Debunking Power Supply Myths

by Christoph Katzer on 9/22/2008 3:00 AM EST
POST A COMMENT

97 Comments

Back to Article

  • Dancer - Thursday, October 22, 2009 - link

    On the face of it, this article seems an excellent, well-researched contribution to a highly vexed question. I do have a concern, however: We know that the power output of a PSU drops as it ages. We also know that this drop depends partly on quality and partly due to random chance. If I'm buying a PSU to last 3, 5 or more years, will this seriously affect the capacity of the PSU I should buy for a given machine?
    Reply
  • internetrush - Saturday, June 06, 2009 - link

    Ok, lets low ball it, im running 200w (average) per graphics card, about 50w cpu (core i7) and three hard drives.

    Lets see

    200 x 2
    +50 + 50 (motherboard chipset)
    +30 Sound card
    +10 cd drive
    +20 (fans)

    During a game, much less a stress test, im lowballing a 600w load on my PSU.

    If i had an 800w PSU that would be 80% of its total output, which thereby increases its heat and decreases its life.

    When you buy a 1000w PSU, not only are you ensuring that you will never watch your computer go up in smoke (had a friend do that to a 350w on an old P4) but you are also not having to replace it whenever you buy a new processor or add something to your system.

    This article is good, however, on a tech website i would expect a bit more consideration for the higher end gamers and common sense.

    Common sense says, if you are a higher end system user, you WILL expand said system!

    For gods sake! Some cards today use up to 500w power (the 4890X2 and new 295 SuperCard).

    As a gamer, id rather have a 2000W PSU that id never have to replace than a 400w that would FRY as soon as i threw on a new video card.
    Reply
  • Christoph Katzer - Saturday, August 08, 2009 - link

    Sorry for the late reply.

    You are probably right when you see it from the perspective of a high-end-hardcore-gamer... But do you know how small the percentage of people is who actually own a real high-end system?
    Reply
  • JohnMD1022 - Sunday, March 15, 2009 - link

    It would be nice if you could periodically update this with newer components. Reply
  • lopri - Tuesday, April 28, 2009 - link

    quote:

    quote
    Reply
  • BillyBuerger - Sunday, October 26, 2008 - link

    Anyone have any info on that Thermaltake QFan 300? That thing looks great efficiency wise. Not normally a Thermaltake fan. And the fan controller looks like it sucks. Just keep it below 150W... Or fan swap. Reply
  • Christoph Katzer - Thursday, October 30, 2008 - link

    I will have a review up soon! Reply
  • Cincybeck - Tuesday, September 30, 2008 - link

    Couple of "knowitall" friends were trying to tell me I was going to need a larger power supply when I built my new system. Which in turned incited the Microcenter sales person saying oh yea you're probably going to need that too. I turned around said I estimated these parts to draw at most around 200, 250 watts, and I have a 500W Seasonic M12. Shut him up pretty quickly, but my friends were still pushing it the whole way home. So now if they ever bring it up again I can print this article and shove it in their face. Thanks =D Reply
  • 0roo0roo - Sunday, September 28, 2008 - link

    i like the graphs:)
    keep it up!
    this is the info we need!
    normally the psu market is just lousy because of the lack of any real information.
    Reply
  • mark84 - Friday, September 26, 2008 - link

    For those quoting that old link for the AtomicMPC graphics card power thread, the new/current one is being maintained here http://forums.atomicmpc.com.au/index.php?showtopic...">http://forums.atomicmpc.com.au/index.php?showtopic... Reply
  • nilepez - Thursday, September 25, 2008 - link

    I've seen this site and countless others constantly promoting the benefit of 750 watt+ PSUs.

    The reality is virtually nobody needs 500W, much less 750 or 1KW.

    Reply
  • AnAverageJoe - Wednesday, September 24, 2008 - link

    Thank you for this article, it is really one of the more organized efforts at getting this kind of info out there that I have read and hopefully will save folks from overdoing it (saving $$ in the process). Couple of comments:

    1. p.1 “If people really took the time to examine system power requirements”: From the perspective of building a new rig, where does one find this information? Any sort of figures, short of the very few articles such as this and public power supply guesstimators, appear non-existent. As stated in the article something is better than nothing but getting precise information appears impossible for any given component. Both Intel and AMD publicly provide technical data on their parts but the only single figure one can really get out of that data is the TDP for the processors. I have failed to find any similar technical data for GPUs and although most graphics card reviews now include power draw it is for the system and not the graphics card itself thus there is is no way to get the discrete graphics card power draw. Ditto for every other component. In sum, I don't see any way for one to gather the required data to compute the power requirements for any planned build short of actually building the thing and putting test equipment to it.

    2. I was left wanting a statement to the effect of "one important goal of choosing a power supply is to maximize efficiency under expected operating loads." The examples do make the point just took a bit longer for me to get it thru my dense skull.

    Again, thanks for spending the time to put this together because this really is I think one of the most overlooked areas of system integration and in my experience impossible to more than generally guesstimate.
    Reply
  • Cincybeck - Tuesday, September 30, 2008 - link

    Try searching "power supply calculator" in Google. The one, eXtreme has, provides a pretty complete list and quite a bit of options. I punched in the low end and mid range computers from the article. The calculator is high according to the article's results, but close enough to give you a good idea. Reply
  • Johnniewalker - Wednesday, September 24, 2008 - link

    I have been waiting for an article like this for years, thanks! I always suspected people were buying PSU's that were too big. Reply
  • rvikul - Tuesday, September 23, 2008 - link

    Anandtech is back! This is exactly the kind of information Anandtech excels at providing. very relevant and very useful. Thanks and keep it up (and please dont go back to discussing social issues etc).


    Reply
  • OddJensen - Wednesday, September 24, 2008 - link

    Strange figure for the 2900XT. Knowing it's one of the most powerguzzling single GPU cards, it should have been much higher, right? Reply
  • CSMR - Tuesday, September 23, 2008 - link

    I'm going to repeat myself because I feel this is important.

    The tables claim to represent actual power consumption of processors and chipsets but the figures are very exaggerated. (There are even people who run whole systems on one or two of the chipsets listed on less power than the power the article claims for just the chipset.)

