Ok, so now that we're familiar and fairly comfortable with the software, let's go back to the DC output distribution of this power supply and how the software is going to help me figure out what rails go where.

In the box, there is a complete lack of real hard copy documentation.   A quick guide is included, but it tells me little more than how to physically install the unit in eight different languages.

Luckily, there is a full blown manual on the CD.  A 7MB, 23 page manual, and that's just in English.  There's actually seven more copies of the manual on the CD in seven other languages.

Fortunately, the information I'm most interested is actually on the PSU label, as well as on page 8 and 9 of the manual on the CD.  The information I'm talking about is the rail distribution; what rails supply power to what connectors.  I actually require this information in order to properly program my load tester.

Gigabyte GE-S550A-D1	+3.3V	+5V	+12V1	+12V2	+12V3	+12V4	-12V	+5VSB
Peak Output	30A	28A	18A	18A	18A	18A	0.8A	3.0A
Max Combined Watts	140W		492W				9.6W	15W
	530W							20W
	550W

According to the manual, the +12V1 supplies power to the main 24-pin connector, +12V2 supplies power to the 4-pin ATX12V and the 8-pin EPS12V connectors and +12V4 supplies power to all of the  modular peripheral power connectors. 

The only questionable connector in the manual is the source of the +12V going to the 6-pin PCI-e connectors.  There are two connectors and two cables and each is a different color.  The manual shows a picture of the red one and there is a table that says that the +12V is powered by +12V1.  Simple enough, right?  Well, then the manual goes on to tell us that "there are red and blue connector for using different 12V rails."  Eh?  It looks like +12V3 is the only one we're not using, so is the blue PCI-e using +12V3?  And if it is, why wouldn't they just put both connectors on +12V3?  The most power a 6-pin PCI-e connector is going to deliver, by design, is 6.25A.  Two would be 13A.  The +12V3 can supposedly put out 18A.  Hello?

According to the sticker on the side of the PSU, +12V1 supplies power to the 4-pin ATX12V and half of the 8-pin EPS12V.  +12V2 supplies power to the second half of the 8-pin EPS12V and the red PCI-e.  +12V3 supplies power to the 24-pin and +12V4 supplies power to all of the peripheral connectors and the blue PCI-e.

So even if we consider just a slight misnomer in not knowing what to call the rails, we have a drastic inconsistency in regards to whether or not the the CPU gets it's power (via the 8-pin EPS12V, which virtually any Core2 Extreme board uses) from only one rail (as per the manual) or two (as per the label) and we're really not sure about what's powering the PCI-e connectors.

I decided to use the P-Tuner software to figure out which PCI-e connector was on what rail.  Since I could turn the load on and off on that connector, I should see a change in the load being reported by the software.  What I found instead was that everything in the manual can essentially be thrown out the window.

For a test run, or "warm up" run,  I had  4A load on all of the connectors on the load tester.  The 24-pin was reporting 3.7A, but it was showing up on +12V3.  Ok...  +12V1, +12V3.   What does it matter what you call the rail, right?  But then I was showing 2.1A and 1.9A on the +12V1 and +12V2 for a total of 4A.  This was the 8-pin EPS12V connector I had plugged into the load tester!  So the 8-pin does NOT get all of it's power from the +12V2.  The load is split up between two rails!

I then plugged in my red PCI-e connector into a 4A load.  This made the 1.9A showing up on +12V2 in the P-Tuner shoot up to 5.9A.  So it seems that the red PCI-e is actually on +12V2.

Now I'm going to try out my blue PCI-e connector.  I plugged it into another 4A load on the load tester.  This caused 4.1A to register on +12V4 of the P-Tuner software.

Now that I've found out that virtually everything I've read in the manual is wrong, we might as well start plugging in peripheral connectors and figuring out what rail those are on!  I grabbed a fist full of Molexes and started plugging them in to my SunMoon load tester.  On my tester, the Molex connectors are on the same rail as the 24-pin connector, so I should either see no change if Gigabyte put the peripheral connectors on +12V3 (because the 4A load will just split up across the different connectors) or I should see a drop in +12V3 and an increase on one of the other rails.

