- Status
sephirotic
Distinguished
- Jan 29, 2009
- 67
- 0
- 18,630
How about realitively cheap PSU with two 12v lines? How do i actually differenciate where my 450w huntkey lines are going? I´ve always hd hard time to know if it was enough for my system. i´ve recently bought a Q6600 G0 and now iam scared to know how hard can i overclock it with an standard intel chipset with 2 hds, 2 ddr2 2.0volts memories and a radeon 4850. as far as only wattage is in account i could be fine doing 3.2 ghz (about 130~140w for the cpu) but taking amperage in account, im starting to ge scared that its not goig to be enough:
15A in 12V1 and 17A in 12V2.
15A in 12V1 and 17A in 12V2.
nukemaster
Titan
- Nov 23, 2004
- 19,626
- 6
- 74,965
The Idal wattage "Should"(Most efficient) be about 2x what you need at average loads(Idle for office systems and load for game systems).
Most power supplies reach maximum efficiency at about 50% (+-5%). This is due to how the systems operate. Transformers like those found in the PSU are rated without any cooling, add cooling and they can[not saying they all are but its an average] be pushed to 190%+ of there fanless rated load. Since they reach maximum efficiency at about 110% load(I know, this sounds impossible).
That said, That is just the best way to maximize efficiency. If that is not your concern just leave your self some extra power for future upgrades and to decrease excess stress on the psu.
If you have a passive PSU, the ideal wattage is full load, but you have to be sure it has a bit of air flow to avoid thermal shut off.
Most power supplies reach maximum efficiency at about 50% (+-5%). This is due to how the systems operate. Transformers like those found in the PSU are rated without any cooling, add cooling and they can[not saying they all are but its an average] be pushed to 190%+ of there fanless rated load. Since they reach maximum efficiency at about 110% load(I know, this sounds impossible).
That said, That is just the best way to maximize efficiency. If that is not your concern just leave your self some extra power for future upgrades and to decrease excess stress on the psu.
If you have a passive PSU, the ideal wattage is full load, but you have to be sure it has a bit of air flow to avoid thermal shut off.
Nukemaster is one of the few who has this correct.
PSU efficiency works on a bell curve, meaning they are most efficient at 50% load. So figure out your total draw under full load, and double that to get the best efficiency for a PSU in your configuration.
And contrary to what some others have posted, a PSU does not draw more voltage from the wall than what your system requires. EG: If your total system draw is 400w..., a 550w PSU, a 750w PSU, and a 1200w PSU will ALL draw 400w from the wall. The 750w PSU will do it most efficiently.
I'd like to see these comments edited for accuracy. It's understandable that some people will get things wrong from time to time, but it's a shame that misinformation can be posted so freely and confuse people. Without editing the content it just looks like a word fight between 2 people and no one knows for sure who is right and who is wrong.
PSU efficiency works on a bell curve, meaning they are most efficient at 50% load. So figure out your total draw under full load, and double that to get the best efficiency for a PSU in your configuration.
And contrary to what some others have posted, a PSU does not draw more voltage from the wall than what your system requires. EG: If your total system draw is 400w..., a 550w PSU, a 750w PSU, and a 1200w PSU will ALL draw 400w from the wall. The 750w PSU will do it most efficiently.
I'd like to see these comments edited for accuracy. It's understandable that some people will get things wrong from time to time, but it's a shame that misinformation can be posted so freely and confuse people. Without editing the content it just looks like a word fight between 2 people and no one knows for sure who is right and who is wrong.
nukemaster
Titan
- Nov 23, 2004
- 19,626
- 6
- 74,965
Well it would be hard to dispute this
The first of Corsairs own image, the rest are just random psu's pulled from spcr(Silent PC Review). The peak varies a bit, but in general most are good around 50%. it all comes down to how they wanted to push the transformers.
The first of Corsairs own image, the rest are just random psu's pulled from spcr(Silent PC Review). The peak varies a bit, but in general most are good around 50%. it all comes down to how they wanted to push the transformers.