    Here are more accurate CPU measurements from xbitlabs:
    http://www.xbitlabs.com/articles/cpu/display/intel...">http://www.xbitlabs.com/articles/cpu/display/intel...
    Behardware gets similar results. (Can't find the link straight away.)

    Must fix this as Anand needs to keep its reputation for good information. You can't have figures (CPU idle power) that are out by a factor of 10!

    The graphics idle power data are better than the CPU data but also too high, by a factor of about 2, compared to existing measurements:
    http://archive.atomicmpc.com.au/forums.asp?s=2&...">http://archive.atomicmpc.com.au/forums.asp?s=2&...
    Reply
  • JarredWalton - Tuesday, September 23, 2008 - link

    The question is: did Christoph measure power incorrectly, or did someone else? From my understanding, he's measuring the current on the various wires leading from the PSU to the components. HDD is of course easy to measure. The ATX12V/EPS12V connector supplies the CPU, so that's simple as well. PCI-E gets 12V from the PEG connector along with the extra four pins on the 24-pin ATX connector. The remaining pins on the 24-pin ATX feed the chipset, RAM, and other motherboard components. Sum all of that together and you get the power draw for the entire system.

    Perhaps the CPU power draw numbers are high and the chipset/mobo numbers are low, but worst case the point of the article is to show that higher wattage PSUs are not required for most systems. A midrange system with similar components might use a bit less than what we estimate, but I'm quite sure it wouldn't need more power than our high estimate.

    I won't guarantee that the individual numbers are 100% accurate, but I doubt that desktop C2D processors are idle at only 7W or less. I know on my own C2Q Q6600 8800GT 4GB system it idles at 176W power draw, and if I put a load on just the CPU (Folding@Home SMP) the draw increases to 262W. Guessing at 80% efficiency, the components are consuming 141W to 210W, which means CPU (and mobo, chipset, and RAM) power use went up 69W. That's pretty close to the QX6850 result (fudging on the 65nm vs. 45nm and Penryn vs. Kentsfield). The Fur benchmark also gives ~262W average (one core is 100% load on the CPU), so the GPU + one CPU core increases power consumption by the same 69W, but some of it goes to the CPU and most of it goes to the GPU. That jives with the 51W increase in power Christoph measured on the 8800 GT. Running both - a "worst case" test - gives a power draw of 315W to 337W, with an average of around 324W.

    So a system somewhat similar to Christoph's "midrange" setup (and estimating efficiency) uses 141W idle and 260W load. That doesn't include trying to tax the HDD or DVD, which might increase the load by another 20W, and it uses a single GPU instead of a 3870X2. What's the specific amount used by the CPU, the GPU, the RAM, the chipset, and the other motherboard components? I can only guesstimate, but stepping back to examine the whole picture I don't have any serious problems with the tables on page 1. CPU power is probably lower, since it sounds like Christoph measured the current going through the ATX12V lines and some of that will feed the VRMs and other bits and pieces on the motherboard.
    Reply
  • Christoph Katzer - Wednesday, September 24, 2008 - link

    Thanks for the explanation. The point being is that I measured how much current is going to the CPU through the 12V rails. What happens later doesn't matter since the power supply needs to deliver X current at that specific time.

    That doesn't mean however that the CPU is actually using all of this delivered power since we loose power at the VRMs and CPU itself. SO if there is X watts going to the CPU it doesn't mean the CPU actually needs that much power but since it is being delivered it should be called the "actual power consumption".

    For example the 6000+ which needs quite too much obviously. We know now it had something to do with the different VRMs at the AM2 and AM2+ boards. The power is being delivered, meaning the power supply needs to provide this amount x. If the CPU actually needs it or not is irrelevant. Of course if I would have known from the difference between AM2 and AM2+ I would have measured it in a different way, no one is perfect and I know now how to do it better next time.

    As for the Chipset it is indeed tricky since every mobo vendor has different additional chips installed that take a different amount of power. So if we do publish an article about this we will have to mention the actual manufacturer of course.

    It's just funny that everyone says the results are too high but I get an Email from Taipei that my numbers are far too low...
    Reply
  • Barack Obama - Tuesday, September 23, 2008 - link

    Notch one up for the great AT team :D Reply
  • computerfarmer - Tuesday, September 23, 2008 - link

    Good article! Based on the number of comments, this has many of us thinking. This gives a way of figuring out our needs vs bigger is better.
    This article gives us enough information to make educated choices. Looking at the 12V rails is a good place to start along with total power. With a single 12V rail like Corsair VX series, the amount each rail carries is a non issue.

    A rule I have always followed is to use a power supply with at least 20% more power than the maximum required.
    Reply
  • dragosmp - Tuesday, September 23, 2008 - link

    Hey, I just wanted to say that after posting this link on OCN, there were tenths of replies. People want to know how much various things consume, as all this 1kW PSU hype hype is getting very tiresome.

    I for one would be curious how did you measure the current thru the PCIe slot - have you modified the slot to access the power lines? Soldering geeks would really like to know :)

    And lastly, browser compatibility. This comment is very hard to write in Chrome (writing overlapping), maybe you or google will fix this.

    Cheers,

    Dragosmp
    Reply
  • Fudus - Tuesday, September 23, 2008 - link

    I am stupid and run my 4850 off a 300W power supply, with a sata power>molex>pci-E power connection. Go Go overloading!

    It seems to work at stock speeds under stress as well for some reason. I really didn't expect it to work this well (C2D e6550, Radeon 4850,300W power supply, 2GB RAm, G33 motherboard, 1 HD/DVD+RW)
    Reply
  • marc1000 - Tuesday, September 23, 2008 - link

    you are not stupid, Fudus. if you sum the power for your c2d (never more than 65w), mobo (40w), ram+HD+DVD (5+15+10) you have 135w. then add the 4850 and you are still fine with a 300w PSU. Like i Said before, I have a 90W CPU (that old Pentium-D, argh!) but run a radeon 3850, so my system consumes about the same energy as yours. and the 300w is working fine with me too. or else I would be a stupid too =P Reply
  • dragosmp - Tuesday, September 23, 2008 - link

    Wow, this gives me hopes to mount a 4670 on a 120W pico-atx powered rig.
    How is the noise, the +12V level?
    Reply
  • oopyseohs - Monday, September 22, 2008 - link

    I have seen upwards of 700W while testing on old QuadFX systems with 8800Ultra SLI. I would imagine that Skulltrail overclocked to 4.0GHz (easy) and 2x Radeon HD4870X2's in CrossFireX would demand significantly more than 700W. Of course these systems are rare, but they do technically validate the "need" for power supplies that output 1000W+.

