly after you mentioned it, and I've seen it up to 356watts while running BattleField2 in network mode.

Here're the specs of the machine:
p4 3.4g
gForce 7800gtx
2 gigs mem
raptop 10k rpm hdd, caviar 400gig, misc 100 gig
20" dell lcd
pioneer elite receiver- for sound
2 cube speakers from Cambridge

I'm suspecting that the graphics is adding a good amount of the load. With CPU maxed without the 3d graphics, it'll normally be around 320. With the CPU at idle and the LCD on, it's normally around 260-280 watts.

LCD off, largely idle, it's 220 or so.

Everything off, it's about 4-6 watts.

> graphics is adding a good amount of the load

The CPU, particularly an Intel CPU, will always be the biggest single power draw in your system.

The graphics card is a close second though. And in extreme cases (low end, recent generation CPU combined with very latest video card) you could have the video card pull more power than the CPU -- at least while gaming.

I like the idea of getting as little watts in a power supply as you need. If you want the best idea of what you'll need, with the most precise guestimates as to how much power your system will use, I don't think you can do much better than here:
http://www.extreme.outervision.com/index.jsp

Every component, broken down according to watts used, and it will even let you know how much more you are using at various degrees of overclocking and at peak cpu/graphics usuage.

Save yourself the 30 dollars and try the extreme PSU calculator instead. (not affilated with this non-commercial site in any way, I just have found it very useful in putting together higher end systems without the unneeded PS megawattage)

The calculator is OK, but it's telling me my recommended PSU wattage is 248w for the system described in the post (see top of this page). That's way, WAY over the actual maximum measured consumption of 133w with two instances of Prime95 running.

The calculator is fine for estimating if you bear in mind it's giving you about 1.5 times the size of the PSU you really need. Thirty bucks is worth it to actually be able to MEASURE what is happening-- instead of passively taking someone else's word for it.

There's a fun thread here at SilentPCReview documenting the maximum amount of "stuff" people have run in a PC using a standard 300 watt power supply. You'd be surprised:

How much will a 300w power supply run?
http://forums.silentpcreview.com/viewtopic.php?t=3075

I have a question. Would my pc (according to the calculator needs a 525 watt psu) use more power with a powersupply closer to what it uses with a small headroom, or would a really big powersupply use more. You have to take in to account that ex. a 200 watt psu running at 150 watts would make more heat than a 400 watt psu running at 150 watts. you would loose power there. My situation is that i am rigging up multiple(6) pc's and i need to use the minimal ammount of power because i am hooking them through 2 power cables. so i need to use the least ammount of power. note these pc's range from a 486, p1-4 and my dual xeonl. I am planning to get the 200 watts for all but the xeon which i will get a larger one to take the overclocking.

> You have to take in to account that ex. a 200 watt psu running at 150 watts would make more heat than a 400 watt psu running at 150 watts. you would loose power there

What you're asking about is efficiency. Most power supplies are less efficient at LOWER loads, so you would want a PSU that runs closer to its actual limit.

I highly recommend reading through the PSU reviews at http://www.silentpcreview.com ; they determine the efficiency graph for each PSU reviewed.

You've neglected to rate the power comsumption of your displays. I plug my computer, display and high end KVM into my APC UPS, which rates the power consuption and usually sit about 220watts.

I wouldn't mind tracking down some of those killawatts for my other devices and such tho!

Incidently, your device (and my apc ups) measure and report AVERAGE flow, but I doubt either one can properly account for sudden, short spikes that might be required.

Then you get in to power per rail on the supply, ability to step into a sudden demand increase etc etc etc. Always buy QUALITY, but that also usually means higher wattage capactity also. Nothing wrong with that, as it gives you room to grow and spike into.

> You've neglected to rate the power comsumption of your displays.

LCDs are about 20-40 watts, as I recall.

> I doubt either one can properly account for sudden, short spikes that might be required

You can see the power spike when you power on the PC (hard drives are notorious for requiring a startup power spike). The kill-a-watt does sample a few times a second.

