POST A COMMENT

39 Comments

Back to Article

  • SodaAnt - Wednesday, September 03, 2014 - link

    I still tend to be slightly annoyed at the low efficiencies with high power PSUs at around 5% load or 50-75W. I noticed the article even said "Obviously using a 1200W PSU to drive a 60W load on a regular basis would be rather odd, so in most use cases the efficiency will be much higher." However, I don't really think that's true. Maybe that would have been the case 5 years ago, but now even monster rigs will easily use less than 100W when they are just idle at the desktop, which is a surprising amount of time. Further, there's quite a large subset of users that get a 1200W power supply, yet use it in a build with a single GPU and CPU, merely because they think they might upgrade later.

    In short, efficiencies at the low end should be higher.
    Reply
  • LtGoonRush - Wednesday, September 03, 2014 - link

    76% efficiency at 5% load is EXTREMELY efficient, especially for a high-current power supply. There's a certain minimal overhead to power the electrical conversion hardware, and while it doesn't affect semi-passive models, the fan alone can account for 5% or more of efficiency losses at 50W draw. This overhead is also affected by the capacity of the power supply, as larger or parallel components increase overhead. The fact that a 1050W power supply is only wasting 12W of power at low load is quite respectable, and means that any improvements would have very minimal benefit. Reply
  • SodaAnt - Wednesday, September 03, 2014 - link

    I think that wasting 12W of power at low load is actually quite a lot, mainly because the time scale is different. I'd say at least for my use case, 70% of the time is spent basically idling (internet browsing, writing a word doc, email, etc) or just not even using the computer at all. Especially for users who don't bother to turn off their computer, the 12W loss starts to add up pretty quick. I know its not easy to do, but I feel that because its not part of the spec for 80 plus, not very much effort is going into making improvements. Reply
  • RaistlinZ - Wednesday, September 03, 2014 - link

    I think you expect too much. Anyone buying this is likely building a high-power overclocked rig, and wouldn't be spending much time at 5% power draw anyway. I also doubt the time, money, and effort into making PSU's 90%> efficient at 0-5% power draw is worth it from a manufacturing standpoint anyway. Reply
  • fokka - Thursday, September 04, 2014 - link

    just because people are overclocking and gaming for a couple hours every day doesn't mean the close-to-idle times are negligable. people will still browse the web, write an email or watch a video or two.

    i'm not saying the inefficiencies at idle have to amount to a big loss compared to running a 1kw-pc at full load, but taken for itself it's still enough to power a complete low powered computer.

    i don't know about the engineering/physics side of things, but i'd say idle efficiency is still something manufacturers should look to improve.
    Reply
  • Kutark - Sunday, September 14, 2014 - link

    The point is primarily that there is a minimum overhead that is needed to operate a power supply, and reducing that overhead is extremely difficult and in the end is not worth it. For example, if it takes say 12w of overhead, reducing that by 25% would basically put you at 9w. The return for the effort is simply not worth it (increased component costs, engineering/design costs, etc). Not many people would pay say a 10% price premium just to have the 5% load efficiency be 10% better so they can save basically save $6 a year in electricity costs.

    I've honestly never understood why people can't seem to pick their battles with these things. I always love when i would see people get into arguments here over a HDD that uses 8.2w instead of another one thats say 9.6w because its a 15% difference and thats massive! Well the reality is, its 1.4w. Now, in a corporate SAN environment where you might have say 500 of these drives in an array, yeah, 15% is a big deal. But to a home user who might have at most 2 of these drives in their computer, you're talking about $4/yr in electricity costs.

    Pick your battles.
    Reply
  • dealcorn - Wednesday, September 03, 2014 - link

    If you use your over-clocking rig to browse the Internet you are wasting lots more than 12W. Get a petite, passively cooled Bay Trail desktop for browsing. Save the big rig for when you need the horsepower. Reply
  • rarson - Sunday, September 21, 2014 - link

    Spend several hundred extra dollars to save a few bucks a year? That doesn't make any sense at all. Reply
  • surt - Wednesday, September 03, 2014 - link

    Why does any power supply need to run a fan at 50w draw. Even if you're only 50% efficient at that load, you shouldn't need to actively cool 25w. If that's your overhead, just start the fan later. Reply
  • Samus - Thursday, September 04, 2014 - link

    I want to chime in and agree with the OP here. Any PSU over 600W is going to be inherently inefficient at idle in many PC's.