    In any case, thanks for this article. I think it does a great job of showing why ultra high-output power supplies are not even close to necessary for 99.9% of the computing world.
    Reply
  • JarredWalton - Monday, September 22, 2008 - link

    I have a Q6600 @ 3.42GHz with dual 3870 cards, and that peaks at around 650W at the outlet. Certainly there are plenty of overclocked PCs that can draw more than 700W... but once you take efficiency into account, my PC is only really using around 520W. I've also tested high-end water-cooled setups with 8800 GTX SLI that topped out at 650-700W power draw as well. I personally think with ultra high-end PSUs that having six 12V rails isn't very useful as well - some of the problems people experience with lesser PSUs simply comes from 12V rail distribution.

    If you're running dual GTX 280 cards and a quad-core (probably overclocked) CPU, I don't think there's anything wrong with 1000W PSUs. In fact, I know Gary has blown a few 1000W PSUs with his overclocking testing in the past. However, I'm running perfectly happy now with a 3.2GHz quad-core and really don't need even that much CPU performance; it's all about the GPU for games, and CPUs are only really taxed in 3D rendering and video encoding it seems.
    Reply
  • Griswold - Monday, September 22, 2008 - link

    Now that you made that clear, give us some of your great reviews for reasonably dimensioned PSUs instead of these 600-1000W bricks. :P Reply
  • whatthehey - Monday, September 22, 2008 - link

    Maybe it's just me, but I sort of got the impression that the reason for this article is precisely that Anandtech is tired of only getting the highest-end power supplies for testing. It's all marketing BS of course: they don't want to limit sales of the $300 1000W PSUs so they only send those out for testing.

    Or maybe the truth is more nefarious: they figure if they send out a top-quality 1000W PSU that has great efficiency and voltage regulation, unsuspecting buyers will buy their lower wattage parts that might not be all that great? The first is more likely, but it wouldn't surprise me if some of the less expensive 400-500W PSUs (even from major brands) use much cheaper components. Which is why we all want to see them tested, and probably also why the companies don't want to send them out for review.
    Reply
  • Martimus - Monday, September 22, 2008 - link

    Since this article is about choosing a PSU for a new computer build, it would have been nice to include new components like the P45 chipset, or the new nVidia or AMD graphics cards. I started reading it with the hope of knowing how well the Antec Earthwatts 500W PSU that came with the Sonata III case I bought would be do with various build options. The problem was that with only old equipment being used, I couldn't come to any real conclusion. It seemed to read more like a report to prove a point rather than an actual guide written to help with a new computer build. I am not trying to berate the article, but I just don't see the point of it. Reply
  • JarredWalton - Monday, September 22, 2008 - link

    If you look at all the included components, it should be pretty clear that the latest parts won't be drastically different from the tested components. Sure, 4870 might use 30W more (or 30W less) than the 3870, and the P45 might use +/-10W relative to the P35. Does that really change anything with the information the article conveys? I don't think so. Midrange PCs are still going to use 150-250W for the most part, whether with last year's components or with the latest stuff. If you want to look at the top-end, then GTX 280 will use more power than the listed GPUs, but even then you're not going to break 600W without overclocking. Reply
  • Martimus - Tuesday, September 23, 2008 - link

    Thanks. I have since calculated out what is needed for the processor and chipset (for a E8400 and P45 MB) and it came to about 8A. I went on the AMD website to find what the current draw was from the 4870. It doesn't say, but does recommend a 500W PSU. I am a little concerned though, because the EA500 isn't on their approved PSU list from Antec, and I have had issues with their power supplies before, plus they only have 17A available on each rail. that should be fine, but I would like to make sure I am not loading the 12V rail too heavily as well. If those components don't work (those are my plan for the moment), then I can always adjust, but I would like to have that peace of mind before I make the purchases. (the computer is a present, and I bought the Sonata III to save money because I heard good things about the Earthwatts brand, but now I am starting to get worried about this particular model.) Reply
  • Matt Campbell - Monday, September 22, 2008 - link

    I would also point out that in addition to all of the data Christoph has pulled together, we have some power consumption numbers on overclocked systems with GTX 280s as well, which seems to be one area people are asking for.

    4.0 GHz QX9770 with SLI GTX 280s: 579W Max
    4.0 GHz QX9650 with Triple 8800 Ultras: 671W Max
    Reply
  • JarredWalton - Monday, September 22, 2008 - link

    Those figures are power at the outlet, though, right? And they're not tested in the same fashion as what Christoph did. Still, if you max out at 671W at the outlet, even with 88% efficiency you're only using 590W - nowhere near 1200W, which is what NVIDIA certifies for 3-way SLI. Quality over quantity, naturally, but there really aren't many terrible 1000W PSUs out there (which is why they all cost over $200). Reply
  • Torched - Monday, September 22, 2008 - link

    For all the hullabaloo about the 12v Rail why nor just recommend a PS that has a single rail. More and more manufacturers are going in that direction anyways since it eliminates the whole power trapping issue. Reply
  • gramboh - Monday, September 22, 2008 - link

    Good article. The underlying point is that you do not need 1000-1500W PSUs to run your system like I see many people claiming you do for SLI setups. I see lots of people with 700-800W PSUs with one graphics card. Insane.

