The Power Supply Thread

[KamiXkaze]

Thanks for the information

KxK

 

[TomMe]

Originally posted by: jonnyGURU
I'm not sure 18A per 12V rail is enough for a pair of 7800GTX's. I'd like to see 20A on one of those rails. Like an 18A/20A or a 20A/20A. Something with 420W or 444W on the combined 12V rails.

I'd like to quote you on this one, because again I'm a bit confused. I get the idea you're saying that the amperages the manufacturers give to the 12V rails are real life physical limits, or am I misinterpreting?

On the SilentPCReview website and forum they say that each 12V rail should be able to draw 20A and that sometimes there's even only one 12V rail although PSUs are labeled as having two.

 

[jonnyGURU]

Again TomMe.... No offense, but you're letting your selective reading get in the way of obtaining factual information. And I'm not saying SPCR's verbiage is not factual... more like it's stated in a "dumbed down" way that isn't clear to you....

Howard asked very much the same question above.... I'll quote it again down here.....

Originally posted by: jonnyGURU

Originally posted by: Howard

If there is no limiter, the PSU could still have two separate rails

I thought all rails drew power from the same inverter/transformer? In that case, what differentiates the different rails?

Even though the power in a power supply all goes through the same rectifiers, transformers, etc (which is why your PSU's total output is never the sum of all of it's rails.) final regulation is still done by an invidual zener diode and power transistor. These are what we call your actual "rails." Not the DC power feeding these circuits. So if there is more than one regulator regulating 12V DC output, it's still more than one regulator, more than one "rail." And if there is no limiter on this circuit, or circuits, that doesn't mean that the regulation doesn't still take place. If you have two 12V rails and they do not have limiters set on them, that doesn't mean that what's hooked up to the regulator 12V1 is not isolated from 12V2 unless the output of these regulators are put into parallel. Of course, the rails are going to be interactive with each other, but that's the case of any dual-12V-rail power supply simply because the source of the voltage being regulated is the same. One will always rob Peter to pay Paul. This is why you can't add multiple 12V rails to find the total combined 12Voutput of a multi-12V-rail power supply. If you only have 444W available to BOTH of the 12V rails, it really doesn't matter if there are limiters restricting only 20A to one rail and 20A to another or no limiter on either, as long as the total output doesn't exceed 444W. But the actual regulation of each 12V rail, down to 12V, and the inherent filtering and isolation of that rail due to the components used still remains.

So to reiterate what SPCR is saying, technically there is only ONE 12V rail where the power is regulated to a single source of 12V. This is why your combined 12V rail is actually MORE IMPORTANT than the value of each individual rail. This rail is then spit into two rails that each have their own regulation and, 99% of the time, their own limiters.

Again, I refer to the ATX12V spec at formfactors.org. According to the specificaion, each rail should have a limiter in place that allows no more than 240VA. That's why you typically see no more than 20A on a label. But in reality, these limiters may actually be set to 22A, 25A or more because quite frankly 20A is not enough... REGARDLESS of what's actually on the label.

But going back to where you quoted me... what is most important is what is reported as the capability of the combined 12V rails. You can now go back and read what I said in my third post in this thread and understand it better....

Originally posted by: jonnyGURU

?Now look at the 12V rails. There?s two 12V rails (I?ll explain why later) and they are rated at 18A each. The maximum combined wattage of those two rails is 32A.

What?!? But 18A plus 18A is 36A? Like the +3.3V and +5V rails I just mentioned, +12V rails are not additive. You can load each one up to 18A, but you can?t load them both up to 18A.

It's not unusual to see a power supply with 18A/18A actually have more capable combined 12V rails than one with 20A/20A. The one with 20A/20A may just have the limiters set higher on the 12V rails, but the regulation of the voltage may not be as tight at 20A as the other PSU would be at 18A.

 

[TomMe]

I read the things you quoted before I asked the question. But I admit that I don't fully understand everything that's being said here, my knowledge on the subject is very limited and I'm doing my best to fit the pieces of the puzzle.

So you are saying that it's impossible to know what the real limits are without testing them and that you can assume they are a bit higher than on the label?

What about 2 PSUs with the following 12V specs:
+12V1@14A, +12V2@15A (combined 348W)
+12V1@17A, +12V2@16A (combined 396W)

What I can see here is that the total output is the sum of the limits. Can I assume these limits are a bit higher than the given specs? Or is it just a coincidence that they add up here to be the total output? (these are specs from the Seasonic S12 430 and 500 btw)

Apologies if I asked something obvious again.