THIS STUDY SUCKS
He never mentions that a power supply with 4-12Volt rails at 18Amps is better than 1 12V rail at 25A. Given a video card pulls around 150Watts(12.5amps) at 12V assuming all power is from 12V and negating the power on the bus. Which means a video card with its own rail is perfectly fine.
Plus it is also good to keep your devices on seperate rails.cd drives and hard drives - feed electrical noise back on the rails(you can see this with an oscilloscope), so if you want max over clocking potential keep your graphics cards on a different rail that is clean. Or hope that your graphics card company put awesome filters on the card which they probably didn't. You can also determine motor speeds of drives with this noise
Hey goof you ever heard of crest factor or %THD.
The goof ball who did this study forgot to add in screen captures of an oscilliscope to demonstrate how much ripple voltage increases with supply load. The biggest factor that separates the high quality supplies from the lowest. Ripple is reduced by using bigger caps with lower ESR right after the switches in the switching mode power supply. As you can imagine these caps are more expensive. Another way to reduce ripple is to play with switching frequencies but you can only overcome so much with this method, but better caps works the best.
This is how to define a cheap power supply from a good one.
Geez can't expect to much from keyboard racer studies these days.
I was surprised he stated efficiency goes up with power draw. Very well said young padwan. Yes switching loses becomes less relevant with higher pwm duty cycles.
If you want to measure true efficiency. Monitor the current into the supply on the 120V line. Measure the voltage of the 120V line. Then monitor each wires current draw and voltage. This will give you instantenous power efficiency since you know power in and out. The rest can be accountanted for as heat loses.
And yes I do not know how to spell or write cuz I a engineer. so forgive me for my grammer. Bash me all you want on grammar but i still just schooled you on power supplies.
He never mentions that a power supply with 4-12Volt rails at 18Amps is better than 1 12V rail at 25A. Given a video card pulls around 150Watts(12.5amps) at 12V assuming all power is from 12V and negating the power on the bus. Which means a video card with its own rail is perfectly fine.
Plus it is also good to keep your devices on seperate rails.cd drives and hard drives - feed electrical noise back on the rails(you can see this with an oscilloscope), so if you want max over clocking potential keep your graphics cards on a different rail that is clean. Or hope that your graphics card company put awesome filters on the card which they probably didn't. You can also determine motor speeds of drives with this noise
Hey goof you ever heard of crest factor or %THD.
The goof ball who did this study forgot to add in screen captures of an oscilliscope to demonstrate how much ripple voltage increases with supply load. The biggest factor that separates the high quality supplies from the lowest. Ripple is reduced by using bigger caps with lower ESR right after the switches in the switching mode power supply. As you can imagine these caps are more expensive. Another way to reduce ripple is to play with switching frequencies but you can only overcome so much with this method, but better caps works the best.
This is how to define a cheap power supply from a good one.
Geez can't expect to much from keyboard racer studies these days.
I was surprised he stated efficiency goes up with power draw. Very well said young padwan. Yes switching loses becomes less relevant with higher pwm duty cycles.
If you want to measure true efficiency. Monitor the current into the supply on the 120V line. Measure the voltage of the 120V line. Then monitor each wires current draw and voltage. This will give you instantenous power efficiency since you know power in and out. The rest can be accountanted for as heat loses.
And yes I do not know how to spell or write cuz I a engineer. so forgive me for my grammer. Bash me all you want on grammar but i still just schooled you on power supplies.
