Power Supply Efficiency FAQ
The Splitting of the +12V railWhat is “efficiency”?
Efficiency is the ratio between the useful output of an energy conversion device and the input. For example, if your computer uses 300W, but pulls 400W from the wall, then the efficiency is 300W/400W, or 75%.
Computer power supplies are typically 75% efficient, especially those units included with computer chassis or units made more than a couple of years ago before power supply efficiency was made such a priority. The BFG GS, LS, MX and ES power supplies are typically 80% efficient or better.
Why is efficiency important?
Quite simply, if your power supply is more efficient, your computer will use less power. Depending on how much you pay for power from your utility company and how much power your computer typically uses, you can save anywhere from $1 to $10 per year, per computer… perhaps more! Furthermore, because any AC power that is NOT converted into DC power is exhausted as heat, a more efficient power supply inherently runs cooler. Not only does this mean your office is going to be cooler, but also allows the power supply manufacturer to use a slower, quieter fans to cool the power supply.
Aren't higher wattage units less efficient at lower loads?
With all things being equal, yes. But you can't always compare brand A with brand B and assume that because brand B is a higher wattage that it's going to be less efficient at lower loads.
It is true that most power supplies are only at their most efficient when the load on them is 20% or more of their capability. So with conventional power supplies, you pretty much have to throw any kind of green initiative out the window when trying to buy a power supply that will allow for future expansion; like buying a second graphics card for SLI or adding more hard drives to run a RAID array.
Of course, even conventional power supplies vary from unit to unit, and since the initiative to be more efficient is a relatively new concept be aware that even if a modern day computer power supply is only 80% efficient from loads of 20% and up, it may be 77% or 78% efficient at a 10% load and this may still be much more efficient than the power supply you're replacing, even at it's best efficiency!
What is 80 Plus?
For a fee, 80 Plus will test your company's power supply to confirm that is over 80% efficient at 20%, 50% and 100% loads. Recently, 80 Plus expanded to include bronze, silver and gold certifications for power supplies that are over 82%, 85% and 87% respectively. Naturally, because of multiple PSU companies sharing the same platform and the cost of certification, the list is not all inclusive... but it's a heck of a brilliant start:http://80plus.org/manu/psu/psu_join.aspx
How Power Supplies Are RatedWhat is "multiple +12V rails", really?
In most cases, multiple +12V rails are actually just a single +12V source just split up into multiple +12V outputs each with a limited output capability.
There are a few units that actually have two +12V sources, but these are typically very high output power supplies. And in most cases these multiple +12V outputs are split up again to form a total of four, five or six +12V rails for even better safety. To be clear: These REAL multiple +12V rail units are very rare and are all 1000W+ units (Enermax Galaxy, Topower/Tagan "Dual Engine", Thermaltake Tough Power 1000W & 1200W, for example.)
In some cases, the two +12V rail outputs are actually combined to create one large +12V output (Ultra X3 1000W, PC Power & Cooling Turbo Cool 1000W, for example.)
So why do they do they split up +12V rails??
Short circuit protection only works if there's minimal to no resistance in the short (like two wires touching or a hot lead touching a ground like the chassis wall, etc.) If the short occurs on a PCB, in a motor, etc. the resistance in this circuit will typically NOT trip short circuit protection. What does happen is the short essentially creates a load. Without an OCP the load just increases and increases until the wire heats up and the insulation melts off and there's a molten pile of flaming plastic at the bottom of the chassis. This is why rails are split up and "capped off" in most power supplies; there is a safety concern.
Is it true that some PSU's that claim to be multiple +12V rails don't have the +12V rail split at all?
Yes, this is true. But it's the exception and not the norm. It's typically seen in Seasonic built units (like the Corsair HX and Antec True Power Trio.) It's actually cheaper to make a single +12V rail PSU because you forego all of the components used in splitting up and limiting each rail and this may be one reason some OEM's will not split the rails, but say they are split. Some system builders adhere very closely to ATX12V specification for liability reasons, so a company that wants to get that business but also save money and reduce R&D costs will often "fib" and say the PSU has it's +12V split when it does not.