This got me interested...

so I just picked up a kill-a-watt, well ok something similar, a no-brand chinese gadget. In HL2, doom3, whatever the highest peak power consumption I've managed to have the thing display is 152 watts...

FYI specs
Sempron 2300 @ 2375mhz 1.85vcore (sisoft sandra 2007 estimates this as 90 watts)
1gig DDR500 @ 2.8v
seagate 250gb 7200
Legend 6600GT 128 AGP 600/1100
various lights and about six fans (25w total)

This is shocking, I've been looking to buy a 600 watt power supply for my SLI rig i'm cooking up, but this is implying for my current ageing rig I don't even need my generic 400 watt power supply? I've got a old 300w somewhere that should do the job on my current system..??

Yep.

That's the advantage of measuring yourself instead of following all the hearsay on the Internet. ;)

How much more does a Sapphire Radeon x850 XT take in watts compaired to a PowerColor Radeon 9800 PRO?

And can a 300/350 watt(not sure if its 300 or 350) PSU handle this?:

17" CRT Screen
AMD 64 3800+ 2,4Ghz Venice
Sapphire Radeon x850 XT
Asus-Socket 939 - ATX Nforce4 (A8N-E) - Gblan/Raid
Kingston HyperX 2x512MB DDR dual channel
2xMaxtor HDD
3xSeagate HDD
Zalman - CNPS-7700-CU (CPU fan)
2 x 80mm chassi fans

ok i came on here wonderin if my ''old'' 300w supply would b ok for the p4 i wanted to buy, and seems thats the answer is a defo yes :) and id need to b a power mad crazy modder to get close to 300w

one question tho if i do some how top the 300w mark is it true that the only affect it will hav (according to sean's post) will b blue screen?

OK, so if you lose 50% of the wall power, then when you buy a UPS and plug your PC into the UPS you are losing 50% of the UPS right there.

IOW if you designed the PC power supply such that the UPS comes AFTER the switching (so the battery is as close as possible to the motherboard) you would automatically, immediately get a 50% boost on UPS capacity.

Not to mention, the wall-to-DC circuitry can be removed from the UPS, as the computer's power supply is now doing that.

That seems a better design to me ....

Anybody know where these beasts might be available - or what they're called so I can search for them?

I googled for "internal UPS" and got a few hits.

Why is this such a specialty item, since the design makes more sense to me, but the regular UPS design is available at every computer shop?

You're measuring steady state operation.

What about turn-on requirements - I have 7 hard drives in one machine - all that inductance would draw TONS of power, right?

It's low power requirement to keep the platters turning once they're up, but could a 300W PSU provide that initial whack?

This is all very good discussion and Jeff, that device looks intriging... good find.

However, one issue was overlooked... heat.

The closer the steady watt usage is to the peak rating of the PSU, the more hot the PSU will get. If a PSU is used close to its peak rating steadily, you not only endanger your PSU, but you also add a little to the heat to the already hot system.

I understand not all areas suffer the 115F summer weather as I do but a suggestion for anyone like me who is very concious of heat (or even sound) in thier system, I would measure your average watt usage when idling and try to get a PSU that is rated for close to, if not, twice that amount. (and I reccomend a 120mm fan to account for sound as well)

I hope my input helps!

I've just got myself a eVga 7800GT and I feel my current Chinese/elswhere manufactured 400w power supply may succumb to a quick death if I shove it in...after searching a bit (and landing on this site) I was wondering if I could just purchase another Chinese/elswhere 400w and dedicate it to the VGA card? from what I've read this particular VGA card WILL suck 50w or so of power from its Pci-express slot would this still lead to an untimely demise for my current installed PSU?

