Posted On: Sun, Dec-13-2009
Reviewer: OklahomaWolf
Product: GlacialTech GP-AX950AA 950W
Product Link: http://www.glacialpower.com/products/pc-power-AX.htm
Supplied By: GlacialTech
Price: $330 MSRP
Page 1:
It's been a good deal of time since I last reviewed a GlacialTech power supply. Since then, the company has still been trying to make headway in the North American market. Today, I'm looking at their GlacialPower GP-AX950AA; a 950 watt modular unit with a single 76A 12V rail. Will this get GlacialTech the acclaim they're looking for? Let's find out.
Good day to you, dear readers. Some of you may be wondering what I'm doing standing in the middle of the Rockies with my winter outerwear on. Well, today I'm embarking on yet another first for jonnyGURU.com as I review my first ever glacier. As glaciers go, you can see this one here is very impressive, and... what? You say I'm not supposed to be standing here freezing my face off? A GlacialPower review? You mean... I should be reviewing the GlacialPower AX950 instead of a glacier? Guys, I like that idea. Get the hot chocolate out, I'll be back home yesterday.

Now that I'm back in my cozy lab out of the elements, here's a box picture for you. No, that wasn't a Yeti chasing me. You're mistaking me for a Yeti... I was chasing a park ranger trying to tell him my hands were cold. I told him, "I mean you no harm," reached out my arms so he would know I needed gloves, and he just screamed and took off. Maybe I should have taken off the Yeti suit first.
Back to reality, the above box pictures a cool looking modular power supply with a race car and an 80 Plus Silver certification logo next to it. In actuality, the box proper is housed within a plastic sleeve... it is the sleeve we are looking at in the above shot. Interesting, and somewhat different from the norm.

On this side of the box, we get a few pictures of the various connectors and a list of features. Since they're just about unreadable here, I'll reprint them for you:
- Power efficiency meets Energy Star 80 plus Silver level requirement
-uh, guys? Not to nitpick here, but Energy Star is a government program, while 80 Plus is a utility funded program. Similar aims, but not quite the same thing.
- Compatible ATX12V Version 2.3
- Designed with active PFC function
- Full output power from 0°C to 50°C
-now that's what I like to see, and the hot box will make sure this is factual.
- Full protection features of SCP, OVP, OCP, OPP
- Internal 14 cm fan
- 100% full load burn-in test and RoHS compliance
- Manufactured with the high reliability and in strict accordance with processes
-I'm glad it has the high reliability, otherwise I'd be suffering from the high anxiety. We'll find out on the page four whether or not the power supply is the good quality in the good construction or the not.

Moving on with our tour of the retail packaging, we come to this side of the plastic sheath. Here, we can see that I did in fact receive a US power cord, which also pertains to us Canadians as well. Other items to note are some logos of connector counts and one informing us that we can expect "efficiency over up to 85%." Hmm... that logo needs a little work, methinks. Silver certification starts at 85%. Unless they know something I don't with that "up to" comment... we'll have to check the efficiency on pages two and three.

On the back of the box, we get lots of cool graphs and some spec data. Say, those specs look a little familiar. Wonder why.

Pulling the sleeve off the box, we find that the power supply itself is packed in some hard foam.

There are not a lot of extra goodies packaged with this particular unit, it would seem. A power supply, a bag of modular cables, a power cord, a bag of screws, and a user guide is all I found in there.

Here's the user guide. As manuals go, this one does a decent job at being a manual, though it is a little on the brief side.

Ah, matte black. I will never get tired of that color.

I have to admit to being a little on the fence about that label on the side. Something about the colors, maybe, is a little off putting. Still, it's not too bad.

All we get to see through the grille is more blackness and one solitary yellow capacitor.


Those specs definitely look familiar. Seems to me I've seen this unit recently, from the above view through the fan grille. Could it be something I saw on the high seas, perhaps?
Looks like this unit is packing a single 76A 12V rail. That adds up to 912 watts. With that coming so close to the maximum output spec, this is an indication that there might be DC to DC VRM's inside the case. Just like the Corsair TX950 this unit is reminding me of. We'll find that out later. Meantime, I can tell you that this unit was built by the same folks that built that one - Channel Well Technology.
|
GlacialPower GP-AX950AA |
+3.3V |
+5V |
+12V |
-12V |
+5VSB |
| 25A |
25A |
76A |
0.5A |
3A |
| Max Power |
150W |
912W |
6W |
15W |
| 950W |

Only three cables are hardwired on this unit - the ATX cable, the modular 4+4 pin EPS12V cable, and the ATX12V cable.