Why don't those PSU companies get in trouble? Because Intel actually lifted the split +12V rail requirement from spec, but they didn't actually "announce" it. They just changed the verbiage from "required" to "recommended" leaving system builders a bit confused as to what the specification really is.
So does splitting the +12V rails provide "cleaner and more stable voltages" like I've been told in the past?
It is true that marketing folks have told us that multiple +12V rails provides "cleaner and more stable voltages", but this is usually a falsehood. Quite frankly, the use this explaination because "offers stability and cleaner power" sounds much more palletable than "won't necessarily catch fire". Like I said before, typically there is only one +12V source and there is typically no additional filtering stage added when the rails are split off that makes the rails any more stable or cleaner than if they weren't split at all.
Why do some people FUD that single is better?
Because there are a few examples of companies that have produced power supplies with four +12V rails, something that in theory should provide MORE than ample power to a high end gaming rig, and screwed up. These PSU companies followed EPS12V specifications, which is for servers, not "gamers". they put ALL of the PCIe connectors on one of the +12V rails instead of a separate +12V rail. The +12V rail was easily overloaded and caused the PSU to shut down. Instead of correcting the problem, they just did away with the splitting of +12V rails altogether. Multiple +12V rail "enthusiast" PSU's today have a +12V rail just for PCIe connectors or may even split four or six PCIe connectors up across two different +12V rails. The rails themselves are capable of far more power output than any PCIe graphics card would ever need. In fact, Nvidia SLI certification these days REQUIRE that the PCIe connectors be on their own +12V rail to avoid any problems from running high end graphics cards on split +12V rail PSU's.
There's less components and less engineering to make a PSU that DOES NOT have the +12V rail split up, so it's cheaper to manufacturer (about $1.50 less on the BOM, $2 to $3 at retail) and typically this cost savings is NOT handed down to the consumer, so it actually behooves marketing to convince you that you only need single +12V rails.
But some people claim they can overclock better, etc. with a single +12V rail PSU
B.S. It's a placebo effect. The reality is that their previous PSU was defective or just wasn't as good as their current unit. If the old PSU was a cheap-o unit with four +12V rails and the new one is a PCP&C with one +12V rail, the new one isn't overclocking better because it's a single +12V rail unit. It's overclocking better because the old PSU was crap. It's only coincidental if the old PSU had multiple +12V rails and the current one has just one.
The only "problem" the occurs with multiple +12V rails is that when a +12V rail is overloaded (for example: more than 20A is being demanded from a rail set to only deliver up to 20A), the PSU shuts down. Since there are no "limits" on single +12V rail PSU's, you can not overload the rails and cause them to shut down..... unless you're using a "too-small" PSU in the first place. Single +12V rails do not have better voltage regulation, do not have better ripple filtering, etc. unless the PSU is better to begin with.
So there are no disadvantages to using a PSU with multiple +12V rails?
No! I wouldn't say that at all. To illustrate potential problems, I'll use these two examples:
Example 1:
An FSP Epsilon 700W has ample power for any SLI rig out there, right? But the unit only comes with two PCIe connectors. The two PCIe connectors on the unit are each on their own +12V rail. Each of these rails provides up to 18A which is almost three times more than what a 6-pin PCIe power connector is designed to deliver! What if I want to run a pair of GTX cards? It would have been ideal if they could put two PCIe connectors on each of those rails instead of just one, but instead those with GTX SLI are forced to use Molex to PCIe adapters. Here comes the problem: When you use the Molex to PCIe adapters, you have now added the load from graphics cards onto the rail that's also supplying power to all of your hard drives, optical drives, fans, CCFL's, water pump.. you name it. Suddenly, during a game, the PC shuts down completely.
Solution: To my knowledge, there aren't one-to-two PCIe adapters. Ideally, you'd want to open that PSU up and solder down another pair of PCIe connectors to the rails the existing PCIe connectors are on, but alas... that is not practical. So even if your PSU has MORE than ample power for your next graphics cards upgrade, if it doesn't come with all of the appropriate connectors, it's time to buy another power supply.