    Below are my overclocked i5-4790 PC's individual power draw specifications taken with my Fluke multimeter, GW INSTEK GDS-122 oscilloscope using additive\deductive component elimination to calculate these measurements at idle within a reasonable margin of error using a PC Power & Cooling 750-QUAD 80-Plus PSU (which is really irrelevant since I am measuring current draw and using ohms law to convert into watts - the PSU has no bearing on the components individual power consumption.) However there is no reliable way to measure below 1 watt so the individual minimum component draw is such.

    Core i5 4690K @ 4500MHz 1.210v draws 12.8w at idle (FIVR reported via CoreTemp)
    2x8GB Crucial LPDDR3 @ 1600MHz 9-9-9 1.35v memory draws 3.2w at idle
    Samsung 840 Pro 256GB SSD 1.0w at idle
    Seagate 4TB 5900RPM 2.8w at idle
    WD 2TB 5200RPM (2.5") 1.7w at idle
    ASRock H87M-ITX measured via ATX connection 3.3+5.0+12.0 deduction 17.5w at idle
    EVGA nVidia Geforce 780Ti 3GB clocked 1025/7400 1.189v 22.8w at idle

    Total system draw before PSU efficiency loss is 62 watts. Since my PSU is only 75% efficient at 5-10% load, like most PSU's, my total system idle measured at the wall (with a somewhat accurate kill-a-watt) is 85 watts.

    However, I need a 750-watt PSU as my load power consumption can surprisingly eclipse 600-watts at full CPU + GPU load and I want to remain below the 80% load barrier where the PSU lifespan is shortened, temperatures and noise escalate and efficiency can REALLY take a dive.

    I think more attention needs to be taken with PSU efficiency at low idle state. There are various ways to do this, the most practical implementation being a dual stage power delivery circuit for different loads. Basically it would be a switching PSU inside a switching PSU, and it wouldn't be very expensive to produce because the first stage (lets say 0-100w) would be a very basic CLASS V, then there would be a voltage regulator and relay system that switches to the next stage with a supercap to isolate any voltage drop while switching.

    I'm surprised nobody has done this yet. Should I patent it?
    Reply
  • ijozic - Thursday, September 04, 2014 - link

    It sounds like it would cost noticeably more for something most people PROBABLY won't really care about. And the market is already quite competitive.. Reply
  • Samus - Thursday, September 04, 2014 - link

    The only complicated factor is space inside the standard ATX housing, but considering there are 500-watt 80+ Gold SFX PSU's that are 1/3rd the size of ATX, making a 100-watt primary stage with a 1000-watt secondary stage shouldn't be impossible. The most expensive part would likely be the cap for the relay (1/4 farad) and obviously that cap is going to be pretty large, most electrolytic caps this size are around 5" long.

    I think the BOM would be ~$20, and engineering it in isn't rocket science. This could totally be done, and if I were paying $200+ for a PSU, I'd expect some level of engineering beyond a good bridge design and PFC.
    Reply
  • jann5s - Thursday, September 04, 2014 - link

    have a look at the pico-PSU, these go up to 160W and are approximately the size of the ATX-plug attached to the umbilical coard of your PSU, this will definately fit. Reply
  • DanNeely - Thursday, September 04, 2014 - link

    With a picoPSU you need to look at not just the 12V to 3.3/5V (smaller models) or 19V to 12/5/3.3V circuitry on the ATX plug; but the size of the external power brick used to feed it. It's still probably feasible volumetrically, but isn't nearly as much of a gimmie as just looking at the size of the ATX plug module implies. Reply
  • DanNeely - Thursday, September 04, 2014 - link