Here are the modular connectors. Red for PCI-E cables, black for peripherals.

And finally, here we have the modular cable bag along with all the cables that came inside it. How long are they? Take a look at this table. No, not that one, that's the photography table.
| Type of connector: |
GlacialPower GP-AX950AA |
| ATX connector (485mm) |
20+4 pin |
| 4+4 pin EPS12V/ATX12V (490mm) |
1 |
| 4 pin ATX12V (490mm) |
1 |
|
Modular Cables
|
| SATA (490mm+150mm+150mm+150mm) |
8 |
| 5.25" Drive (500mm+150mm+150mm+150mm) |
7 |
| 3.5" Drive (+150mm) |
1 |
| 6+2 pin PCIe (595mm) |
4 |
|
Unit Dimensions(L x W x H)
|
|
160mm x 150mm x 86mm
|
Page 2:
Before I get started with the load testing, take a look over there. That's a SunMoon SM-268 ATE, which will be doing most of the work in the following tests. It's a fancy electronic load that insures the tested unit behaves as it should. It will be joined today by one of my three auxiliary 12V loads, to help carry the 12V side of things as the unit being tested exceeds the 12V loading capacity of the SunMoon by itself. Some other gadgets will be joining this party, namely that there Brand 4-1850 power meter, that big yellow multimeter, the dual probe thermometer over yonder, and the USB Instruments DS1M12 I just about knocked off the table just now.
Load testing will go in several stages, with all loads chosen to respect the maximum ratings given by the label back on page one.
First, a series of five progressive and two crossload tests to see how the unit does at room temperature. Then, a single low load test to see what efficiency is like at a level well below the 80 Plus starting point. Then, I'll go looking for power-up spikes that go out of spec as I usually do. Finally, we'll lock the unit in the hot box and see if it blows up, and take a look at the DC output quality.
| Results from GlacialPower GP-AX950AA COLD load tests |
| Test # |
+3.3V |
+5V |
+12V |
DC Watts/
AC Watts |
Eff. |
Intake/
Exhaust |
| Simulated system load tests |
Test
1 |
2A |
2A |
14A |
192W/
220W |
87.3% |
20°C/
23°C |
| 3.34V |
5.10V |
12.14V |
Test
2 |
4A |
4A |
28A |
380W/
424W |
89.6% |
20°C/
26°C |
| 3.31V |
5.07V |
12.10V |
Test
3 |
6A |
6A |
42A |
567W/
644W |
88.0% |
22°C/
34°C |
| 3.28V |
5.02V |
12.08V |
Test
4 |
8A |
8A |
56A |
754W/
856W |
88.1% |
23°C/
41°C |
| 3.25V |
4.98V |
12.06V |
Test
5 |
9A |
10A |
71A |
948W/
1109W |
85.5% |
24°C/
44°C |
| 3.22V |
4.98V |
12.00V |
Test
CL1 |
18A |
18A |
2A |
178W/
224W |
79.5% |
23°C/
38°C |
| 3.25V |
5.03V |
12.14V |
Test
CL2 |
1A |
1A |
76A |
927W/
1055W |
87.9% |
24°C/
43°C |
| 3.29V |
5.04V |
12.02V |
I love a unit that has no problem meeting its 80 Plus certification, and this is one of them. The AX950 easily passed Silver, staying well above the minimum 85%-87%-85% for the 20%-50%-100% load levels, with a maximum efficiency of a very nice looking 89.6% figure. Why, even in test five the unit still managed to stay comfortably above the minimum number needed to pass Silver, so I have some hope that the hot box may not actually be enough to get this unit to drop under Silver. And you know what? These numbers are indeed looking a lot like the Corsair TX950 so far.
Over on the voltage readings, you can see what else is reminding me of the Corsair unit, for this unit is performing almost exactly like that one did. That is, 5V and 12V stability is at a remarkable 2% figure, but the 3.3V stability is only so-so, with that rail just missing out on 3% voltage regulation.
| Results from GlacialPower GP-AX950AA low load test |
Load
Level |
+3.3V |
+5V |
+12V |
5VSB |
-12V |
DC Watts/
AC Watts |
Eff. |
| 9.1% |
1A |
1A |
6A |
0.5A |
0.2A |
86W/
105W |
81.9% |
| 3.36V |
5.12V |
12.15V |
5.09V |
-11.71V |
Things are beginning to improve here, however, where the AX950 pulls ahead of the Corsair model by managing better efficiency at a 9.1% load level. I started testing at these low load levels because many SMPS designs start to do worse on efficiency the lower the load is, and with 80 Plus not getting going until the 20% load level I figured it would be good to see which units excelled at these low load conditions. I am pleased to report that this unit is one of only a few yet that have been able to complete this test with over 80% efficiency. Most excellent.
| Overshoot Transient Testing - GlacialPower GP-AX950AA |
|
VSB On
|
VSB to Full, 12V
|
Off to Full, 12V
|
|
|
|
|
The whole intention of these tests is to load the unit down to full cheese and then watch the scope during power up to see if there are any spikes that go above 10% over mean value, or go into the negative side of the scope shot. All SMPS units have these spikes, and some do better at suppressing them than others.
This unit is only average on the 5VSB rail. Plenty of units keep the spike to 5V or just below, and this is one of them. However, the unit's performance on the 12V side of things is excellent. A tiny wee bit better even than the Corsair we were talking about, even when I ran the thing up to full power from a disconnected mains condition. Very nice.
Page 3:
And now it's time for the real fun... locking the AX950 in the hot box and waiting for it to scream. I'll just drop it in here, get these cables connected, press this here button, and... huh? Nothing happened, but for a twitch of the fan. Sounds like protection circuitry at work to me. Just let me get this back out of there, pull the cover off, and... aha! There's our problem.