My sys specs are:
- Pentium 2.8 HT(cooled w/ Gigabyte Rocket cooler)
- 768MB ram
- Audigy 2 soundcard
- VGA In-build Intel 915 Xtreme Graphics
- Western Digital 80GB sata
- 1 x Gigabye DVD writer, 1 x Gigabyte Combo drive

hi, i wanted to know how much power consumption my laptop does for a hour. my laptop spec are:
Intel core 2 duo.
1Gb ram.
Geforce go 7400 (256MB)
15.4" wide screen display.
lightscribe DVD writer (Double Layer burning).

could you please tell me how can i measure the power rating.

I have owned a "Kill-a-watt" device for some time. I purchased it to monitor the (60) cycles of my 2 RV generators. One MUST run @ 3600RPM to produce 60 cycles.....but I digress....
.
My "Kill-a-watt" NEVER reads 60 cycles, even on the grid. It ALWAYS shows 59.9 cycles. Anyone else notice such an error in cycles? Makes me wonder the accuracy of the other functions.

For a detailed breakdown of AC power demand under idle and max load conditions for many different types of systems, have a look at the updated REAL SYSTEM POWER REQUIREMENTS table on page 4 of SPCR's Power Supply Fundamentals -- http://www.silentpcreview.com/article28-page4.html

Most of the systems are using high efficiency ATX PSUs; the highest one being 84~85% at the top of its efficiency curve. The max power draw of even an OC'd Pentium D950 + ATI X1950XTX system is just 256W.

I need a lil help im running 8 19" monitors. 4 off a ATI firemv 2400 pci-e, 2 off a nvidia gefore fx5200, and another 2 off a ATI radeon X600, in addition I have a 250GB WDC 7200rpm sata system hard drive, a maxtor 200gb external usb hard disk, a phillips 5.1 soundblaster pci audio, 4GB ddr SDRAM, and I also have a AMD Athlon x62 4800+. I have a 350W PSU and my CPU keeps crashing what you guys suggest I get

The reason people insist on big power supplies is because you need them for reliable operation. Components usually have surge powers that are 2-3 times larger than steady-state powers. Your hard disk uses 10W under normal operations, but 25-30W during spin-up. Same thing goes for your DVD/CD drive. How about your motherboard and other non-mechanical electronics? When you hit the power button, there's a ton of capacitors with large in-rush currents. If your computer implements power saving (AMD Cool-n-Quiet or Intel Speedstep), it is periodically adjusting voltages on those capacitors, again, with big currents.

Beyond that, you want some safety margin. The most common system failures are either mechanical (fans and hard disks), or hot capacitors failing, typically in the power supply. If you've got a power supply with cooling such that it is rated at 250W pushing out 250W of power, it will be very close to the failure point of those caps. If it's a 500W one, you'll be halfway to room temperature. Your time to failure will go from quite low to very high.

Measuring beats guessing and calculating, but you've got to have basic clue about failure mechanisms, failure rates, and EE to do the measurements properly.

Thanks for the article Jeff, got to see if I can get one of these Kill-a-Watt things over here in Blighty =)

I will have hours of fun measuring devices all over the house, mainly so I can tell my dad he is wasting more energy running around turning things off standby than he is saving!!

BTW, some people seem to think their display is going to affect power drawn from the PSU - this is only true for laptops though.

All desktop setups use separate power supplies for the monitor (and speakers). A few (older) PSUs have a output kettle socket that you can connect your monitor power lead to, but that is just a pass-through connector and doesn't take power from the PSU.

Thanks for the article Jeff, I will have to see if I can get one of these Kill-a-Watt things over here in Blighty. =)

I will have hours of fun measuring devices all over the house, mainly so I can tell my dad he is wasting more energy running around turning things off standby than he is saving!!

BTW, some people seem to think their display is going to affect power drawn from the PSU - this is only true for laptops though.

All desktop setups use separate power supplies for the monitor (and speakers). A few (older) PSUs have a output kettle socket that you can connect your monitor power lead to, but that is just a pass-through connector and doesn't take power from the PSU.

Surely you need a larger psu as it is the way it distributes the power to all the different components in your computer.