Example 2:
Thermal Electric Coolers take a lot of power and are typically powered by Molex power connectors. I, for one, prefer to run TEC's on their own power supply. But that's not always an option. If you had a power supply with split +12V rails and powered your TEC's with Molexes, you would be putting your TEC's on the same +12V rail as the hard drives, optical drives, fans, CCFL's, water pump.. you name it, just as you did with the Molex to PCIe adapters. The power supply could, essentially, shut down on you in the middle of using it. A power supply with a single, non-split, +12V rail would not have any kind of limit as to how much power is delivered to any particular group of connectors, so one could essentially run several TEC's off of Molex power connectors and not experience any problems if one had a single +12V rail PSU.
Typical multiple +12V rail configurations:
* 2 x 12V rails
o Original ATX12V specification's division of +12V rails.
o One rail to the CPU, one rail to everything else.
o VERY old school since it's very likely that "everything else" may include a graphics card that requires a PCIe connector.
o Typically only seen on PSU's < 600W.
* 3 x 12V rails
o A "modified" ATX12V specification that takes into consideration PCIe power connectors.
o One rail to the CPU, one rail to everything else but the PCIe connectors and a third rail just for PCIe connectors.
o Works perfectly for SLI, but not good for PC's requiring four PCIe connectors.
* 4 x 12V rails (EPS12V style)
o Originally implemented in EPS12V specification
o Because typical application meant deployment in dual processor machine, two +12V rails went to CPU cores via the 8-pin CPU power connector.
o "Everything else" is typically split up between two other +12V rails. Sometimes 24-pin's two +12V would share with SATA and Molex would go on fourth rail.
o Not really good for high end SLI because a graphics card always has to share with something.
o Currently Nvidia will NOT SLI certify PSU's using this layout because they now require PCIe connectors to get their own rail.
o In the non-server, enthusiast/gaming market we don't see this anymore. The "mistake" of implementing this layout was only done initially by two or three PSU companies in PSU's between 600W and 850W and only for about a year's time.
* 4 x 12V rails (Most common arrangement for "enthusiast" PC)
o A "modified" ATX12V, very much like 3 x 12V rails except the two, four or even six PCIe power connectors are split up across the additional two +12V rails.
o If the PSU supports 8-pin PCIe or has three PCIe power connectors on each of the +12V rails, it's not uncommon for their +12V rail to support a good deal more than just 20A.
o This is most common in 700W to 1000W power supplies, although for 800W and up power supplies it's not unusual to see +12V ratings greater than 20A per rail.
* 5 x 12V rails
o This is very much what one could call an EPS12V/ATX12V hybrid.
o Dual processors still each get their own rail, but so do the PCIe power connectors.
o This can typically be found in 850W to 1000W power supplies.
* 6 x 12V rails
o This is the mack daddy because it satisfies EPS12V specifications AND four or six PCIe power connectors without having to exceed 20A on any +12V rail
o Two +12V rails are dedicated to CPU cores just like an EPS12V power supply.
o 24-pin's +12V, SATA, Molex, etc. are split up across two more +12V rails.
o PCIe power connectors are split up across the last two +12V rails.
o This is typically only seen in 1000W and up power supplies.
Ok... What's the bottom line?
The bottom line is, for 99% of the folks out there single vs. multiple +12V rails is a NON ISSUE. It's something that has been hyped up by marketing folks on BOTH SIDES of the fence. Too often we see mis-prioritized requests for PSU advice: Asking "what single +12V rail PSU should I get" when the person isn't even running SLI! Unless you're running a plethora of Peltiers in your machine, it should be a non-issue assuming that the PSU has all of the connectors your machine requires and there are no need for "splitters" (see Example 1 in the previous bullet point).
The criteria for buying a PSU should be:
* Does the PSU provide enough power for my machine?
* Does the PSU have all of the connectors I require (6-pin for high end PCIe, two 6-pin, four 6-pin or even the newer 8-pin PCIe connector)?