    ... especially since this could potentially be done by beefing up the existing Vsb circuit instead of adding something entirely new. JonnyGuru's testing shows current Vsb implementations aren't particularly impressive at only around 80%; but depending on how much worse current models perform below 10% some gains might still be reasonable. With the Titanium spec mandating 90% efficiency at 10% load though I'm not sure it's really worth it. On the 300-700W units that are suitable for ~90% of enthusiast systems 10% of rated power is generally enough to cover idle loads as long as you don't do anything silly (eg running an IGP powered low power mITX board with an SLI capable PSU). OTOH idle loads are steadily improving with each new CPU/GPU generation and the 80+ group will need to extend the standard to be tougher somewhere for whatever comes after titanium. Reply
  • TelstarTOS - Thursday, September 04, 2014 - link

    I totally agree. What's holding me to get a 1000W+ PSU is the efficiency at low loads. With CPUs getting more and more green, PSUs should do the same and not only when under full load (which is a countersense if you really think about it). Reply
  • DanNeely - Thursday, September 04, 2014 - link

    More attention is being paid to lower end efficiency than in the past; and this PSU is a prime example. While it falls short of the (new) 90% efficient at 10% load requirement for 80+ titantium, it's ~89% efficiency is much better than any of the other PSUs reviewed earlier this year did; none of which did better than the low to mid 80's at that level. Reply
  • Phillip Wager - Monday, September 08, 2014 - link

    i like the dual stage idea but i am assuming most consumers are not willing to pay more for additional cost/complexity because that small amount of power that you are wasting turns out to be a very small amount of money over the lifetime of the system. I am somewhat suprised with your test results with your system because my computer has nearly identical specs and i have a 600 watt power supply and i have never had so much as a hiccup. Maybe get a different device for measuring power draw or perhaps your power supply is not as inefficient as you think. Reply
  • bsim500 - Thursday, September 04, 2014 - link

    I agree, but there's not much you can do. It's the nature of high-wattage PSU's. If they came out with a 95% efficient "super platinum" PSU, lower wattage PSU's would adopt the same technology and the "idle gap" would still remain.

    Personally, I have a Seasonic G360 (yes, 360W PSU) that perfectly feeds an i5 + 7870 GFX card, and pulls (measured at the wall) : 182w peak Prime + Furmark / 110-150w typical gaming load / 88w 4T Prime or X264 encode load (CPU only) and 37w idle (which falls to 26w if discrete GFX card is removed, and that's with 1600MHz Ivy Bridge idle, not 800MHz Haswell), never pulling more than 50% PSU draw even under combined synthetic CPU+GPU "power viruses", and spending 90% of its time at +90% PSU efficiency with a virtually silent fan.

    That 26-37w idle wattage on these PSU's is barely 50-60% efficiency level, yet the number of people who "swear" they "need" a 1200w PSU just for a single mid-range dGPU gaming rig is quite comical. ;-)
    Reply
  • Samus - Thursday, September 04, 2014 - link

    I know, "right sizing' a PSU is something few people understand. The worst thing you could do to your wallet is spend a fortune on a 600+ watt PSU and spend another fortune paying for it to inefficiently idle over the years for that i5 with integrated graphics and a single hard disk... Reply
  • roncat - Thursday, September 04, 2014 - link

    Why would you not just use a separate power supply just for the video card? Could you not just make the commons "common" for all PSU, and use the appropriate size for each application to be the most efficient. I assume the PSU outputs are fully floated. Anyone who knows why this may not work (ground loop hell, etc.), please reply. Reply
  • Phillip Wager - Monday, September 08, 2014 - link

    this used to be an actual thing that people would do about 10 years ago when sli and crossfire just started coming out. back then everybody had full towers anyway that could house two PSUs the top one would be fore the system and drives and the bottom psu would be for your graphics card. back then 850-1000 watt power supplies were like $600 so getting two 350-500 watt power supplies was a more cost effective option. also cooler master and evga and a lot of brands made power supplies that just had 2-4 VGA eight pins coming off of it and it was marketed to gamers as video card power supplies and it was going to provide more stable power to your videocards and blah blah blah. Now the trend is to make everything SMALLER the majority of regular people who have home pcs their only home pc is actually a laptop or an all-in-one. The trend for gaming pcs is to make small game-console sized PCs and most full sized chasis these days dont even have the screw holes to fit a 2nd power supply on top of the motherboard anymore. and like i said in another post spending $70-150 to save at most 80 dollars of power over the life of the system is hardly worth the initial upfront cost especailly to deal with fitting another bulky heat source in your chasis. Reply
  • bebimbap - Thursday, September 04, 2014 - link