 

[jonnyGURU]

It is very rare that the rails are additive. Those Seasonic's you quote are an exception to the rule. Is it possible that the limiters aren't actually set to 17A and 16A? Absolutely. I bet they're set to at least 20A each. AT LEAST. But one of the things that makes specs specs is not just a limiter that prevents the power supply from providing power over a given number, but more so "quality control" parameters that a power supply company gives themselves to maintain a certain quality.

Let's say "Brand S" gives their 12V rails a tolerance of +/- 1% and a .01V@1ms fluctuation at 50C. At 16A, the 12V passes, but at 17A, the rail droops by .02V and there's a fluctuation of .03V@1ms. This is a high standard, but certainly some companies hold themselves to a higher standard than others. This is a "quality" power supply.

Now let's look at a not-so-high-quality power supply: "brand A." They give their 12V rails a tolerance of +/- 3% and allow for a .05V@1ms fluctuation at 20C. Not only does "brand A" pass at 16A, but also 18A.

So "Brand S" has 16A on the label for the 12V rail and "brand A" lists 18A. But now knowing how they test this, clearly "Brand A" is NOT better. In fact, put "Brand A" in the 50C environment that "Brand S" uses and hold it to the "Brand S" standards and it may not even pass at 14A!!! This is where simple "brand trust" comes into play, because the actual test environment and parameters are not always disclosed. The PCP&C Turbo-Cool ads illustrate this so well.

It's tough on the consumer.

I know for a fact that "Brand U" is working on a new 600W with more juice on it's 12V rails because ATI and nVidia say that the 420W on the combined 12's may not be enough for quad-GPU or X1900 Crossfire builds. Yet "Website M" is kicking "Brand T" to the curb because the current 600W version of "Brand U" schools "Brand T" on their X1900XTX crossfire and 7900GTX SLI bench builds, even though "Brand T" claims to have 480W on the combined 12's!

So you really need to do your homework. And I don't mean forums where people bitch about their wife's cousin's neighbor's problem with a rebate type of problems, but real builds put in real life situations for extended periods of time.

 

[TomMe]

Thanks again for your elaborate explanation! I'm starting to see things a bit more clearly now. :thumbsup:

 

[KoolDrew]

You may want to mention where to find the UL number. Just so other people know, it should be on the label below the backwards "UR" symbol.

 

[surv1v0r]

"There are several calculators on-line that allow you to "add up" your computer's power" - could you name some of the (more reliable) calculators please ? I realise that they may not be entirely reliable, but trying to find the information from Gigabyte, MSI, AMD, Maxtor, Crucial, etc. websites is incredibly difficult.

Many thanks

 

[Howard]

Originally posted by: surv1v0r
"There are several calculators on-line that allow you to "add up" your computer's power" - could you name some of the (more reliable) calculators please ? I realise that they may not be entirely reliable, but trying to find the information from Gigabyte, MSI, AMD, Maxtor, Crucial, etc. websites is incredibly difficult.

Many thanks

sig

 

[EpsilonLyrae]

Last summer, a client of mine got a case with a stock power supply, and the power supply was missing the -5 volt wire on the 20-pin main mobo connector.

Without the -5 volt, the Abit EQ motherboard monitor was sounding a continuous alarm, which of course got them nervous. And, the thing didn't seem to be Prime 95 stable. So I talked them into springing for an Enermax 465 power supply like mine. It has all twenty pins, that power supply does, including the -5 volt.

The Mobile Athlon XP on their motherboard wasn't able to reach 2200Mhz stably either without the -5 volt or due to the stock nature of the power supply. I didn't know whether it was a bad power supply, or whether the problem was due to the missing -5 volt. I had no idea that the -5 volt wire was disappearing, and I thought at the time that we had just simply received a faulty power supply with the case. The Enermax replacement worked just fine.

--
--

Anyway, yesterday, when I installed my new Enermax 485 Power supply into my own Athlon XP machine with a Mobile XP processor running at 2200Mhz, everything looked fine. But just recently when I loaded up Abit EQ, I get the warning siren going off, and then I see that I'm missing the -5 volt as well.