* If using SLI or Crossfire, is the unit SLI or Crossfire certified (doesn't matter if a PSU is certified for one or the other as long as it has the correct connectors. If it passed certification for EITHER that means it's been real world tested with dual graphics cards in a worst case scenario).
Figure out if there are any variables that may affect the actual output capability of the PSU:
* What temperature is the PSU rated at? Room (25° to 30°C) or actual operating temperature (40°C to 50°C)
* If room temperature, what's the derating curve? As a PSU runs hotter, it's capability to put out power is diminished. If no de-rate can be found, assume that a PSU rated at room temperature may only be able to put out around 75% of it's rated capability once installed in a PC.
After that, narrow selection down with finer details that may be more important to others than it may be to you....
* Does the unit have power factor correction?
* Is the unit efficient?
* Is the unit quiet?
* Is the unit modular?
* Am I paying extra for bling?
* Do I want bling?
Hope this helps eliminate some of the questions that come up over and over again. I'll be editing this throughout the day and hope to be putting up more FAQ material later!
Power Factor Correction FAQWhat is the "wattage" number (i.e. 650W, 800W, etc.) actually telling me?
When you see the wattage rating of a power supply, you’re seeing the total maximum output capability of that particular power supply, but a computer has multiple voltage needs, and newer computers require more of the power supply’s capability to be on the +12V DC output rail. CPU’s and GPU’s regulate their power off of the +12V DC rail. Also, all of the computer’s motors run off of +12V DC: hard drive and optical drive motors, fan motors, pumps for water-cooling, etc. It wasn't too long ago that graphics cards did not require auxiliary +12V power and CPU's use to regulate their voltage from the +5V rail. An older power supply may have a lower percentage of it's power on the +12V than a more current unit.
What is the difference between "continuous" and "peak" ratings?
Some power supply units are rated for continuous output while others are rated at peak. "Continuous" means that the power supply is rated to run at it's maximum capability for no pre-determined period of time, while "Peak" indicates that the power supply will only run at the specified wattage for a brief period of time, possibly only a few seconds or up to a minute. This number is typically about 100W more than the power supply's actual continuous rating.
How does the temperature inside of my case affect the performance of my power supply?
Power supplies can perform differently depending on the temperature at which they are operating at. When a power supply is rated for it's total output wattage, it is rated to do so at a particular temperature. Anything beyond this temperature may take away from the power supply's capability. A power supply that is rated to put out 550W at 25°C or 30°C (room temperature) may only be able to put out 75% of that at 40°C or 50°C (actual operating temperature). This difference is called the "de-rating curve". A normal operating temperature for a power supply is 40°C.
Is the temperature at which MTBF is measured at an indication of what temperature the power supply's output rating is measured at?
Unfortunately, no. It's a tough race out there and there are a lot of guys rating their PSU's MTBF at room temperature, even if they rate their PSU at operating temperature. Fact of the matter is, MTBF can be a significantly, often exponentially, lower number when going from 25°C to 40°C. For example, one unit with an MTBF at 100,000 hours @ 25°C can have an MTBF of 20,000 hours at 40°C. That's a pretty big difference! So it's not unusual for a manufacturer to use the higher MTBF number at the lower temperature and, in most cases, not tell us at what temperature that MTBF is derived at. But even when they do tell us the MTBF temperature, this doesn't mean the PSU is rated at this. A PSU's output capability may not be seriously compromised by heat. If the PSU does 700W continuous @ 25°C and only 600W @ 40°C, the difference may not be significant enough to the manufacturer to increase their continuous wattage claims, so although they may measure MTBF at 25°C., they may very well be rating the PSU at 25, 40 or even 50°C. Unfortunately, it all comes down to marketing. It's easier to market a PSU that runs at what it's rated at 40°C then it is to market a significantly lower MTBF at the same temperature.
What is "power factor"?
Power factor, or “PF” for short, is the ratio of the real power to the apparent power.
Real power is the capacity of the circuit for performing work in a particular time and is measured in Watts.
Apparent power is the product of the voltage and current (V x A) of the circuit and is measured in volt-amperage (or “VA”.)