[citation][nom]demonhorde665[/nom]"It'd be nice to see a test of where the cards are taking their power from - how much from motherboard and how much from pci-e connectors. That would be useful information indeed for multi rail PSU users."any oen with half a wit , would drop messign with multi +12 volt rail PSU's. the better PSU's for gaming have always been single HEAVY duty +12 volt railed. PC power n' cooling, Seasonic, and Corsair , all have a wide variety of Great single +12 volt PSU models , ranging from a 450 (or 500) watt one with a +12 @ 33 amps up to their 850 watt models that offer a 60 (or 90) amp +12 volt , the benifit of using a single heavy duty +12 volt rail is in power efficeny , smaller split up rails are genrlaly weaker to odd power surges or the occasional power hic up that occurs with alternating current (the acutally ammout of power coming out of your wall flucuates constantly) weaker rails can be subject to failure or spikes more than heavyier rails. lets take a dual rail set up with 2 +12's at say 20 amps each , well say ther is a random (yet unnotacible) power flucuation in your wall outlet one of tehse railes dont get the power they need for jsut afew second and the rail drops for a 20 amp turn out down to a 10 amp turn out , and say you system needs at elast 34 amps to keep any thign from crashing ... well now your 12 volt seciton is only giving out 32 amps now .... CRASH ... blue screen (or worse) ok take a single heavy duty 60 amp +12 , since it charges more pwoer onto the rail than a smaler rail can do , the same fluctation would only cos teh rail to lose 5 amps instead of ten , even on abeefeier machine that reuqires 45 amps to stay stable , it would still be gettgin 55 amps durrign such a spike or drop and your system would enver even realise that ther was apower drop on teh +12 volt now reverse teh situation , and lest say you ahve a pwoer spike of 10 amps , on teh smaler rail such a spike could likely burn the rail out alltogehter *pop sizzle sizzle(* you jsut burned up one +12 volt rail if not both adn possibly fried your video card . However on the bigger 60 amp rail even on abig machienthat is puling 45 amps teh rail is only puling/out putting the 45 also beign made of more ruably ehavy wireing , the spike would ahrdly fry said rail only increasing the rails opull by 0 amps the extra amps would get dispersed as heat with no burn out *note this also depends on acutalal wattage fluctuation as well but teh same concept applies* bigger single rails are much more sturdy than split up smaller rails[/citation]
Odd power surges and hic ups what are you talking about. These things are choked like crazy on the input. Not only that you can put very noisy wall electricity in these and get the same output on the rails. You can also power these things with square waves. The power is almost always being supplied by the reactive sources. It also acts as a buffer for incase the input voltage/current to the switch drops low enough that at 100% duty the it can no longer make power for your "hic up" or phantom power issues.
Odd power surges and hic ups what are you talking about. These things are choked like crazy on the input. Not only that you can put very noisy wall electricity in these and get the same output on the rails. You can also power these things with square waves. The power is almost always being supplied by the reactive sources. It also acts as a buffer for incase the input voltage/current to the switch drops low enough that at 100% duty the it can no longer make power for your "hic up" or phantom power issues.
nukemaster
Titan
- Nov 23, 2004
- 19,626
- 6
- 74,965
Ummmmm almost ALL current power supplies, including those 4 x 18amp rail ones have ONE rail NOT 4. The 4 "Virtual Rails" are there to meet the ATX guidelines.
To protect against overheating the wires do to pulling too much power through them they said no more then 20AMPs on any given rail. Most psu makers go 18 to be safe. That said they ALL get there power from one main rail. The only thing that makes the rails is an over current unit on each one to prevent more then 18amps from passing.
It would be vastly inefficient to actually run 4 rails in the manner you speak of. PC Power and Cooling have decided to just ditch the multi rail thing and give on large rail. As long as the user does not try to pull 4 video cards on on molex connector there is no harm in doing this.
A general image of how most multirail psu's work. One large rail and the virtual rails are just broken up by way of an overload to prevent more then a preset amount of power through.
To protect against overheating the wires do to pulling too much power through them they said no more then 20AMPs on any given rail. Most psu makers go 18 to be safe. That said they ALL get there power from one main rail. The only thing that makes the rails is an over current unit on each one to prevent more then 18amps from passing.
It would be vastly inefficient to actually run 4 rails in the manner you speak of. PC Power and Cooling have decided to just ditch the multi rail thing and give on large rail. As long as the user does not try to pull 4 video cards on on molex connector there is no harm in doing this.