    From an economical standpoint, "fixing" the lower efficiencies at almost no power draw is not worth your money. for example, and extra 15w of power 24/365 is only 131kwh most places in the us is about $.10 a kwh. why not just let it go to sleep, or shut off, most computers with ssds boot up to desktop in less than 30 seconds. But I digress, so continuing on, the inefficiency will cost you $13 more year for never using your computer. you have to subtract any time you are at non-idle state. but if you have that power supply for 8 years and not use the computer that entire time of course it adds up to $100 for keeping a computer idle 100% of the time 24/365. But if a psu was made to be 90% efficient at 10w AND 90% efficient at 1200w i guarantee you, you'd pay $200+ more per power supply for that efficiency or it'd be the size of 2 power supplies since you'd basically need 2. wasting 2x the money is also inefficient. Reply
  • Laststop311 - Wednesday, September 03, 2014 - link

    230 is pretty expenive for the 1050 watt. I got the 80+ platinum 1000 watt seasonic for 185 45 more dollars for 50 watts is pretty crap Reply
  • Iketh - Wednesday, September 03, 2014 - link

    yea I was hoping to get a quick explanation of the differences between the 1050W and their 1000W, especially considering the price difference Reply
  • Iketh - Wednesday, September 03, 2014 - link

    Nicely written article. The level of detail was superb. Reply
  • Freakie - Wednesday, September 03, 2014 - link

    Dat voltage regulation though.

    I'm actually surprised the ripple was so high, usually Seasonic units completely beat every other unit in their class. But still, only PSU's I will put in my personal systems are Seasonic. Pretty legendary quality and performance.
    Reply
  • SammichPG - Thursday, September 04, 2014 - link

    While I understand that there's a market for this kind of unit and manifacturers send them for the halo effect, try to get some samples of low wattage units please.

    Most people will never go over 150-200W real world power consumption and others might be building their own nas so quality and silent power supplies in the 300W range are not only interesting but useful to a much bigger audience.
    Reply
  • RussianSensation - Thursday, September 04, 2014 - link

    Considering modern GPU's by themselves use 250-280W, and many gamers on our site use 2 of these, just in GPU power alone a solid 500W PSU is the minimum for a modern gaming rig. If all you are building is a core i3 system with no discrete GPU, then sure your system will use only 150W total.

    I mean really dual overclocked 780s, overclocked i7 5820K and a 1000W PSU that runs 600-700W at load keeps the system quiet.
    Reply
  • romrunning - Thursday, September 04, 2014 - link

    I think you're validating his point more than not. Today Nvidia and AMD are looking at dropping the power consumption of their GPUs, not increasing. On this site, the R9 285 was listed at 190W, and most of the R9 28x series were listed at 250W. You would have to move up to a huge card (like a Titan or R9 295x2) or dual-GPUs (SLI/Crossfire) to exceed the capabilities of a 450-550W PSU.

    I would say most gamers are going with the most powerful single card that they can afford, more so than dual-GPU options. Also, I think most gamers don't buy the extreme high-end, but they usually go for the one that provides the most bang-for-the-buck.

    So, frankly, these reviews of high-output PSUs benefit very few. I believe more reviews of 450-550W PSUs along with looks at the power requirements of a few typical system builds a gamer might build with his own money would be more beneficial than a review of 1000W+ PSUs. Sure, do a few of them once in a while, but do more of the lower-output, more commonly-used PSUs.
    Reply
  • Impulses - Thursday, September 04, 2014 - link

    I'm not sure I buy that NV and AMD are putting efficiency at the forefront, seems it's kinda cyclical for each ands then they regress.

    Case in point, I recently had to buy a new PSU because the 750W that could easily handle two AMD HD6950 in CFa couple years ago wouldn't be enough for two new R9 290. Ended up with a Seasonic X-1250 just because it was on sale (was looking at 1K units otherwise).

    Seasonic's biggest competition seems to be themselves (they now have three top end lines? the Gold X models and two Platinum lines?) and the various brands that resell Seasonic designs (XFX etc).
    Reply
  • romrunning - Thursday, September 04, 2014 - link

    I'm curious - how much power does your system actually draw?