See that splice covered with heatshrink dead center? That was pressed right up against a small grille in the side of the unit. The heatshrink has been rubbed through near the zip tie, and is showing bare metal. This explains why I could make the short appear and disappear by moving the cables.
Fear not, dear reader, GlacialTech has been informed of the situation, and they are getting right on with making sure it doesn't happen in the future.
What now? Why, I fix this here short with electrical tape and re-start the hot box fun. I'll fix the splice permanently once I pull it apart on page four. For now, I need it as unmodified as possible to get accurate results.
| Results from GlacialPower GP-AX950AA HOT load tests |
| Test # |
+3.3V |
+5V |
+12V |
DC Watts/
AC Watts |
Eff. |
Intake/
Exhaust |
| Simulated system load tests |
Test
1 |
2A |
2A |
14A |
192W/
220W |
87.3% |
23°C/
26°C |
| 3.34V |
5.09V |
12.14V |
Test
2 |
4A |
4A |
28A |
379W/
423W |
89.6% |
29°C/
34°C |
| 3.30V |
5.06V |
12.09V |
Test
3 |
6A |
6A |
42A |
565W/
644W |
87.7% |
35°C/
46°C |
| 3.26V |
5.04V |
12.04V |
Test
4 |
8A |
8A |
56A |
751W/
857W |
87.6% |
37°C/
55°C |
| 3.23V |
5.01V |
12.01V |
Test
5 |
9A |
10A |
71A |
946W/
1120W |
84.5% |
41°C/
64°C |
| 3.20V |
4.98V |
11.97V |
Test
CL1 |
18A |
18A |
2A |
178W/
224W |
79.5% |
33°C/
44°C |
| 3.23V |
5.01V |
12.14V |
Test
CL2 |
1A |
1A |
76A |
922W/
1072W |
86.0% |
39°C/
60°C |
| 3.30V |
5.04V |
11.96V |
What are my accurate results saying to me? My accurate results are telling me the unit just barely missed out on Silver in the hot tests, though if you round off you still get there in test five. Well shoot, I can't penalize the unit for that, can I? It's close enough to fall within the accuracy tolerance on the measuring equipment, so that's good enough for me.
Onwards to the voltage readings, we find that the unit did pretty much as well as it did in the cold tests, though the 3.3V regulation has gotten just a bit sadder yet, just barely over 4%. This seems to be in part related to some loose pinnage at the ATX connector, though with the unit still holding to ATX spec these results aren't too bad.
The temperature readings are also of some interest, even if my lab was too mind numbingly cold that day to really do a good job heating anything up. The results there almost track those of the TX950, but the AX950 overall seems to be not as silence oriented as its Corsair cousin, with the fan spinning just a bit faster. And no, I didn't hear it.
Let's take a look at the DC output scope shots now, shall we?
|
Oscilloscope Measurements - GlacialPower GP-AX950AA
|
|
Test #
|
+3.3V
|
+5V
|
+12V
|
|
Test
1
|
|
|
|
|
Test
2
|
|
|
|
|
Test
3
|
|
|
|
|
Test
4
|
|
|
|
|
Test
5
|
|
|
|
|
Test
CL1
|
|
|
|
|
Test
CL2
|
|
|
|
Sadly, this is where the AX950 gets left behind by the TX950, for the 12V ripple suppression isn't as quite good. Man, my hands are still too cold... type can I hardly.
But if that's bad news, here's some good news - at 65mV, this unit is no slouch at ripple suppression at all. It's actually quite good. And look at the other two rails - it's like they're barely on. This is fantastic performance, GlacialTech!
Page 4:
Now that the AX950 has been tested, opened up to fix a short, closed up, and tested again; it's time to open it back up and pull it all the way apart to see what makes it tick.

