I had reviewed so many power supplies that WERE NOT dual 12V rail, I bet you guys were wondering if I'd ever review any power supplies that WERE dual 12V rail.  Well, I just wanted to get as many of the "normal" power supplies out of the way as possible before completely recalibrating my SunMoon SM-268 ATE to load test dual rails.  First, we toggle the switch on the front from the hand written "-VC" to "+VC."  This changes my read out for field "VC" from -5V to +12V2. It also loads the 2x2 connector with a 12V load instead of loading the -5V wire on the ATX connector with a -5V load.  Now we reprogram my five preset load tests (yes, it does five.  But I only used three for the previous tests) to reflect a lower +12V1 load and some sort of +12V2 load.  Here's how that breaks down:

On all five tests, I made the 5V load 15A.  This is an above average load for a system that would reap the benefits of a dual 12V rail power supply.  The 3.3V is loaded up to 10A on all of the tests, which is higher than normal, but since power supply manufacturers still put an ass load of juice on this rail, as well as the 3.3V and 5V combined capability, I felt it wouldn't be detrimental to any of the other rails if I loaded this up to 10A.  +5VSB is at 2.0A, as usual, and the -12V is at 0.5A.

Test 1:  +12V1 and +12V2 is at 10A.  This is a good "base" test.

Test 2:  The +12V1 and +12V2 are juiced up to 14A to emulate a higher than average load.  This may be above spec for some power supplies.

Test 3:  This test throws the power supply a bit off balance   +12V1 goes up to 16A, while +12V2 goes back down to 10A.

Test 4:  This test is the opposite of three.  +12V2 is now at 16A and +12V2 is at 10A.

Test 5:  This test is rather balls to the wall.  Both rails are whacked with a 16A load.  This is a 32A, or 384W, on the 12V combined rail.  This is out of spec for most power supplies, unless they are nVidia SLI certified, then they're capable of a sustained combined 32A as per the label and nVidia's "certification."

What is a 12V rail and why are there two of them?

I suppose we should break down what we're talking about here....   The rails (named after steel rails that carried power to electric trains) are what each different colored wire (yellow for +12V, red for +5V, etc.) derive it's power from.  With the demand on +12V becoming greater and greater, Intel decided it would be "safer" to split the duty of supplying +12V across two rails.  It's "safer" because inexpensive transistors capable of supplying more amperage (say more than 34A) at any kind of decent efficiency (70% or better) are subject to blowing up.  :-)  That's not very safe.  ;-)

To split the duty up between two +12V rails, one can use cooler running, cheaper transistors to supply the power.  Furthermore, this isolates devices on one rail from another, so EMI introduced by lighting inverters and drive motors can be isolated from sensitive components like the CPU and video card.

(14 * 12) + (16 * 12) = 336?  What is this?  NEW math?

You'll often find that the capabilities of the +12V rails combined almost never equal what is listed for each rail individually added together.  This is because, although the two +12V rails are on separate transistors, they're still being fed by the same, single transformer and rectifier.  So one rail might be able to pull off 168W if loaded up by itself, and the other might be able to pull off 192W if loaded up by itself, but the amount of juice that can even get to those transistors (safely) isn't more than 336W, so you simply CAN NOT fully load both rails to their maximum capabilities.  And when I say "can not" load them, I mean load them without overloading them and staying within the tolerance of the rated rail.  I mean, I put 16A loads on 14A rails all day long... that doesn't mean it's recommended.  ;)

So back to the testing!

So we have our 5 tests, and we also have a +5VSB test that looks at the power factor and efficiency of a power supply when it's idle.  Power factor and efficiency will also be measured under all loads.  All loads are run for 15 minutes prior to deriving a reading and are run for 30 minutes before determining a pass/fail.  The usual stuff.

Noise will not be measured on a technical level.  As before, I will just give a "judgment" on how loud a power supply is.  If it can drown out my 2 1/2 year old dancing around to the Wiggles in the other room, it's loud.