    A link here (http://www.anandtech.com/show/8422/amd-announces-r... said the R9 290 draws 250W. Even if they spike to 275-300W under heavy load, you still have 150W for everything else. Are you running a 220W AMD CPU or something?
    Reply
  • Impulses - Friday, September 05, 2014 - link

    I could maybe get away with the 750W unit, but I wasn't really looking forward to running it at 100% load thru hours of gaming, and/or unexpected crashes while OC'ing etc. Reply
  • DanNeely - Thursday, September 04, 2014 - link

    There's something wrong with the -12V column in the table. 3A @ -12V would be 36 watts of power not 6. 3A also seems unusually high for a current generation model. Reply
  • E.Fyll - Thursday, September 04, 2014 - link

    Thank you for pointing that out Dan. The 5VSB and -12V columns were reversed, that was a typo mistake on my part. Reply
  • WilbertCane29 - Friday, September 05, 2014 - link

    I found this to be really interesting. Are there any more sources on this? I would like to know more about this. Reply
  • Oxford Guy - Wednesday, December 03, 2014 - link

    Typical review site doublespeak:

    The power supply can be "ear splitting" and "be heard from rooms away" but:

    "we do believe that very few users actually require their system to remain silent while it is heavily loaded."

    Give me a freakin' break. 50+ db is related to a desire for silence how, exactly? Not having something cause someone to develop tinnitus during the warranty period has to do with a quest for silence how, exactly?

    "it is known that such powerful units have a very limited target audience."

    Yes, the deaf.
    Reply
  • Oxford Guy - Wednesday, December 03, 2014 - link

    It's long past time for these antique standards like ATX with 120mm fans be replaced. If you're going to provide 1000 watts of power or more, move to a new form factor. How about a 200mm fan with a vapor chamber?

    Think outside of the bloody box! And, stop apologizing for these poor thoughtless designs. This power supply is unsuitable for use out side of a server room.
    Reply
  • Oxford Guy - Wednesday, December 03, 2014 - link

    It pisses me off to no end to see the same regurgitation of anachronism year after year, with the most miniscule changes being hailed as great innovation. Why can't companies actually innovate?

    Make the entire top of a PC case a triple width 360mm radiator instead of these absurd tower designs with redundant narrow radiators and fans splattered all over the place, competing with each other. Link in the GPUs, the CPU, the motherboard VRMs, and the power supply with heat pipes and a giant vapor chamber. Do something instead of this piecemeal and pathetic series of attempts to shoehorn high wattage equipment into form factors that don't work.

    I remember when Apple told people to drop their Apple IIIs onto desks to reseat the chips that popped out from heat — because Apple designed the case to be oh-so-pretty before it finished engineering the internals! This is the same nonsensical mentality that dominates the "enthusiast" equipment industry and review sites' appraisals. The assumption is that the form factors already exist and must be slavishly followed, no matter how idiotic the result.

    But, while I sit around unemployed in disgust in the midst of cornfields I get to see the parade of banality continue...

    I realize that Silverstone has done a bit with its triple and dual 180mm fan rotated designs but those are not adequate for really quiet computing with high wattage multiple GPU machines. People even suggest pairing blower GPUs with those cases. Blower GPUs! Blowers on GPUs should have been consigned to the relic bin 10 years ago. Sure... let's pair a high wattage part with a teeny tiny little fan and form factor. Then, let's put two or three into the system to double or triple the stupidity. That makes sense!

    Linked heat pipes, linked radiators, big vapor chambers, single air path, the case acting as a heat sink as well... Centralized non-redundant non-competitive cooling mechanisms. That is the future, not more of the same 50+ db just for the power supply BS, teeny tiny fans, 5 or 6 airflow paths in different directions, and general anachronistic chaos disguised as cleverness.

    I know it makes more money for manufacturers to keep people in the upgrade mouse wheel, trying to get a decent product by cobbling together all the half-hearted efforts but it's time to start delivering a holistic well-made product without paying Mac Pro prices or worrying about Apple's obsession with glamour.

    Get with it, people.
    Reply

Log in

Don't have an account? Sign up now