PSU's---2 x 12v---The Dual Rail Myth

F

FastEddie

Guest
The Dual Rail Myth

In looking at the newest ATX standard, the ATX12VPower Supply Design Guide, ver 2.01, we find two major changes from the past in that 2 x 12v rails are now advocated, rather than a single 12v rail, in addition to a 24-pin main ATX power connector interface.  How does this dynamically impacts us AMD users?  First, let?s look at some of the changes made from the ATX 1.3 standard and see how they impact us, as power users.

The most important change is the incorporation of 2 x 12v rails, or ?Dual Rail,? technology, where the processor now has basically its very own power interface that is separated from all other power consuming devices on the motherboard.  A Quote from Section 1.2.1, entitled ?Increased +12VDC Output Capability? of the Guide:

?System components that use 12V are continuing to increase in power. In cases where
expected current requirements is greater than 18A a second 12 V rail should be made
available. ATX12V power supplies should be designed to accommodate these increased
+12 VDC currents.?

This may sound like a good thing, and I?m sure that one day it will be a major improvement over past power supply interfaces, however, as an AMD user, the major drawback is that this newest standard (being designed around, and in support of, Intel based technology), took no consideration of AMD/Sli.  ?Our? problem isn?t that AUX ATX 12v Rail-1 is isolated with anywhere from 15amps to 18amps of power for the processor alone, but that the remainder of the entire system must depend upon ATX 12v Rail-2, with it?s 15amps to 18amps, to power all the other devices on the motherboard.  This is a lopsided balance of power in regulating and controlling devices in the average computer system, and becomes more than simple balancing act when this ?design? becomes the limiting factor for system stability in our high powered enthusiast systems.

What?s added to ATX 12v Rail-2 that makes system stability an ?issue? is related to the plethora of new technologies that must all be powered by this single rail.  Aside from SATA/SATA II, standard PCI, in addition to all ?onboard? peripheral interface devices such as multiple gig-lan network connections, multiple (up to ten) USB 2.0 connections, IEEE 1394 interfaces, standard serial/parallel port interfaces, legacy IDE interfaces and audio functions, the ?Guide? takes into consideration a new power hungry interface, being PCI-E, that consumes anywhere from 40watts to 120 watts of power in a single card.  The cpu no longer has the distinction of being the highest power consumer in a computer system, when one considers the amount of power a 6800 Ultra needs to operate properly.  This is a vast amount of power to take into consideration when building a new computer system, and one in which is compounded two fold with the release of nForce 4 SLI, where two such power hungry graphic adapters can coexist in the same system.

Many of you have followed the Dual Rail ?vs- Single Rail argument as it?s been played out in the implementation of NF4 Sli, where I have maintained that dual rail psu?s, as they exist today, are not an efficient use of the power resources these power supplies provide.  Where AUX ATX 12v Rail-1 (for the processor) would run at approximately 35%-50% of it?s available output 60%-70% of the time,  ATX 12v Rail-2 would run at or near 100% of its available output 100% of the time.  Power Supplies are not designed to work under these conditions and it seems that the ?Guide? agrees and has taken this into consideration, where at Sec. 1.2.2. Minimum Efficiency, it points out:

?Minimum measured efficiency is required to be 70% at full and typical (~50%) load and
60% at light (~20%) load. New recommended guidance has been added to provide
direction for future requirements.?

What this tells us is that, indeed, ATX 12v Rail-2 shouldn?t be required to run at more than 70% MAX of it?s available output on a constant ?full and typical? basis.  And it means that Dual Rail psu?s are not the best choice for powering nForce 4 Sli as these power supplies weren?t designed to work in such a lopsided environment.  It is much more efficient to use a high output Single Rail psu with at lease 26a on that 12v rail, where 28a for the Sli dual 6800 Ultra?s is a much more stable environment to run these top of the line cards.  And for those that say, ?well these boards were designed around the ATX 2.01 standard, so they should work just fine,? my answer is that they weren?t ?designed? around NF4 Sli by nature of the limitations the ?Guide? itself places on that second 12v rail.  Another response regards accumulation of the two 12v rails, where 18a & 18a = 36a and plenty to power any motherboard.  My response is this is a total misconception as these power supplies have no ability what-so-ever to accumulate their output current from the two totally separate rails on the fly.  Sure, there are one or two psu?s that do have the ability to combine their output, via a switch, but that ?switch? turns them into essentially a more powerful single rail unit, with less amperage than the total of both 12v rails. Until a psu is designed for Sli, don?t be drawn into the marketing hype about ?Sli Ready? and ?Sli Capable? as their use goes contrary to the very guide that implements them.  You would be much better served with a single rail power supply that can ?efficiently? power that Sli gamers box.  :cool:


 
Ah, just saw on their site that they still sell power cables .... http://www.qtec.info/products/product.htm?artnr=13201
At least that's something ....
And do look bottom right ... it's even HQ -> High Quality!!!
 
New Silverstone 54a single rail.

http://www.phoronix.com/scan.php?page=article&item=644&num=1

http://www.jonnyguru.com/review_details.php?id=100

 
I looked into this. My old man is an electrician. Dual Rails are instantly in-efficient compared to Single Rails. Due to the fact of standard electrics.