    I've been running a Corsair HX520 (520W) for a year and a half now. My current config is:

    Asus P5B-Deluxe (P965) mobo
    Q6700 G0 @ 3.3GHz (1.42v actual voltage reported by Speedfan)
    4x1GB PC2-6400 DDR2 memory
    EVGA GTX280 video card overclocked (675-1350-2422)
    3 hard drives (7200.11 1TB + 2x 7200.10 500GB)
    1 optical (18x Samsung DVDRW SATA)
    3 case fans ~1500rpm + CPU fan 1200rpm
    SB X-Fi XtremeGamer audio

    Zero issues and I'm sure I have headroom to spare. Some people on forums told me I should upgrade my PSU to at least a 650W Corsair or 750-800 of other brands to run the 6700/GTX280, just goes to show people have fallen for the marketing hype of PSUs.
    Reply
  • cesthree - Monday, September 22, 2008 - link

    I have had a few different systems in the last 6 months. Using the Zalman MFC-2 I can see the wattage being used in real-time. How accurate it is, I do not know, but it gets me in the ballpark.

    All systems have had the same 4 WD1600YS HDD, Lite-On DVD burner, 700W OCZ GameXStream PSU, and OCZ DDR3 1333Mhz 2x1 GB RAM. I also run 4 x 120mm fans including the one mounted to my TRUE-120.

    Set-ups:

    1. EVGA 790i Ultra (JUNK) w/ Q6600 @ 3.2Ghz 1.35VCORE, 1400Mhz FSB, 8800GTS 320MB SLI
    Idle Load: 225W, Prime Load: 275W, 3DMARK06: 325W.
    Never saw higher than 400-425W during highest loads.

    2. DFI X48-T3RS (PWNING 790i ULTRA) w/ Q6600 @ 3.0Ghz 1.32VCORE, 1333Mhz FSB, EVGA 9800GTX
    Idle Load: 200W, Prime Load: 250W, 3DMARK06: 290W.
    Never saw higher than 350-375W during highest loads.

    3. DFI X48-T3RS w/ E8400 @ 3.0Ghz 1.245VCORE, 1333MHZ FSB, EVGA 9800GTX
    Idle Load: 175W, Prime Load: 200W, 3DMARK06: 250W.
    Never saw higher than 300-325W during highest loads.

    BTW, all Watts are averages. Prime Load is averages between blend testing and small FFT's.

    I really like the meter on the MFC-2. If it is at all accurate, then it supports the logic that it isn't always necessary to have a 2000KW PSU.

    I could see needing a higher wattage, QUALITY PSU, for CF or SLI + the latest GPU'S, non-extreme CPU, with everything OC'd 25-50%, and maybe a single loop WC as well.

    Just my 2 cents.
    Reply
  • Spacecomber - Monday, September 22, 2008 - link

    I was wondering if it might have been helpful to add some comments on the importance (or not) or having 8-pin 12-volt motherboard connectors, instead of the more standard 4-pin motherboard connector. I don't think that this motherboard connector, which I think largely powers the CPU, was given much attention relative to the discussion of the additional 4-pin connector for the main motherboard connector (20 + 4) and the PCI-E power connectors.

    Overall, this was a very good article. I like how it places technical details about power supplies into a context of everyday use.
    Reply
  • CSMR - Monday, September 22, 2008 - link

    The power consumption data on page 1 is completely overblown.
    The tables claim to represent actual power consumption of processors and chipsets but the figures are completely exaggerated.

    There are even people who run whole systems on one or two of the chipsets listed on less power than the power the article claims for just the chipset.

    Here are actual CPU measurements from behardware and xbitlabs:
    http://www.xbitlabs.com/articles/cpu/display/intel...">http://www.xbitlabs.com/articles/cpu/display/intel...
    http://www.xbitlabs.com/articles/cpu/display/intel...">http://www.xbitlabs.com/articles/cpu/display/intel...

    I haven't seen measurements of chipset power but here is a list of TDPs:
    http://www.silentpcreview.com/forums/viewtopic.php...">http://www.silentpcreview.com/forums/viewtopic.php...

    Must fix this soon as Anand needs to keep its reputation for good information.
    Reply
  • Zoatebix - Monday, September 22, 2008 - link

    Seriously. This guy didn't go too far our of his way to make a 30-40 watt system: http://www.silentpcreview.com/forums/viewtopic.php...">http://www.silentpcreview.com/forums/vi...tdays=0&... Reply
  • BernardP - Monday, September 22, 2008 - link

    The following paragraph from the article has me puzzled:

    "It's important to have one 12V rail supply the CPU with power and the second rail for the PCI-E slots and 6-pin connector. Unfortunately, many companies make a tremendous mistake when it comes to power distribution. We have seen several power supplies that use one 12V rail for the 6-pin PEG connector and then a second 12V rail for the CPU and 24-pin ATX connector. That means if you have a graphics card that doesn't include a 6-pin jack, both the CPU and GPU will use the same 12V rail for power. In this case, the second 12V rail goes completely unused, and users risk drawing too much current on the remaining 12V rail."

    I have an Antec Eartwatts 380. How can I find out if Antec has made the tremendous mistake or not? I want to make sure that the 6-pin connector and PCI-E slot are on their own 12V rail. Antec litterature on this PS says:

    "Dual 12V outputs: 12V2 for motherboard and peripherals, 12V1 for processor"

    It would seem OK, assuming "peripherals" includes the 6-pin connector.
    Reply
  • 7Enigma - Monday, September 22, 2008 - link

    What is your graphics card? From the article it seems to me they are saying if you do NOT have a card that requires a 6-pin PEG connector then you could possibly have an issue. If you have a card requiring a PEG connector you don't have to worry. Reply
  • BernardP - Monday, September 22, 2008 - link

    Exactly. I am planning to add a 9500GT to my existing system (integrated graphics). No power connector on that card. Reply
  • Dribble - Monday, September 22, 2008 - link

    http://extreme.outervision.com/powercalc.jsp">http://extreme.outervision.com/powercalc.jsp

    Seems to return sensible values, and not only does it cover pretty well every component you might come across, but it also understands overclocking, over volting, and allows you to enter a value for capacitor ageing.

    Also, here is a thread which someone has helpfully listed real power requirements (as given in reviews) for pretty well all graphic cards on the market right now:
    http://archive.atomicmpc.com.au/forums.asp?s=2&...">http://archive.atomicmpc.com.au/forums.asp?s=2&...
    Reply
  • drank12quartsstrohsbeer - Monday, September 22, 2008 - link

    Hey Guys: Remember that decibels is a logrithmic scale of measurement! Using a linear scale on the graph leads to inappropriate conclusions being drawn from the data. Reply
  • 7Enigma - Monday, September 22, 2008 - link

    I don't think it does. Maybe a quick *note* at the beginning of the acoustics section mentioning its logrithmic, but it is very easy to read a linear scale.