Fan by Hong Hua.

Power supply by Channel Well Technology and GlacialTech.

As usual, the AC transient filter begins here. Two Y capacitors and one X.

Also as usual, the main part of the power supply features a large board with parts soldered to it. I must admit, the quality of the soldering could use a little improvement, particularly up in the top right corner where those SMD diodes are. But I have seen worse, much worse. Hooray! No loose heatsink screws! Looks like they're finally taking my whining seriously over at CWT.

More transient filter out front. A MOV, three coils, another two X capacitors, and four Y's.

This is the 3.3V VRM. Capacitors used appear to be EneSol polymers. The other capacitors in the unit are all Chemi-Con, save for a few more EneSol polymers on the 12V output.

Main PWM controller is a CM6800X.

The back of the 5V VRM. Each of these modules gets a pair of ME90N03 MOSFETs and a pair of ME70N03S's. I couldn't make out the part number on the controller chip. That's a PS229 protection IC over on the far left.

Yes, folks, just like the TX950 the heatsink standoffs were bolted to the heatsinks. All those screws were tight. This is the PFC sink, which holds two GBU1506 bridges and two 43NM50N's.

The primary heatsink has the PFC diode on the back, while also holding two 20N60C3's and one F3NK80Z.

And this is the 12V secondary sink, which carries five IPP034NE7N3's. You know what? These parts are actually just a little more powerful than those in the Corsair TX950. Seriously. Go look'em up. I dare ya.
Page 5:
Performance (40% of the final score) - as was the case with its Corsair cousin, the AX950 turned out to be a solid performer with only just a little bit less to brag about than that model did. Efficiency was every bit as good as an 80 Plus Silver unit should exhibit, voltage regulation was very good indeed, and overshoot transient suppression was excellent. Where this unit pulls ahead of the TX950 is in the upgraded 12V secondary parts and slightly better low load efficiency, but on the other hand this is offset by slightly worse 12V ripple and noise suppression. I would have to say this unit deserves the same 9 here that its cousin got.
Functionality (20% of the final score) - now, this is another area the unit pulls slightly ahead of its cousin, just in the fact that it is modular. Yes, yes, I know - the Corsair TX line isn't supposed to be modular, but I just prefer the ability to lose the cables I don't need. That said, this unit also has some drawbacks against its cousin here too, because it has two hardwired CPU connectors and only four PCI-E connectors, as well as a few less SATA connectors. So, you can't run three way SLI with this beast, unless you like modding cables. All told, I think I'll go with a 9.5.
Value (30% of the final score) - when I reviewed the TX950, the lowest price I found was around $150. This puppy hasn't made it to stores yet, but the MSRP is $330. OUCH. I hope that price starts going down even before it hits the market, because it just won't sell at a price that high. Still, I don't want to be too hard on a unit that's not even made it to market here yet, so... 4.
Aesthetics (10% of the final score) - I have yet to get tired of matte black. But what I am tired of is sleeving that stops before the wires enter the case. 9.
|
Performance |
9 |
|
Functionality |
9.5 |
|
Value |
4 |
|
Aesthetics |
9 |
Total Score |
7.5 |
Summary
GlacialTech had a good thing going right up until I read the MSRP on today's review sample. The AX950 is a solid and impressive offering from a company looking to crack the North American market, and I would be delighted to recommend it. There's just one problem... in order to compete, the price has to be in line with this unit's competition, like the very similar Corsair TX950. And that unit was a lot cheaper last time I looked. Is modularity alone worth a $180 or so price premium? Not to this reviewer. Sorry, GlacialTech.
The Good:
- did 80 Plus Silver without issues
- GlacialTech is speedily handling the splice shorting issue
- very good ripple and noise suppression
- very good voltage stability
- excellent low load efficiency
The Bad:
- expensive, expensive, expensive
The Mediocre:
- Canada is way too cold in winter