If you output two splits from the same supply there is DOUBLE the resistance. Thus obviously proving that Dual Rails can't supply the combined amounts of rail A + rail B. Standard electrics an Intel overlook it. When really it's blatantly obvious.
 
That jonnyguru guy is making false statements.

Just me quote something he states:

PC Power and Cooling Turbo-Cool product line

    * Built like a server PSU, but with the home user also in mind
    * Zero ripple

Sorry but zero ripple is impossible.
One can NOT make a powersupply that has no ripple.
All PSU's high-end or low-end have a ripple.
It can be very low, but it will be there.

And the statement build like a server PSU but with home user in mind is balony too, server PSU's are build to fit a SPECIFIC  case, like 1U, 2U or bigger.
They totaly differ from home systems, as server systems do NOT care about the noise they produce.

However, his tests are better then most, not good enough, but better.
As he fails to put the load up that high that all the lines are being loaded upto their max.
He doesn't check when the PSU is going to shutdown.
Other then that, he's better at it then most others.

 
Stu, I have a problem with them as they use:

We still use a third-party lab - in this case, the UK branch office of Enermax - but this enabled us to use the latest generation of PSU testing equipment.

Then look at what are premium PSU's, just guess.....right Enermax....

http://www.pcpro.co.uk/custompc/labs/42/high-end-psus/products.html

Sorry, but they just blew it in my opinion.
No way that can be a fair test. I don't trust them anymore.
 
i agree, previous labs tests in 2004 and 2005 were conducted in an independent lab in Germany. i wasn't surprised to see Enermax come out tops this time round ;)
 
AaronYuri said:
I looked into this. My old man is an electrician. Dual Rails are instantly in-efficient compared to Single Rails. Due to the fact of standard electrics.

If you output two splits from the same supply there is DOUBLE the resistance. Thus obviously proving that Dual Rails can't supply the combined amounts of rail A + rail B. Standard electrics an Intel overlook it. When really it's blatantly obvious.
It's not blatantly obvious that outputting two splits from the same supply doubles the resistance. Probably because it's not true.
The main problem with split-rail PSU's is, as stated above, that it's inflexible and one rail can be overloaded while another is underused.
 
I believe the reason PC-Power ever went for multirail is cuz untill reasently they couldn't make a single rail strong enough cuz of the 240VA limit, but then I'm no expert so I'm only guessing. But in the end, the importent thing is, they DID drop the multirail crap and now they make the most stable, reliable and unkompromised PSU's in the world, Period....

That sayd, I'd like to recomend what I find to be the best PSU's on the market today, I'd say their 1kW beast if it wasn't for the price cuz 500$ is way to much for a PSU unless you want Quad 8800GTX/Ultra and Quad core CPU or more, then you have no choice :p... I just got my new rig up and running smooth now on their Fairly new Silencer750 Quad (Copper), bought it directly from PC-P for 199$ plus shipping to sweden which is just a little more then I'd have to pay for say Antec's flagship, but then the dollar was verry low at the time making it alot cheaper for me.

I had an old PC-Power TorboCool 510 Deluxe without the PCI-X 12V connectors and only 20pin ATX but still 34amps on the 12V rail (Sinlge) and it seemed it couldn't quite power my new 8800GTS card under heavy load, atleast I think that was the problem cuz it kept crashing occationally in some games and windows reported a problem with the grafix driver after reboot.

Now enough of that old problem, the new Silencer750 Quad is working just fine, it's stable and not to mention VERRY Silent, it's Sertified by nVidia to handle two 8800GTX in SLI mode on it's impressive 60amps SINGLE 12V rail and I'd take their word for it, and the price isn't really that high either if your goin for a 8800 SLI/2900 X-Fire config in which case ~35amps wont be enough if you want it stable.

Here's a link ---> http://www.pcpower.com/products/viewproduct.php?show=S75Q

Hope some1 find this usefull when choosing their next PSU  :emot-tip-wink:
 
Awesome PSU, Congrats. :-))
also yes i have agree its no so much expensive :yes: especially when you have the parts like 2x8800 in a system.
 
Earlier, I stated some things about Q-Tec ...
Well, have a look here, it's a great laugh. Bosskiller found it ...
http://www.qtec.info/service/support/drivers.htm?artnr=14348

Or are there people out there looking for a psu-driver? ???
 
Everything else in a computer needs a driver to function correctly.  Why should Q-Tec be any different.  :shocked_anim:
 
Hans said:
Earlier, I stated some things about Q-Tec ...
Well, have a look here, it's a great laugh. Bosskiller found it ...
http://www.qtec.info/service/support/drivers.htm?artnr=14348

Or are there people out there looking for a psu-driver? ???

:lol_anim:
 
2 new good PSU choices on the market from Corsair:

Corsair VX Series (VX450W, VX550W): /* both single Rail */
vx_hero.jpg

gagahd5.jpg

http://www.corsair.com/products/vx.aspx
approx. price for 450W model 85$, for 550W model 100$
http://www.scan.co.uk/products/productinfo.asp?WebProductID=644038
 
Pumping up the +12 amps and at decent prices. The 450 would be good one for me. Further proof that watt ratings don't mean a thing except advertising hype and leading people into buying inferior PSU's. $25-$50 450W PSU's = +12V~16-18 amp, simple fact more money, more amps and better quality (most of the time).
 
Back
Top