    Also, the majority of the tested systems fall well below the floor of most systems (20 decibels), so it is a moot point anyway.
    Reply
  • gmkmay - Monday, September 22, 2008 - link

    I'll start off by saying good article, however I would have liked to see a few additions.

    Other than the aforementioned new cards and overclocking information I think it would have been helpful to include common watercooling pumps and case/system fans. There is most likely a large enough set reading this that would have liked to see those added.

    The problem with the power supply issue is you have to be really careful not to get something too weak...and its really easy to forget a few small items that can quickly add up (for instance 2 pumps, 8 120mm fans, etc).
    Reply
  • mindless1 - Thursday, September 25, 2008 - link

    Nobody building a PC needs 8 x 120mm fans. Let's suppose you throttle down the fans enough that you might actually have good use for so many to have them all at very low RPM. That would tend to cause under 150mA per fan or barely over 1A total, a relatively trivial amount of power considering that even spinning up any one hard drive causes a larger momentary spike.

    A couple pumps shouldn't use all that much power either, but if you're pouring enough money into the system to have it that elaborate then why would you be on the fence about choosing a marginally capable PSU versus one with plenty of reserve power to the point where 3A one way or the other isn't a factor?
    Reply
  • Anubis - Monday, September 22, 2008 - link

    would be interesting to see numbers on just how much power OCing pulls over a non OCed system Reply
  • Christoph Katzer - Monday, September 22, 2008 - link

    I had a QX9770 just slightly overclocked with basic bios functions (for an oc-noob like me) and already then it had an increased power draw of 20% at full load compared to normal. Reply
  • 7Enigma - Monday, September 22, 2008 - link

    20% greater system power? or just for the cpu? If just the CPU what did it equate to system-wise if you don't mind me asking?

    In an earlier post I mentioned a high-end PSU possibly being better than the mid-grade if you were going to moderately overclock (it was right at the overlap point under heavy load) from a sound and efficiency standpoint. But that was assuming an increase of 20-30% overall.
    Reply
  • CK804 - Monday, September 22, 2008 - link

    Finally, an article from Anandtech that will really open people's eyes on how much power they actually need. Reply
  • Beenthere - Monday, September 22, 2008 - link

    When purchasing a PSU, ignoring the importance of PSU quality and output, in favor of noise and efficiency is foolhardy.

    Many PSUs do not provide clean power or the rated power - especially under heavy laod. In additional while the article touched on it, depending on the 12V rail design, many PSUs can't deliver the proper power (wattage) to the 12V rail(s) even though the PSU total wattage rating may be more than sufficient . While I'm all for green it is always better to buy a quality PSU that delivers at least 20% more power than you current needs, to provide update headroom and maintain good PSU efficiency and low noise.

    If you're not comparing PSU quality, power output per rail and warranty before considering efficiency, noise and cost, then you've missed the point of buying a proper PSU. While most folks do not need a 1000W PSU, many need a quality PSU that can deliver the correct power to each rail and a PSU that will last.
    Reply
  • 7Enigma - Monday, September 22, 2008 - link

    Wouldn't a high efficiency PSU by design be a quality PSU? Generally inferior parts/design are the reason for poor efficiency... Reply
  • mindless1 - Thursday, September 25, 2008 - link

    High efficiency doesn't automatically mean high quality per it's own ratings, and an old design not attempting to have high efficiency can still be using reasonably good quality parts and design, unless all your criteria revolve around efficiency being a necessary factor before you'd call a PSU "quality".

    Take server PSU for example, many don't have such high efficiency but many are higher quality than those used in PCs.

    Do you realize that more elaborate filters will reduce efficiency? To some extent, trying to maximize efficiency limits how much quality can be present.
    Reply
  • marc1000 - Monday, September 22, 2008 - link

    and it is powered by a old Pentium-D 945 (3.4ghz, 90w TDP, and it gets quite hot actually). I also have 1 HDD, 1 DVD-RW, 2GB of ddr2 ram, and one Radeon 3850 512mb.

    I was in fact using a high-quality 250W PSU, the one that came bundled with the system (I believe it is high-quality because of the build quality and the clear specs, and also because it is a system I bought from HP with "free upgrades option", so I upgraded the CPU+GPU by my own. I wanted a Core2Duo, but my mobo will not accept it.)

    whatever. the 250w PSU was working fine, even when I ran old games or 3dmark01. but on 3dmark03/05/06 and newer games, the system was turning itself off after 5 or 10 minutes. so i bought this very quiet AKASA 300w PSU with a single PEG, and now I have a relatively quiet computer that works just fine with no power problems.

    this is quite a good gaming machine, if you want to know. I am OK with 20/30 fps, as I am not a hardcore gamer. and I can play GRID at 1680x1050 with almost everything high and 2xAA at this frame rate. also Crysis run fine with everything to medium at this resolution and frame-rate (but this is not so enjoyable because it is a fast-paced FPS). I know that my CPU is the bottleneck, but maybe next year I will change the mobo+CPU.

    It is a modest PC, with a modest PSU, for a modest gamer! =)
    Reply
  • Insomniac - Monday, September 22, 2008 - link

    You said the Corsair VX450W performs best of the low power usage system PSUs. But looking at the charts, it seems the Amacrox Calmer 300W is the best. Its efficiency curve is the highest through the range and its noise curve is the lowest through the range. Was this an oversight or was there a reason this PSU would not work for the midrange system?

    If it was, that seems like a great PSU. It would be close to the other mid-range PSUs in efficiency and the best for noise. It would be great for a low power system, but has quite a bit of headroom as well.
    Reply
  • Christoph Katzer - Monday, September 22, 2008 - link

    Since it is passively cooled it has a very limited usability... If you run a few fans it won't be a problem but then you still have the huge price difference between both units. Reply
  • Insomniac - Monday, September 22, 2008 - link

    Thanks for the information! Reply
  • duploxxx - Monday, September 22, 2008 - link

    Very nice article it really explains the desires and needs what to buy for PSU altough i am missing some top psu's like seasonic for example.

