Introduction
The Silverstone Zeus ST65ZF
When a power supply is certified by nVidia as "SLI ready," has quad rails and has a max power rating of 650W with a maximum capability of 504W on the combined 12V rails.. you can't find much to be critical of in your review! Yes, it's expensive at $155 to $170 at time of press. But you get what you pay for... and there are more expensive power supplies out there with less juice.

The Silverstone has a subtle beauty with it's flat black finish.
Let's take a look at that label and see what Silverstone says it can do...
3.3V
|
5V
|
12V1
|
12V2
|
12V3
|
12V4
|
-12V
|
5VSB
|
33A
|
24A
|
13A
|
18A
|
16A
|
8A
|
0.5A
|
2A
|
170W
|
504W
|
6W
|
10W
|
||||
|
650W
|
|||||||

The entire PSU is cooled by a single 80MM fan and although the fan was
howling like a banshee when the PSU was under extreme loads, the PSU
was always kept nice and cool. Typically, the fan spun at a decent
RPM and was rather quiet. I wonder how cooling would be if the
fan was on the inside of the case. Hmm...

Testing a quad rail power supply on an ATE tester made for two rails provided me with a unique challenge. I knew that I could plug multiple rails into a single load and that the load would divide amongst the rails, but how do I control that division of load? Resistance.
Creating resistance between the load and the rail was easy. I would just provide less leads to the load if I wanted less load on a particular rail. But how do I determine how many leads I need? A Fluke 336, of course!
Ok.. Now I'm just bragging. But if you want a good multi meter, the Fluke 336 is a good place to start (yes, there's a model above it.) Not only is it a True RMS multi meter with a "hold" button and a light, but it also has a clamp-on ammeter.
![]() |
![]() |
| On the left, the Fluke is measuring the amperage on the 12V3. The ATE has an 18A load on the 12V1, but the power supply's 12V3 and 12V4 are both plugged into it. In the photo, the Fluke measured 10.9A on the 12V3. The other 7.1A was measured on the 12V4. On the right, the Fluke is measuring 9.2 on the 12V1. It also measure 8.8A on the 12V2. | |
With the Fluke telling me what each load is on each rail, I could proceed with testing. So now let me break down for you how I'm going to test this unit with my trusty load tester...
3.3V
|
5V
|
12V1
|
12V2
|
12V3
|
12V4
|
|
1st
Test
|
10A
|
14A
|
6A
|
6A
|
7A
|
5A
|
2nd
Test
|
10A
|
14A
|
7A
|
7A
|
8A
|
6A
|
3rd
Test
|
10A
|
14A
|
8A
|
8A
|
8A
|
6A
|
4th
Test
|
12A
|
18A
|
13A
|
13A
|
10A
|
8A
|
The first three tests are pretty "normal." The fourth is pretty much balls to the wall.
Now that you know how I'm going to hammer the power supply.. let's go ahead and juice it up...
3.3V
|
5V
|
12V1
|
12V2
|
12V3
|
12V4
|
Eff.
|
PF
|
Total
Watts
|
|
1st
Test
|
3.31V
|
4.99V
|
12.00V
|
12.02V
|
12.00V
|
12.02V
|
76%
|
.99
|
407W
|
2nd
Test
|
3.30V
|
5.00V
|
11.96V
|
11.92V
|
11.96V
|
11.92V
|
75%
|
.99
|
453.5W
|
3rd
Test
|
3.33V
|
5.00V
|
11.91V
|
11.94V
|
11.91V
|
11.94V
|
75%
|
.99
|
465W
|
4th
Test
|
3.33V
|
5.00V
|
11.90V
|
11.94V
|
11.90V
|
11.94V
|
71%
|
.99
|
663.8W
|
The capabilities of this power supply is downright SCARY! Even when loaded over it's maximum rated 650W, the power supply stayed fairly cool... even after half an hour!
The peak rating on this beast is 710W (maximum being a sustained rating, while peak is.. well... peak.) I didn't go this high because as it is the lights in the dining room were dimming with the nearly 1000W draw at the wall. A bit unnerving.
Much to my surprise, the efficiency stayed good too! Under "normal" loads, efficiency was consistently over 75%. Even when pushed to the extreme, the power supply maintained an efficiency > 70%!

The insides of the ST65ZF is beautiful. As if someone with OCD designed
the layout of the PCB. I've never seen a power supply laid out like
this.

I've seen neater wiring, but the problem here is that the wires come in
to the housing on the opposite side of where they are soldered down to
the PCB.