I know that it almost sounds as if Watts and VA are the same thing, and in DC they are (240W DC is equal to 240VA DC, for example) but because energy stored in the load of a device using alternating current (AC) is returned to the source, or due to a non-linear load that distorts the wave shape of the current drawn from the source, the apparent power can actually be greater than the real power. This would give you a power factor of less than 1. Power factor below .70 is generally considered poor power factor.
So what if I have a poor power factor?
Most consumers are charged per kWh (kilowatt hour) by their utility company. Fortunately, for the customer, poor power factor does not typically affect how much wattage your computer uses. But poor power factor does has an affect on how much power the utility companies can deliver. This means that the utility companies either have to increase their grid's capacity to compensate for the increased power load, charge per kVA instead of per kWh (some commercial/industrial accounts are charged per kVA while residential customers are still charged per kWh), charge a "power factor penalty charge" (which can be applied to customers with power factors even as high as .95!) or impose a martial law of sorts requiring all appliances sold in the country to have a power factor of .96 or better.
The European Union believed the latter of these to be the best solution, so as of January 1st of 2001 the EN61000-3-2 was put into place imposing limits on the harmonic currents drawn from the mains. In other words, if you're in the EU, you are REQUIRED to have a power supply with power factor correction. Power factor correction is not (yet) a requirement in the U.S.
Poor power factor can also limit how much current you can draw from a circuit. If you’re using a 20A breaker and are drawing a total of 15A in “real power” and the power factor is only .70, then you are drawing an apparent 21.4A, thus overloading the breaker.
Also, one of the requirements for a computer power supply to be considered "Energy Star" compliant is that it has a power factor of at least .90.
What is power factor correction?
Poor power factor can be corrected by adding some form of power factor correction to the AC input of the power supply. Power Factor Correction comes in two forms: Active Power Factor Correction, or APFC, and Passive Power Factor Correction.
Computer power supplies can create harmonics of the same frequency as the input current, due to the non-linear load caused by the bridge-rectifier doing the AC to DC conversion, and typically have poor power factor (typically 0.55 to 0.65).
Passive Power Factor Correction uses a filter that kills any harmonic current and passes current only at line frequency (typically 60Hz in the U.S.) The filters typically come in the form of large, high-value inductors.
Active Power Factor Correction is done by using a boost converter in between the bridge-rectifier and main input capacitors. The boost converter attempts to maintain a constant output voltage while drawing a current that is always in phase and at the same frequency as the line voltage.
Power factor correction won’t make your power supply more efficient (convert more DC output power with less AC input power), but can allow for more devices to be plugged into the same circuit. If you have a number of PCs on the same circuit, say in the event of a LAN party where a number of computers are plugged into a single power strip, it is easier to overload that circuit if a number of the PCs have poor power factor. Say for example you have a 20A breaker and there are five PCs plugged into the outlets on this breaker. Let’s assume the PCs are each drawing 115V at 3A from the wall, or 345W each, for a total of 1725W. This isn’t a lot of power and something the breaker should be able to handle without problem, but if the computers in question lack power factor correction, the “apparent” current draw could be as high as 27A (assuming a power factor of .55)! This will easily trip the breaker.
So I'm in the U.S. PFC is a non-issue, right?
Well... yes and no.
You may not NEED power factor correction, but "green" is more marketable and it costs a PSU factory less to make a bunch of the same platform, even if it includes PFC, then split production up between non-PFC and PFC designs.
These days, PFC is typically integrated into the design of the platform. So much so that the PSU manufacturers couldn't even really REMOVE the PFC in an effort to cut costs. That's ok because they're making up for it in larger quantities being able to sell their product around the globe. It also reduces returns because with active PFC it's impossible to plug the PSU into 230V while the switch is set to 115V (BOOM!)
If you're looking at a PSU over 550W continuous and it DOES NOT have PFC, be suspicious of it. It's likely an old, recycled platform that's inefficient and probably not suitable for the high +12V loads of today's computers. Or worse: It's really a 500W PSU labeled as a 700W!