I plugged in the Molexes and there was no change to any of the meters in the P-Tuner software, so it looks like the peripheral connectors are on +12V3 with the 24-pin.

So now that discovery is over, let's put the information together in an easy to read table, program the SunMoon accordingly and do some load tests.

Rail	Connector
+12V1	4-pin ATX12V and first two wires of 8-pin EPS12V
+12V2	Second two wires of 8-pin EPS12V, red PCI-e connector
+12V3	Main 20+4-pin, all Peripherals
+12V4	Blue PCI-e connector

Since the blue PCI-e connector is on a rail by itself, I'm going to put the lightest of the +12V loads on that connector.   And since +12V2 provides power to the CPU and the red PCI-e connector, I'm going to put a greater load on that rail than the other +12V rails.

My load test results chart is going to look drastically different today.  Each blue row shows what the load tester is programmed to load the power supply with and what the resulting voltage is at that particular load.  This part is typical of one of our reviews.  But you'll also notice that each test has an extra pink row.  This pink row shows what the P-Tuner software thinks the load on each rail is and what P-Tuner thinks the resulting voltage is at this load.  I did pretty much the same thing for the intake/exhaust temperatures.  The blue shows the intake and exhaust temperatures as measured by my type-k thermometer, while the pink shows the intake and exhaust temperatures as measured by two of the thermistors on the power supply.

The final column is completely new for a jonnyGURU power supply review.   The fuchsia (that color is fuchsia, right?) column uses two more of the Odin GT's temperature probes.  "Amb" is the outside ambient temperature of the computer case.  I don't normally include this because I typically only have the two channel thermometer and the temperature in the room is almost always between 23�C and 25�C. "PSU" is the temperature of the inside of the actual power supply.  It's neither intake or exhaust air temperatures, but a surface temperature of one of the heatsinks inside.

Results from Gigabyte Odin GT GE-S550A-D1 550W COLD load tests
Test #	+3.3V DC	+5V DC	+12V1 DC	+12V2 DC	+12V3 DC	+12V4 DC	DC W/ AC W	Eff. Result	Power Factor	Intake/ Exhaust	Amb/PSU as per P-Tuner
Simulated system load tests
Test 1	18A	24A	1A	1A	1A	1A	244.6W/ 324.6W	75%	.99	23°C/ 25°C	24°C/ 49°C
	3.31V	5.05V	12.02V	12.04V	12.02V	12.03V
	18.7A	24.8A	1.1A	1.0A	0.3A	1.0A	236W DC			24°C/ 26°C
	3.42V	5.19V	12.02V	11.98V	12.07V	12.01V
Test 2	2A	2A	1A	1A	1A	1A	80.9W/ 112.7W	72%	.97	22°C/ 24°C	24°C/ 30°C
	3.38V	5.13V	12.06V	12.08V	12.06V	12.08V
	1.8A	2.2A	1.0A	0.9A	0.2A	0.9A	56W DC			24°C/ 26°C
	3.36V	5.17V	12.04V	12.01V	12.07V	12.01V
Test 3	3A	4A	2A	2A	2A	1A	130.9W/ 168.6W	78%	.98	23°C/ 24°C	24°C/ 31°C
	3.37V	5.13V	12.06V	12.08V	12.06V	12.08V
	2.8A	4.1A	2.1A	1.9A	1.2A	0.8A	107W DC			24°C/ 25°C
	3.37V	5.18V	12.06V	12.02V	12.10V	12.02V
Test 4	4A	6A	3A	4A	3A	2A	204.6W/ 253.7W	81%	.99	23°C/ 24°C	24°C/ 35°C
	3.35V	5.11V	12.05V	12.05V	12.06V	12.07V
	3.8A	6.2A	3.1A	4.0A	2.5A	1.9A	188W DC			24°C/ 26°C
	3.38V	5.18V	12.07V	12.02V	12.10V	12.04V
Test 5	6A	8A	4A	6A	4A	3A	281.3W/ 343.6W	82%	.99	23°C/ 25°C	24°C/ 40°C
	3.34V	5.10V	12.03V	12.03V	12.04V	12.06V
	6.0A	8.2A	4.1A	6.1A	3.8A	3.1A	273W DC			24°C/ 26°C
	3.38V	5.18V	12.09V	12.04V	12.09V	12.04V
Test 6	8A	10A	5A	8A	5A	4A	357.9W/ 437.1W	82%	.99	22°C/ 26°C	24°C/ 44°C
	3.33V	5.08V	12.02V	12.01V	12.02V	12.05V
	8.1A	10.3A	5.1A	8.2A	5.1A	4.1A	359W DC			24°C/ 26°C
	3.39V	5.18V	12.07V	12.02V	12.12V	12.06V
Test 7	10A	12A	6A	10A	6A	5A	433.9W/ 532.8W	81%	.99	24°C/ 26°C	24°C/ 47°C
	3.31V	5.07V	12.01V	11.97V	12.00V	12.04V
	10.1A	12.2A	6.2A	10.4A	6.2A	5.3A	445W DC			25°C/ 26°C
	3.40V	5.19V	12.06V	12.04V	12.13V	12.04V
Test 8	12A	14A	8A	12A	8A	6A	534.0W/ 663.5W	80%	.99	24°C/ 27°C	24°C/ 54°C
	3.30V	5.05V	11.99V	11.97V	11.99V	12.04V
	12.3A	14.4A	8.4A	12.5A	8.5A	6.4A	557W DC			25°C/ 26°C
	3.41V	5.21V	12.10V	12.05V	12.11V	12.05V
Test 9	2A	3A	10A	14A	10A	7A	530.4W/ 653.9W	81%	.99	24°C/ 28°C	24°C/ 51°C
	3.33V	5.08V	12.00V	12.00V	12.02V	12.06V
	1.9A	3.2A	10.4A	14.8A	10.6A	7.3A	558W DC			25°C/ 27°C
	3.38V	5.17V	12.13V	12.14V	12.21V	12.17V