For example a 600w psu has a maximum output of 600w yes, but this is split between +-3v +-5v and 12v and other things. So as in your example if all your components need say 12v and need about 300w of it, you may need to have a psu that can supply "in total" 750 or 800. Each PSU is different and distributes it's power differently... just something to mull over or prove wrong.

Tom

Couple of things that don't seem to have been covered. One is that the power rating of a PSU doesn't necessarily indicate its efficiency. This is a figure that is notably lacking from most specifications. In theory, a 500W (max) reasonably efficient PSU running a certain system (say requiring 180W between all the rails) could consume *less* mains power than a 300W inefficient PSU powering the same system.

The second is the thing about total energy auditing. It's all very well to get a new machine which consumes 100 or 120 Watts less power, but how many people stop and think how many Joules of energy it took to produce that machine? What if, for example, in order to recoup the energy burnt creating the machine you would have to use it instead of your old box for 10000 hours?

No one ever stops to add it all up. The most environmentally friendly industry wide option would be to abandon Windows Vasti^HVista so that people can continue using the machines they already have, instead of junking them by the million for a bit of visual glitter and no extra usability.

Similarly for cars. IIRC the energy required to mine the necessary raw materials and build a new car amounts to about half as much as it will ever consume in its lifetime. So, replacing your old car with a new one has done half the damage it will ever do before you even take it out of the showroom. It is actually greener' to repair your old car and continue using it for an extra five years or so than it is to scrap it and get the latest "I'm green" status symbol.

People don't think about the energy audits, so many want to be *seen* to be environmentally conscious so much more than they actually want to *be* so.

Dave J.

400W benchmark is come and gone. On old PC's would've been fine with 400-500W PSU but latest ones will not work with less then 500W. My 2900Pro graphics card alone chews up 150W on standby and 350W on full load. Add a CPU (65-150W), all the ram, optical and hard drives etc, and you have whooping 600W. Throw in another graphics card and you got yourself looking for the latest(at least 85% efficient) 1000W PSU.

Someone mentioned that their computer blue screened, and they changed the power supply and that fixed it. A new power supply will have slightly different voltages outputs for 3v, 5v, 12v, and 28v; or whatever voltages it's designed to produce. As soon as you change the DCPS, you have new voltages being applied to your various devices/loads, which changes the timing in all your circuits. In failure analysis I often felt I could solve a timing problem just by adjusting the voltages. Voltage inputs can straighten up a lot of the leading and trailing edges on your signals, and can eliminate timing problems that may contribute to blue screens.

The eXtreme Power Supply calculator mentioned above is pretty accurate now. Plus it has been updated on a regular basis with latest components. Now they have two versions: Lite (free) and Pro (small fee). Pro version has Amperage on main PSU rails, which is very useful since only Wattage is not enough to decide which PSU you need.

Highly recommended.
http://www.extreme.outervision.com/psucalculatorlite.jsp

How many watts does the Kill A Watt use?

>> How many watts does the Kill A Watt use?

My no-name KillAWatt copy apparently uses .67 watts.

Can anyone explain this for me -
I plugged my two monitors into the meter and it read approx 60 watts.
Then I plugged my computer in and it read approx 69 watts.

However when I plugged the double adapter that hooks up both the monitors and the computer, it read 80 watts. Shouldn't that be closer to 130?

Is there a good explanation for this, or are my suspicions correct that the accuracy of the meter is not very good.

I did get more predictable results with the computer and monitors turned off - monitors sleeping at 14 watts, computer at 23 watts; combined through the double adapter is 38 watts.

Worth noting that leaving the computer turned off, but left on at the wall, for a year, adds up to 332kWh = 480kg greenhouse gases (using figures for my town)... or about the same as a 1800km car trip in a medium sized car.

Thanks for this great article. I've been wondering about my laptops power usage, and this really helps with getting some real answers.

I'm going to have to buy me one of those devices. Would be real helpful when I go looking at HDTV's too.