Subsequently, looking at pictures on NewEgg, all of the new 24 pin power supplies are missing the -5 volt from what I can tell. They don't list it and they don't have it, and there is a hole there, 19 pins in the 20 pin section. Now, I'm starting to see that all new power supplies come without the -5 volt. I take it that the Athlon 64 Mobo's don't require the -5 volt supply. And, then I even read in this thread up above where the new PSU specs exclude the -5 volt wire.

A couple of the remaining no-name brand PSU's with the old 20 pin connector seem to have a -5 volt, but just as many have no -5 volt connector. I think I found a 20+4 splittable PSU that still has the -5 volt, the cheapo Thermaltake 430 that is popular. But, most 20+4 splittables are really 19+4 if you look at the pictures or read and see what voltages they supply to the main connector. I don't think any of the true 24 pin mobo connectors have a -5 volt pin or wire at all, they are all really 23 pin connectors.

My machine seems to be running fine right now without any -5 volt, once I figured out how to turn off the Alarm in Abit EQ. I'm typing this right now without a -5 volt.

I can't find anything that tells me if it is fine to just go ahead without -5 volts with an Athlon XP and just take it forward business as usual. I wanted a PSU that was ready for the Athlon 64 when I'm ready to go there. Apparently -5 volt isn't needed or wanted for the new Athlon 64 stuff, but my old ABIT NF7-S motherboard monitor reports it missing when it is missing.

I can't find anything in writing that explains the trend of removing the -5 volt and explains whether it is safe for an Athlon XP mobo to be run with no -5 volt supply for a long period of time.

Anyway, if it is safe to go forward without a -5, this current power supply unit is actually set up and supposedly ready for Athlon 64 and PCI Express.

Anyway, you can probably see why I'm a bit nervous.

I ran Prime 95 for an hour or so, and this new Enermax 385 is Prime 95 Stable with my NF7-S V2 Abit Motherboard and the Mobile Athlon XP that I have. Nothing indicates that there is anything wrong with running without the -5 volt wire. And, I tried a USB device, and it worked fine.

But, today, when I stuck in a new Sound Blaster Live 24-bit card, there was a bunch of static hiss in the background, and a hiss sound when using slider bars on IE, and the game I tested sounded crappy with a static hiss going on in the background the whole time. Then it makes me think that the SoundBlaster live 24-bit card needs the -5 volt supply from the mobo that it isn't getting from the new Enermax 385 PSU.

Anyway, I'm Prime 95 stable, and my Geforce 6800GS seems to be just fine without the -5 volt supply. But I'm not sure whether the nforce 2 motherboard will deteriorate over time without the -5 volt supply, and I'm not sure whether I got a lemon SB live on my hands or whether the -5 volt supply is needed in order for my mobo to power the SB Live. But, the SB Live seems to do well enough on the desktop sounds, and only mucks up in games and on background sounds, like selecting, sliding, and rolling the central roller on the mouse to scroll the screen. Then the SB Live goes yuck when I scroll but is fine otherwise with Windows sounds, so I don't know what a missing -5 would be doing to selectively give bad stuff in games while at other times causing the SB Live to work properly.

Does some guru out there know if an nforce 2 Athlon XP based mobo can work just fine without a -5? What was the -5 volt supply supposed to do? What's going to get messed up without a -5 volt supply being fed into my motherboard?

I hope I don't get anyone angry at me, but this seemed like the best place to post my questions about the missing -5 volt pin on my new Enermax Power Supply and its possible effect on my Athlon XP nforce 2 system.

 

[JEDIYoda]

from what I know about that the -5 volt is going bye bye....

 

[jonnyGURU]

Originally posted by: EpsilonLyrae

I can't find anything that tells me if it is fine to just go ahead without -5 volts with an Athlon XP and just take it forward business as usual. I wanted a PSU that was ready for the Athlon 64 when I'm ready to go there. Apparently -5 volt isn't needed or wanted for the new Athlon 64 stuff, but my old ABIT NF7-S motherboard monitor reports it missing when it is missing.

I can't find anything in writing that explains the trend of removing the -5 volt and explains whether it is safe for an Athlon XP mobo to be run with no -5 volt supply for a long period of time.

Another candidate for reading the ATX12V specifications on formfactors.org.

The -5V rail was removed from ATX12V spec because most motherboards do not use it. The only one I can think of that does is some MSI boards that need it for the onboard sound.