A general image of how most multirail psu's work. One large rail and the virtual rails are just broken up by way of an overload to prevent more then a preset amount of power through.
aganainotoki
Distinguished
- Dec 6, 2008
- 3
- 0
- 18,510
Well, all power supplies sold in the US are calculated with "Max power", not actual power. The problem is that different manufacturers have different ways to calculate their products. Many brand name supplies have different current strength even though they are marked with same max power. For example: Rosewill RD500-2SB +12V is 31A at about 350W, while ENERMAX ELT500AWT 500W has 46A at about 400W. Both of them are ATX 12V 2.2. So it's better to look the nameplate carefully when choosing a power supply.
nukemaster
Titan
- Nov 23, 2004
- 19,626
- 6
- 74,965
They are both 500. One just puts it all on the unused 3.3 and 5 volt rails.
I fully agree this is BS, but there is no law against it since it is not a lie. it WILL put out 500 watts just more of that 500 is dedicated to the 3.3 and 5 volt lines that no one needs so much power on.
I fully agree this is BS, but there is no law against it since it is not a lie. it WILL put out 500 watts just more of that 500 is dedicated to the 3.3 and 5 volt lines that no one needs so much power on.
pangloss44
Distinguished
- Feb 3, 2009
- 2
- 0
- 18,510
unbelievable, thank you so much for this valuable resource. can't believe how much detail is on here.
Guys and Mr. Tino Kreiss, please help me with your professional advise. I was going to replace my GeForce 8800 GT (512 MB)with Radeon HD 4870 (512 MB). I currently have PSU Huntkey LW-6500H that has dual rails +12V1 (with 15A, +12V2 (with 18A) output, which in total will have 33A for max load. Do you think it would be enough for Radeon HD 4870? Or should I replace my PSU before replacing video cards?
Thank you beforehand.
Thank you beforehand.
Onus
Titan
- Jan 27, 2006
- 37,037
- 12
- 89,965
The last reviews of Huntkey PSUs I saw had them crapping out well below their rated load. I think that might have been at hardwaresecrets.com, but it may have been hardocp.com. I don't think it was jonnyguru.
Also, on most PSUs, the rails are not additive. Some Antecs are an exception to that, but typically the total amount of power available is somewhat less.
Also, on most PSUs, the rails are not additive. Some Antecs are an exception to that, but typically the total amount of power available is somewhat less.
wingmaster
Distinguished
- Sep 1, 2006
- 66
- 0
- 18,630
I think theres a mistake on the 8800GT and 9600GT showing the ones with 1024 memory using less wattage that the ones with 512. Maybe I'm wrong here but just guessing the ones with 2 times more memory would use more power.
Excellent article, really takes the guesswork out of this stuff.
Question. What about when a PSU has two 12V channels?
My PSU has a 12V1 rated 22A and a 12V2 rated 24A. Looking at the connector descriptions in the manual, the video power connectors are 12V2. Now, according to your article, my video solution needs about 22A, so I'm in good shape. What I was wondering though was, if there is unused amperage on 12V1, can my video cards draw on it also, through the PCI Express bus? That would mean -- if my other components have amps left over -- I could go to an even more power hungry video setup without changing the PSU.
Question. What about when a PSU has two 12V channels?
My PSU has a 12V1 rated 22A and a 12V2 rated 24A. Looking at the connector descriptions in the manual, the video power connectors are 12V2. Now, according to your article, my video solution needs about 22A, so I'm in good shape. What I was wondering though was, if there is unused amperage on 12V1, can my video cards draw on it also, through the PCI Express bus? That would mean -- if my other components have amps left over -- I could go to an even more power hungry video setup without changing the PSU.
thank you for this article it is a good read and informative. i would like more info about multi rail psu and how the total power is split up between the different voltages? i have two different 500w psu both of which are dual rails but one has higher 3.3v rating and the other has a higher 12v rating. is the 12v the most important? the article seems to focus on 12v. i have read through the comments and am curious if the multi rail thing is just to satisfy UL standards for the wire used. they can't have a 30 amp load on a single 18ag wire so the one rail is split in two.
also, one of the psu has a rail dedicated to the 4 pin mb connection and the other shares that rail with the sata connectors. is there an advantage to this? does the video card get some of it's power from the 4 pin mb connection?
also, one of the psu has a rail dedicated to the 4 pin mb connection and the other shares that rail with the sata connectors. is there an advantage to this? does the video card get some of it's power from the 4 pin mb connection?
aganainotoki
Distinguished
- Dec 6, 2008
- 3
- 0
- 18,510
Seems like PSU's are headed for a problem.