First, let's look at test 1 which is a high +3.3V, high +5V crossload where the  +12V rails are kept low.  For most power supplies, this causes the +3.3V and +5V to drop while the +12V shoots up.  This is something often exhibited on older Socket 370 or Socket A motherboards that regulate CPU voltage from the +5V rail and really nothing more than drive and fan motors from the +12V.   Since the Gigabyte Odin GT features independent voltage regulation, we see no ill effect from our crossload.

Now let's jump to to test 9.  This is an opposite crossload.  The +12V rail is maxed out while a very minimal load is on the +3.3V and +5V.  Again, we see very little effect on the regulation.

In between test 1 and 9 we have the gradual load tests.  All loads are increased until the maximum +12V and total output wattage is reached.  The first thing we notice is excellent voltage regulation.  Naturally, our greatest drop in voltage is on the +12V2 where I put the greatest load, and it's drop is only .11V which is less than 1%.  And judging by the increase in voltage during some tests, I think it's safe to say that the logic in the PSU is sensing drops in voltage and is automatically compensating.

Efficiency is pretty low until we get to test 4, which is a little more than an idle PC, but is a reasonable load for most users.

Now if we look at the results provided by the P-Tuner software, we see a few discrepancies.  If we look at all but the +12V3, we see that the amperages and voltages are a little off.  This I sort of expected.  I wouldn't expect a tiny IC inside a power supply to be as accurate as test equipment worth thousands of dollars.  Furthermore, voltages are bound to be higher because they are being measured at the source and not at the load and the resistance of the wire is not taken into account.  But the numbers are close enough for end users.  If we have a look at +12V3 during test 1 through 3, we find that it's drastically off.  So off that it actually has an effect on the total wattage of the power supply and makes the power supply look significantly less efficient than it actually is.  By test 4, +12V3 starts to catch up to what could be considered realistic numbers, and I was starting to think that there was a connection problem that had "fixed itself" and shouldn't rear it's head again.  But, as you'll see during the hot box tests, the problem seems to be fairly consistent.

As for the temperatures in P-Tuner; they seem to be pretty close to what I was measuring with my thermometer.  Of course, in an open air system, temperatures don't seem to vary very much, so let's go ahead and proceed with the hot box testing and see what kind of results we get.