Your friends ability to not overclock Mobile XP has nothing to do with a lack of -5V rail.

 

[foodfightr]

Cliffs: OCZ PowerStream 600W 4TW

Just playin... but seriously with my overclocked volt hungry machine it doesn't even budge when I go to full 100% dual core load from idle.... measured with a digital multimeter

 

[GalvanizedYankee]

Originally posted by: foodfightr
Cliffs: OCZ PowerStream 600W 4TW

Just playin... but seriously with my overclocked volt hungry machine it doesn't even budge when I go to full 100% dual core load from idle.... measured with a digital multimeter

I have some very strong opinions about PSUs.

I also favor some brands over others. I have a couple high end PSUs that I would love to rave about.....But not in this thread..OK!!

I wish the silly fanboies that have posted in this stickied thread would edit out the WHORAH crap so this thread can remain stickied for a long while.

This thread is about information, NOT promotion of brand loyalty.

This thread would be half as long if the children would do this type of posting in the general PSU threads, there are many, so it can be done often


...Galvanized

 

[JEDIYoda]

maybe there is a way we could get a moderator to go through and edit out anything that has to do with brand loyaltie.

It is a real shame that this thread can get off topic when its purly an general informational thread that has nothing to do with brand loyalties...

 

[Havlock]

EDIT: Never mind. Wrong thread.

 

[GalvanizedYankee]

A request, asked for myself as well as others.
There as been a bit of low key worry posted about PSU acoustics. I'm not talking about fans but coils and caps.
Since so many are striving for quiet yet powerful systems, this means that the acoustic qualities of other system componets will stand out because cooling fans aren't masking them anymore.

If you would, please explain why so much silicone is used on coils and caps.
Try to define the acoustics of caps and coils. I doubt you have sound files. A long Google advanced search pulled up nothing. It would be nice to hear. *shrugs*

Hitachi has sound files of HDD problems but no PSU provider does, normal or otherwise.

From my hour long homework romp via Google it was learned that Good audio PSUs have a very low dBA spec for their PSUs and that the spec for desktop units is more demanding than workstation or server PSUs. Once again my refferance is not to fans but the noise emmision of electronic componets.

TY for your consideration of the questions posed.

...Galvanized

 

[Showtime]

A little off topic.... is it worth getting a power cleaner for my system and which one? I run a few different APC surge protectors, but I saw some cleaners at compusa and wondered if that would do anything more than surge protectors.

 

[GalvanizedYankee]

Originally posted by: Showtime
A little off topic.... is it worth getting a power cleaner for my system and which one? I run a few different APC surge protectors, but I saw some cleaners at compusa and wondered if that would do anything more than surge protectors.

Start a thread in GH on that topic. This STICKIED thread is about PSU operation.

A surge protector that offers AVR is a plus. A UPS with AVR is better. That has been addressed already in this long thread. Read the thread. MMmmKay

...Galvanized

 

[AJeightFive]

Sorry if this has been answered somewhere in these 9 pages...

Should I be concerned if my hardware monitor shows an undervoltage of roughly 0.05V on the the 3.3V and +5.0V rails and ~0.22V on the +12V rail? Vcore is under by 0.04V. Are these within normal tollerances or should I look for a beefier PSU?

 

[chr6]

i will never stray from my pc p&c psu. <3

 

[Howard]

Originally posted by: AJeightFive
Sorry if this has been answered somewhere in these 9 pages...

Should I be concerned if my hardware monitor shows an undervoltage of roughly 0.05V on the the 3.3V and +5.0V rails and ~0.22V on the +12V rail? Vcore is under by 0.04V. Are these within normal tollerances or should I look for a beefier PSU?

You're well within tolerance.

 

[Hari Seldon]

This is my first post.
After a long search, I came across this thread in this forum (a site I have had bookmarked for a year because I used to subscribe to CPU, but I had rarely returned to it), and I was initially convinced that this thread was going to answer all my questions.
Although it has been a great source of knowledge, it has failed to answer the questions that led me to this thread in the first place.
But since I am substantially more aware of what the heck is going on with the PSU now, I must say thank you for putting your original pieces together and publishing them.