If components can ratchet down their wattage when not under load -- like the newer 260/280's, then your PSU needs a high maximum to handle a high-load draw, but when your system is idle -- it seems likely that the PSU might fall into the 50% usage range. This means the PSU isn't going to be operating in its 'comfort zone' for efficiency, no?
This especially gets true for planned Intel CPU's that will be able to turn cores off or on depending on system load. I can easily see that 50% idle/load dropping to 33%idle/load -- meaning PSU's will need to be efficient over a wide range of outputs but especially down to 50% or 30% of max if newer 'green' tech's are implemented throughout a computer.
Are there any PSU's with that type of usage?
Also, the draw of memory can't be ignored -- Dell's workstation machines have memory expandable to 64G -- and they suggest going with a 1kW PSU vs. a 750W if you want to go for more than 32G. That would suggest they are allocating 250 extra watts for 32G of memory? Is that overkill, or do they just have incredibly inefficient memory?
If components can ratchet down their wattage when not under load -- like the newer 260/280's, then your PSU needs a high maximum to handle a high-load draw, but when your system is idle -- it seems likely that the PSU might fall into the 50% usage range. This means the PSU isn't going to be operating in its 'comfort zone' for efficiency, no?
This especially gets true for planned Intel CPU's that will be able to turn cores off or on depending on system load. I can easily see that 50% idle/load dropping to 33%idle/load -- meaning PSU's will need to be efficient over a wide range of outputs but especially down to 50% or 30% of max if newer 'green' tech's are implemented throughout a computer.
Are there any PSU's with that type of usage?
Also, the draw of memory can't be ignored -- Dell's workstation machines have memory expandable to 64G -- and they suggest going with a 1kW PSU vs. a 750W if you want to go for more than 32G. That would suggest they are allocating 250 extra watts for 32G of memory? Is that overkill, or do they just have incredibly inefficient memory?
Onus
Titan
- Jan 27, 2006
- 37,037
- 12
- 89,965
80+% certified PSUs must be at least 80% efficient on as low as 20% load. I think 50% and 80% are the other load values that must also be at least 80% efficient; not sure about 100%, but you wouldn't want to run a PSU at 100% anyway, at least not for long. An exception might be the Antec Signatures, which can support over 100% of their labeled rating.
Tino, I'm don't understand how you get to the conclusion that the optimal operating range for PSU's is 65-85%?
Looking at the performance graphs on the test reports for various PSU's at http://www.80plus.org/manu/psu/psu_join.aspx, most of the one's I've looked at have a peak efficiency at ~50% load.
Are you suggesting that as a peak in order to balance against idle usage where efficiency also drops?
I guess to get the most apropriate unit for each user you would need to chart their power usage over time and calculate an average W rating then buy a PSU that operates at 50% efficiency at that wattage...
Are there any cheap devices or software that can do that?
Looking at the performance graphs on the test reports for various PSU's at http://www.80plus.org/manu/psu/psu_join.aspx, most of the one's I've looked at have a peak efficiency at ~50% load.
Are you suggesting that as a peak in order to balance against idle usage where efficiency also drops?
I guess to get the most apropriate unit for each user you would need to chart their power usage over time and calculate an average W rating then buy a PSU that operates at 50% efficiency at that wattage...
Are there any cheap devices or software that can do that?
- Status
TRENDING THREADS
-
Discussion ABS Eurus Aqua gaming pc with Gigabyte z790 ud ac crashing and BSOD
- Started by Icarry2
- Replies: 3
-
-
-
-
News Windows 11 will reportedly display a watermark if your PC does not support AI requirements- Started by Admin
- Replies: 31
Facebook
Twitter
Instagram