    Only unfortunate is that some measurements of hardware are way out of range... especially in the motherboard parts.

    And if you want to be stay out of who is best... you know the always existing rival that a site has a preferred vendor, take the latest hardware from both sides, if not leave it out.
    Reply
  • kuraegomon - Monday, September 22, 2008 - link

    Aargh. The most important argument for ensuring that your PS has plenty of headroom is ... lifespan!

    The most knowledgeable PS people out there will all tell you the same thing: running even a quality PS at consistently more than 80% or so of its rated output is all but guaranteed to reduce its operational lifespan. It's also a catch-22 because the longer a PS is run at high load, the less the maximum load it can support becomes. This isn't a terribly quick process, but quite a sure one. Track down any number of JonnyGuru's comments/reviews out there for more info.

    Do all the math stated in this article, and figure out what your idle and load draws approximate too, then make sure you've got 30% headroom on top of your load requirement. This is BEFORE taking into account any expansion plans. Also, try to remember that 850 PCP&C supply described here STILL isn't being used the way a power user uses their system. It's spending more time at or near peak load, but it's also quite likely spending a fair bit of time on the shelf.
    Reply
  • 7Enigma - Monday, September 22, 2008 - link

    What is your point? In the article the PSU's for each category are well under 80% utilization. If you look back at the charts, in the rated range for each system:

    -low end is utilized <50% of rated wattage

    -mid-range is utilized <75% of rated wattage

    -high-end is utilized <80% except for the Neopower Blue (which honestly looks pretty crappy both from a efficiency and sound standpoint)

    That is across the board. In each category the higher-rated parts are obviously utilized less than those percentages.

    Even the comments below each chart bares this out. For the low end, for example, they state that a 250w PSU would be perfect, while even a 200w would suffice. With a total system draw of 140w, the 250w would be near 50% utilization (56% if you want to be picky), and the 200w would still be <75% utilization.

    I don't see fault in this article from that standpoint.
    Reply
  • vlado08 - Monday, September 22, 2008 - link

    The problem is that even if you calculate the expected power draw of your system you have to trust the label on the power supply and to be sure that if it says 500w then it is so. Well then you just end up to trust the trade mark or some reviews for the model you are going to by.
    Or you trust somebody who is going to assemble your new computer for you.
    Reply
  • 7Enigma - Monday, September 22, 2008 - link

    Thank you very much for this article. As someone building a system by the new year I appreciate it greatly!

    One interesting thing is that there are times where the higher wattage supplies actually make more sense due to efficiency (and probably more connectors/warranty/etc.).

    The Enermax Pro82+ 625w is definitely the best mid-range you have listed IMO for a stock system (ie non-OC'd), but for someone looking to OC their system I think the Zalman850 from the high-end section is probably the better buy (both efficiency and soundwise). There is a crossover point very close to idle power levels (if you take into account another 25-50w for an OC'd system), and so anything above idle will have better efficiency and at load quieter levels.

    But I haven't checked the price difference, which I'm assuming is quite large. A 1-2% efficiency difference between the Enermax and Zalman is probably not worth the increase in price from both a ROI (from power savings and increased case temp from inefficiency).

    Thanks again for the great review!
    Reply
  • vlado08 - Monday, September 22, 2008 - link

    In the article you didn't mention how did you measure the power draw of different components for example the CPU or the draw from PCI express? And haw did you test the efficiency of the different power supplies? Reply
  • Don Tonino - Monday, September 22, 2008 - link

    I found the remarks concerning the efficiency charts a bit misleading... why give the range of efficiency of the high end system, for example, as between 85% and 89%, when the first number refers to the efficiency with 90VAC? the numbers given out are not consistent, as the systems at 230VAC show in reality the following efficiencies approx:

    low end - 74 to 80 %
    midrange - 82 to 88 %
    high end - 87.5 to 89.5%

    Based on the numbers above, the PSU is actually quite well suited to the high system as the efficiency changes by a meager 2% between idle and load. It would be even better with some extra load, so to place the idle/load range between 450 and 700 W.

    As far as the point to make is to show how efficiency changes with the load, it would have been as meaningful to give data just for the 230VAC, as it was already stated that efficiency with 120VAC or 90VAC would be even lower.
    Reply
  • Insomniac - Monday, September 22, 2008 - link

    The range isn't for 90VAC to 240VAC, it's to cover the idle load to full load range of the sample system. Reply
  • Don Tonino - Monday, September 22, 2008 - link

    Check the chart. For every system the lower efficiency, the one given for the sistem at idle, has a value that at that particular power load (respectively given as 90W, 168W and 310W) lies on the red line, the one representing efficiency of the PSU when running at 90VAC.

    This is most evident if you take the high end system, which is stated will make the PSU run at an efficiency between 85% and 89%; those values, if you move on the blu line (PSU running at 230VAC) means a power load between 200W and 650W).

    Giving the idle efficiency with the PSU running at 90VAC and the load efficiency with the PSU running at 230VAC gives a much higher change in efficiency than real. The only real meaning for it would be to say: "with such a system and such a PSU you will have an efficiency between A% and B%, based on the current the PSU is running on"... and I seriously doubt that anyone at home have an electrical system that changes VAC on the run.
    Reply
  • Insomniac - Monday, September 22, 2008 - link

    I see what you are saying now. I misunderstood what you said before. It seems a table would be better to show the efficiency range for each, or the values for one curve only (the article seems to say it was supposed to only be 230 VAC). Reply
  • JarredWalton - Monday, September 22, 2008 - link

    Sorry for the error - not sure how we missed that, but yes the efficiency with the high-end system and UCP 900W is higher than stated initially. Must have been confused with the other systems, but I'll correct the text now. Reply
  • JarredWalton - Monday, September 22, 2008 - link

    Edit: Christoph's text reflect the range for 90VAC to 230VAC, but my editing made that a little less clear. I've added in "input voltage" comments to clarify things. Reply
  • poohbear - Monday, September 22, 2008 - link

    thanks for this article!!! im planning on running 2 8800gt's in SLI on a 80% efficient enermax 420wt psu. it has 29a on the 12v+ line so im confident it can run it. All this BS about needing 500+wts psus is nonsense if you know your cards power needs. Reply
  • bela - Monday, September 22, 2008 - link

    Where did you get those power figures? Dream last night or what?