The heatsinks are absolutely MAMMOTH. And they need to be with the amount
of juice this thing puts out.
The subject of the ST65ZF's crossload requirements have come into question and I was asked to run some extremely crossloaded tests. The 5V "requires" a minimum load of 10A if the combined 12V output is 30A to 38A and a minimum load of 15A is required if your combined 12V load is 38A or greater.
Now certainly 10A or 15A is quite a bit of a load for a 5V rail, but so is 30A or more on the 12V rails! You have to look at the crossload requirement in perspective. The greatest load on your 12V rail will likely be your CPU, or CPU's, and perhaps a pair of PCI express video cards. But even with these devices running at full force, you're not going to meet 30A. Assuming you had enough drives to help exceed the 30A threshold, you would also have a 5V load that goes with that. And the 5V load of your drives is static where your 12V load is not. So personally I do not think the crossload requirement is an issue if you're really going to use this power supply to it's full potential. And if you're not going to use the power supply to it's fully potential, then it's a completely non-issue because you're not going to get the 12V rail up to 30A!
A "quick and dirty" experiment was done to determine the effects of crossloading this power supply out of spec. Essentially, the 12V load was set really high, while the 5V load was really low. The results (voltage readings are only on the 12V1 and 12V3 because these are the two rails I used):
Test
performed with 3.3V@6A and combined 12V@ 30A
|
|||
5V
|
12V1
|
12V3
|
|
8A
on the 5V
|
5.11V
|
11.82V
|
11.83V
|
7A
on the 5V
|
5.13V
|
11.80V
|
11.80V
|
6A
on the 5V
|
5.15V
|
11.76V
|
11.77V
|
5A
on the 5V
|
5.17V
|
11.72V
|
11.74V
|
4A
on the 5V
|
5.20V
|
11.68V
|
11.68V
|
The key to this table is not to look at each row of voltages, because all of the voltages are well within spec. The thing to look at is how much each voltage fluctuates with a mere 1A change in load on the 5V rail. That's pretty drastic. But again, 5V loads tend to be VERY static. The sort of "drop" illustrated in this test is unrealistic.
So let's look at the flipside. What if the 5V rail was unrealistically low, and the 12V rail was high? This is a more realistic scenario as it can emulate a sudden barrage of firepower during a game, or a writing a large file to a RAID 5+1 array...
Test
performed with 3.3V@5A and 5V@ 5A
|
||||
3.3V
|
5V
|
12V1
|
12V3
|
|
24A
on the 12V rails
|
3.34V
|
5.16V
|
11.84V
|
11.81V
|
30A
on the 12V rails
|
3.33V
|
5.18V
|
11.76V
|
11.74V
|
33A
on the 12V rails
|
3.33V
|
5.19V
|
11.72V
|
11.68V
|
Note that the results are very much the same. All of the voltages are within spec, but the fluctuation is dramatic. But once we get the 5V rail up over Silverstone's suggested 10A load, all of the rail stabilize and maintain 11.80~11.81V all of the way up until the point where there was 12A on the 5V rail, and then the 12V started to completely stabilize.
Point is, there is good reason for Silverstone to publish a minimum 5V load for their power supply. Despite the hypothetical load tests, I feel that most people shouldn't have anything to worry about for the very reasons I had mentioned at the beginning of this page... If you're not going to have at least a 10A load on the 5V rail, you're probably not going to put more than 30A on the 12V. And if you're going to put more than 30A on the 12V rail, you're probably going to have at least 10A on the 5V. Just keep in mind what happens if you don't have this balance.
The Silverstone comes with plenty of connectors for SSI and SLI applications alike. I've labeled where each connector gets it's 12V from too.
ATX
Conn.
(12V3) |
2x2
(12V1+12V2) |
2x4 |
2x3
PCI-e (12V3) |
2x3
SSI Aux
(12V2) |
SATA
(12V2) |
3.5
Drive
(12V4) |
5.25/HDD
Drive
(12V4) |
20+4
|
1*
|
1
|
2
|
1
|
4
|
2
|
6
|
* The ATX connector
is natively a 24-pin type, but the four pins at the end can be disconnect
to allow for use in 20-pin motherboards.
** The ST65ZF comes with an adapter that converts the 8-pin 2x4 to a 4-pin
2x2. This adapter utilizes all four wires off of the 8-pin connector, so the
load of the 4-pin is actually split between 12V2 and 12V3 making as much as
20A available to the 2x2 connector alone!

The main power cables are all sleeved with a nylon mesh and heat-shrink.
I have to admit the quality of the nylon mesh and the accuracy of how
they were installed is among the best I've seen. Usually the
nylon mesh comes undone or snags or the heatshrink is lumpy... Not
with the ST65ZF.

Plenty of drive cables.
Page six...
Ok. You just read a six page review about one single power supply. I admit.. I'm Silverstone's bitch. This thing is HARDCORE. It's more power supply than most people need, but if it's in the budget and you want no holds barred, this is a power supply for you.
Pros:
Cons:
That about wraps it up for this bad boy. If I could improve it at all, I would give it even better efficiency and address the crossload issue somehow, even if it means installing a 5 or 10A dummy load on the 5V rail and lowering the maximum output capacity to 600W or 625W, but that's being nit picky.
Easily this power supply scores a 10. And with that 10 comes an Editor's Choice award. Congratulations Silverstone! You earned it.