I bought my computer a year ago, built it myself (the most fun I've had since I got my first bicycle) and made a few mistakes. The Antec Sonata came with a TruePower 380 watt PSU, and although everything seemed at the time to be cutting edge, I can see now how the PSU manufacturers fell behind (and are still behind I worry) what the power community requires.
I figured I'd look first at what Antec has got. I'm not loyal, not promoting them and not adverse to hearing criticism of their products if it's fair. I will look at competitor's products after I fully understand the whole PSU subject.
I found three models that seemed to be adequate: the TruePower II 550, the Phantom500 and the NeoHD 550.

My concern, which this thread has not fully answered is: How many amps is enough on each of all the rails? All the rails means the +12V, the +5V, the 3.3V, (and now that I'm writing this I must add) the +5Vsb and the -12V rails.

What prompted me to want to know what I should have as a minimum for ALL the rails is that I noticed something stiking about the three models I have referred to. Pay special attention to the +3.3V and the +5V rails please.

Model ------- +3.3V ---- +5V ----- +12V1 --- +12V2 --- +12V3 --- +5Vsb ---- 12V

TPII 550 ------ 32A ------ 40A ----- 19A ------ 19A ------- n/a ------ 2.0A ----- 1.0A
Phantom ------ 30A ------ 30A ----- 17A ------ 18A ------- n/a ------ 2.0A ----- 0.5A
NeoHD 550 --- 24A ------ 20A ----- 18A ------ 18A ------- 18A ----- 2.5A ----- 0.8A

Sorry about the graph. I didn't know how to make one of these things in a forum post.

Notice the significant drop on the +3.3V and +5V rails and how the +3.3V and the +5V ratings on the NeoHD 550 (which is the one that has 42 amps available on the combined +12V rails) have dropped all the way down to 24A and 20A.
I'm worried when I see that huge drop and I wanted to make sure that in getting a more efficient and versatile PSU than the TruePower 380 that I have now, that I don't get a PSU without enough juice on all the other rails besides the all-important +12V rail which you have completely covered.

Notice also the smaller variations on the rarely-mentioned +5Vsb and the -12V rails. I'm not too concerned about these, but it would be nice to have that knowledge.

Later on, I?ll give you a little formula I use to calculate how much power I think I need on the 12V rail (the most important rail in a modern PC.) 3.3V and 5V just follow.

What do you mean they "just follow"? I don't know.

Given that most power supplies give you 20 to 30A on the 3.3V (which is way high by today's standards, but even 30A on the 3.3V is only 100W)...

What do you mean by "today's standards"? I don't know what they are.

I?ve seen power supplies with the PCI express connectors on 12V2 and even some with one PCI-e connector on each of the two 12V rails. THESE are the kind of dual rail power supplies you need to look for.

Without favoring any particular brand, can you provide a reference list so that those of us who are interested can compare those that are on the list, and in the future when the list is outdated, it can still be referred to as a part of this information? It would be helpful now and in the future to see some examples.

Calculating how much juice you need on the 12V rail:

Can you do the exact same thing for all of the rails? You see, you were quite right in your belief that many of us, like me, were going out and buying a PSU based upon the total wattage, without knowing much else. But then you didn't follow completely thru with this conviction of yours. In other words, I really did not know what components are using the +12v rail, how many amps they require and how they need to be managed.
After reading this thread, I know alot more than I did before, but I still don't know anything about all the other rails. You have only covered the so-called most important rail, but what about all the other rails? What components use them? I don't know. What minimums should they have and why?

Good job. This is the best thread of information I have read in the year I've had my computer. Please complete this in-depth study you began.

 

[Bobthelost]

The reason there is no calculation for the other rails is that they are almost never the point of failure. To see just how irrelevant they have become, and how little modern PCs use them read this :http://www.silentpcreview.com/article265-page1.html

 

[Hari Seldon]

Yes a very interesting read, especially this point:

1) It seems to be highly unlikely that a modern system will ever overload either the +5V or +3.3V lines of a ATX12V 2.x compliant power supply. In our systems, neither of these lines ever drew more than 5A under any circumstance, and many power supplies rate them above 20A. The power draw on these lines tended to be quite stable and did not fluctuate much with load.

I guess I wasted alot of time and space with my questions. Still, it would have been nice to know everything there is to know about all the other rails too. Even if the major drop in the ratings (as I pointed out in my post above) for the +3.3V and +5V rails does not matter. It's just some missing pieces. However I agree with you that they have become irrelevant as far as something to worry about.