    This is totally made up bullshit.

    The ANTI-AMD war continues @ Anandtech

    you compare 2 year old 90nm AMD DC with new 45nm Intel DC, is that a fair peer group?

    6000+ 160W load? Even with 90nm this ist ridicoulus, it should be around 110W, a new 65nm 6000+ needs less then 80 Watt, a 65nm 5000+ less then 60W, so talk about making Intel look good.
    Reply
  • elaar - Monday, September 22, 2008 - link

    bela, you seem to be incredibly rude and have also missed the entire point of the article, if it makes you that annoyed then why not do us all a favour and stop reading articles and commenting in the future.

    I for one found the article incredibly useful especially when you consider the sheer number of people who go out and buy way too powerful psu's and have no idea what they're doing.

    It doesn't matter what processor or graphics cards power stats were listed, they were just there as examples for the article, god knows how you've managed to get so confused with paranoia to believe it was an anti AMD campaign.

    Thanks Anandtech for a superb article.
    Reply
  • npp - Monday, September 22, 2008 - link

    The power draw figures for the X2 6000+ are a bit off-scale (and yes, it is an older 90nm die, apparently), it's a tiny bit, however. You can have a look at the charts here (damn, the stupid link button doesn't work):

    http://xbitlabs.com/articles/cpu/display/dualcore-...">http://xbitlabs.com/articles/cpu/display/dualcore-...

    The system equiped with an X2 6000+ was measured to draw about 304W at full load and 180W at idle. Adding ~25W to that difference makes for ~150 total power consumption, which comes close to what was stated in the article. Just because you thought "it should be around 110W" doesn't make you automaticaly right. Learn living with the truth and stop behaving like a small child.

    Furthermore, as it was properly stated, those figures were intended to draw a frame around the best and worst case scenarios, representing some of the CPUs typicaly found in a system today. They weren't meant as a CPU-to-CPU comparison.

    That old dark sense of anti-AMD or anti-Intel paranoia continues to be abundant in every discussion nowadays... What a triumph for the PR brainwashers at both camps.
    Reply
  • bela - Friday, September 26, 2008 - link

    No, they are not of scale, they are bullshit, nothing else but made up numbers.

    Look at this:

    X2 6400+ WITH Voltage Regulator, depending on Board 85,9 or 103,3 Watt


    http://ht4u.net/reviews/2008/amd_phenom_leistungsa...">http://ht4u.net/reviews/2008/amd_phenom_leistungsa...

    This ist the truth, nothing else
    Reply
  • Kiijibari - Monday, September 22, 2008 - link

    Could this be a typo ?
    106W is ok, 160W is a little bit out of the "normal" scale ..

    Furthermore .. which 6000+ is it ?

    There are 3 different models:

    one 90nm "normal" model: 125W 3,0 GHz; 2x1MB L2 (ADX6000IAA6CZ)
    one 90nm EE model: 89W 3,0 GHz; 2x1MB L2 (ADA6000IAA6CZ)
    one 65nm model: 89W 3,1 GHz; 2x512kB L2 Cache (ADV6000IAA5DO)

    cheers

    Kiiji
    Reply
  • JPForums - Monday, September 22, 2008 - link

    This is a curiosity for me as well.
    I have an A64 X2 6400+ 125W, 3.2GHz, 2x1MB L2 (don't remember the model number off hand) that doesn't seem to require near that power.

    The 6400+ is running on an nForce 570SLI with 8Gb (4x2Gb) DDR2-800.
    The video card is an 8800GTS 512Mb.
    I have 4 HDDs 2 optical drives and 8 fans (7 case + CPU fan) that according to specifications run at 8.6W when at full speed (how I have them while gaming).
    If I use the (presumably lower) power ratings used for the 6000+ and the 8800GT, and I exclude the power of usb components and the fan controller/sensor overhead, my total system consumption at load (using the values from the article) is around 450W.

    The curiosity is that the same Enermax Pro82+ 385W PSU mentioned in the article has no issue running this system. (Ironically emphasizing the point of the article) Using a basic kill-a-watt meter, I found a power draw of 378.2 was as high as it got during benchmarking, gaming, stressing the system. For reference, I tried 3DMarks 2006/Vantage, Stalker, Crysis, C&C3, and a combination of 2xPrime95 + ATItool's GPU heat up mode (rotating fuzzy block). The ATI tool combo offered the largest power draw in my system. Granted, the kill-a-watt may not be as accurate and I may not have stressed the system as well as in the article, but I suspect the power draw numbers for the 6000+ are lower than the article suggests.

    That small inconsistency aside, this was a nice article. I would like to see those power draw blocks that you overlayed on the power efficiency and noise curves included in future PSU reviews. It would be a quick and easy way to let people know how applicable the PSU being reviewed is to them. It would also be interesting to see how high the power draw gets with water cooling systems, case mods (I.E. cold cathodes), and the likes.
    Reply
  • Christoph Katzer - Monday, September 22, 2008 - link

    My 6000+ was 90nm, yours? Reply
  • Kiijibari - Monday, September 22, 2008 - link

    He has a 6400+, that CPU is 90nm only (so far).

    But he has a AM2 mainboard, maybe you had a AM2+ board, and the onboard VRMs are running badly with a AM2 CPU ?

    cheers

    Kiiji
    Reply
  • Christoph Katzer - Monday, September 22, 2008 - link

    Yes I guess that was the problem. Thanks for pointing that out! Reply
  • ViRGE - Monday, September 22, 2008 - link

    Christoph, thanks for the excellent article. There is one thing I noticed however that has me a bit confused. In a few of your video card tests, you have the cards pulling upwards of 100 watts from the PCIe slot. It has been my understanding, and is mentioned in AT's X38 chipset article*, that the limit for a PCIe slot is 75W. Your results put this value in contention, so I'm not sure what to believe at this point.

    Is it that I am mistaken that the limit is 75W, or are your video cards pulling too much power from the slot, or is there a problem with the data?
    Reply
  • Christoph Katzer - Monday, September 22, 2008 - link

    Thanks. The data is correct since the cards draw power through the PCIE slot and the additional PEG connectors (6-pin or 8-pin). Towards the end of the article you will find a list which card drew how much power from the PCIE slot and the PEG connector. 75 watts is however correct for the older PCIE slots, PCIE2.0 can draw up to 150 watts. Reply
  • mobutu - Monday, September 22, 2008 - link

    Where is AMD 4850?
    Also almost all the graph cards used are kind of old, previous generation.
    Reply
  • Dobs - Monday, September 22, 2008 - link

    Here is a link to a comprehensive list of all graphics cards.

    Reply
  • Dobs - Monday, September 22, 2008 - link

    damn... link didn't work

    Copy and paste :)

    http://archive.atomicmpc.com.au/forums.asp?s=2&...">http://archive.atomicmpc.com.au/forums.asp?s=2&...
    Reply
  • Clauzii - Monday, September 22, 2008 - link

    Nice lists, thanks :) Reply
  • Christoph Katzer - Monday, September 22, 2008 - link

    As stated I didn't have all of the new stuff here and this article anyway focuses primarily on power supplies and how to choose them and not the power consumption of the components... but it might come soon, who knows. Reply
  • LTG - Monday, September 22, 2008 - link

    >>our labs don't have the latest GPUs available for testing

    Isn't this kind of a bad thing for a site with so many enthusiast readers?

    First two caveats:

    1) There are a lot of article bashers out there who nit-pick, give rude feedback, or are just plain wrong, I don't support these guys or want to be one.

    2) The article written was very good and helpful, thank you.

    However it doesn't matter that the article "primarily focused on how to choose power supplies".

    - Leaving out a GTX 280 was a big omission. Even better to have an stock OC'd product.

    - Leaving out very basic overclock numbers was a big omission. For example it's very common for readers to have an Intel QC clocked at 3.6Ghz since it's so doable.

    Having the last two data points would be very helpful for those building top systems, and would even be interesting for those who are not.



    Reply
  • 7Enigma - Monday, September 22, 2008 - link

    As mentioned, the article is about the myth that we all need 1k PSU's for the "best experience". Bottom line this article shows that the vast majority of people do not need more than a 500-600w psu. How is leaving out any stock product an omission? Just look at the TDP of the product and go from there. All stock products will fall within a narrow range.

    Once you get to OC'ing I also think this is a difficult thing to do. Individual chips, even from the same batch can have very different properties and require different voltages to reach the same speeds. I'm sure if the author had put that his Intel quad drew 120w at 3.6GHz, someone would have complained that their's took 135w, and someone else only 110w...Once you go above stock, it is not cut and dry and as the article states you should have a 20-30% extra buffer if you wish to OC.

    I feel the scope of the article just doesn't require these extra requests. Sure it would be nice to see just how crazy of a system you can make (if I remember correctly they did that a while ago to actually use the >1000w supplies), but this article was to show what a normal base, mid-grade, and high-end setup would require.

    I've complained in the past about articles, but this one I don't see any serious faults.
    Reply
  • LTG - Monday, September 22, 2008 - link

    You can't just look at the TDP's, that's the problem.

    When you overclock the power demands can increase in a very non-linear way.

    So for example I can't plan for one of the most popular 280 cards like the EVGA FTW 280 GTX. It's 11% core over clocked, what does that mean at the outlet? No good way to guess.

    Same for a 3.6Ghz QC CPU - this is a very common overclock, yet there is no direct way to know it's power requirements.

    I'm just saying this is not esoteric information, this would be data people really could use and can't get from the manufacturer.

    Reply
  • xaris106 - Friday, November 07, 2008 - link

    But you can. All you need is stock power consumption at load(Pstock), stock voltage(Vstock) and stock frequency(Fstock) The oc power is then:
    Poc = Pstock * (Foc/Fstock) * (Voc/Vstock)^2
    Reply
  • nubie - Monday, September 22, 2008 - link

    Unless they edited this, you are operating on a false assumption.

    (Unfortunately, our power supply testing labs didn't have the latest GPUs available for testing.)

    A power supply testing lab doesn't need to have every component on hand because it uses a test bench to load the supplies.

    Great article, way to dispel myths, I guess since I only plan to overclock with a single video card and one or two hard drives my PCPower Silencer 470 will be enough power for many years to come (which is what I hoped when I bought it, the only downside is the single 6-pin for the video card, when it can clearly handle much more.)
    Reply
  • JarredWalton - Monday, September 22, 2008 - link

    Just for the curious, AnandTech staff is scattered far and wide around the globe (well, at least the US and Europe). I'm west coast, Wes is east coast, Anand and Derek are in NC, Gary is in TX, and we have Johan and Liz in Belgium with Christoph in France. (That's not everyone, but you get the point.) Since we tend to focus on our own areas of testing, Derek and Anand have the most CPU/GPU hardware, I have laptops and displays, Gary has motherboards, etc. I can definitely say that Christoph isn't the only one without 48x0 and GTX 2x0 hardware. [Pardon me while I go cry in a corner now....] Reply
  • hyvonen - Wednesday, June 03, 2009 - link

    Oh, so in order to get this power draw info on more components, I should beg Anand? :) Reply
  • LTG - Monday, September 22, 2008 - link

    Totally understood, many companies now days are distributed and can't have every physical resource available to every person.

    However I would volunteer to send Christof a new 280 GTX to test if he decides it's worth it.

    Serious - Just please send it back whenever finished :). And I waive all claims if it is accidentally fried by that fancy Chroma thingy.

    Reply
  • ineedaname - Tuesday, November 02, 2010 - link

    This article is well written and tried to put real life numbers and situations to the test on PSUs.

    However i felt that they should mention one other thing for people who are novice to computers. They should mention that just because a PSU is rated for 500w it doesn't necessarily mean that it will do the job even if your computer will only suck about 150w max. Because a 500w psu that comes along with a $50 case just won't do the job. Not because of the wattage rating but because they use crappy parts and workmanship; it'll just die in 3 months when the warranty is over.
    Reply

Log in

Don't have an account? Sign up now