Global Win Sapphire SAF450 Review

The box

When I first heard of the “SAF” prefix on the model name, I thought it was simply a shortened form of “Sapphire.” The box, however tells a different story. It appears the it actually means Super Silent, Active PFC, Fan Control. The bottom of the box is divided into two halves. The back half explains how to work out the “effective current”, “invalid current” and “total power”. The front half tells us that the unit has opposite dual fans, auto ranging APFC, and auto thermal fan control and what power cord it comes with. The table on the right says that the 5v rail can deliver more than double what the 12v rail can, suggesting that this is an outdated design. PCs for the last several years all draw more current from the 12v rail.

_

 

_

Ahh, more features. Let’s go through them:

Front:
– ATP (Asynchronous Thermal Protection)
1. 3 more minutes cooling after system off
2. Prevent temperatures from keeping rising after system off, like your CPU, chipsets, VGA, Micro IC… (Components don’t generate heat when they are off, but it’s still a nice feature)
3. Prolong the life circle of your computer and no suffering form invisible damage caused by overheat. (The life circle?)

– Super quiet AUTO fan control function
1. AUTO fan control keeps the noise level below 22.5dBA at 80% load of power supply (You know guys, you don’t have to yell the word “auto” at us)
2. Special thermal controlled case fan connectors to minimize noise fron case fan(s)

Back:
1. Gold plated connector pin (Which pin?)
2. Auto thermal fan control
3. Electric-shock-free protection (I want to be protected from getting electric shocks, not from being free of them)
4. Scan-disc-free protection (PS_off # >2mS)
5. Over voltage protection
6. Over current protection
7. Short circuit protection
8. Stable regulation for +5V, +3.3V, +12V at ± 5% (What about the −12V and 5Vsb?)
9. Auto ranging Active PFC (PF Value 0.96~0.989)
10. Titanium plated metal case enclosure
11. Compatibility test passed for dual AMD K8 Opteron 246 CPU
12. Interference reduction of EMI

  • It is effective to lower down EMI between 350Hz~440Hz by titanium-plated power housing
  • All DC output cables are twisted for the reduction of EMI

13. PEAK LOAD 100W more for 60sec (I’m getting really tired of being yelled at now)
14. Warranty: 3 years (Not once I’m done with it)

The PSU

Global Win is not exactly a mainstream PSU manufacturer that you’re likely to find at your local retailer. In fact, in all my years, this is the first time I’ve ever heard of one.

_

 

The label says that this PSU is both a Sapphire 19dBA and an Onyx 18dBA. So which one is it? Well, since the box and that small sticker just to the right both say it’s a Sapphire 19dBA, I’ll assume it’s that one. In addition to having a more powerful 5v rail, the unit actually has a −5v rail and only has a 20pin ATX connector, making it seem more dated now, since the need for a −5v rail was removed form ATX specifications years ago. It has a total of 6 fan-only connectors (3 peripheral connectors and 3 motherboard connectors) which are temperature controlled. While it is a nice feature, it is preferable to have the case fan speed adjusted based on the ambient temperature of the case, not the power supply temperature. One other noteworthy thing about the label is that it also mentions TOP-450 as the model, suggesting that the real manufacturer could have been Topower. Looking up the UL number (E130843) in the UL certifications database confirms this. The mirrored finish (which shows a very clear reflection of the carpet surface where the photograph was taken)  and the blue-tinted windows give it a very nice look, but looks can be deceiving, so let’s get on the testing and see how it performs.

Load Testing

Test 1 (118.19W Load – Cold)

Rail Load Voltage Ripple
12V 4.83A 12.08V 43.8mV
5V 5.02A 5.02V 36.3mV
3.3V 10.24A 3.38V 46.9mV
−12V 0A −12.44V 18.7mV
5Vsb 0A 4.99V 34.4mV
AC Power 145.6
Efficiency 81.17%
Power Factor 0.96

 

Test 2 (205.97W Load – Cold)

Rail Load Voltage Ripple
12V 9.61A 12.01V 52.5mV
5V 10.02A 5.01V 43.8mV
3.3V 10.18A 3.36V 63.8mV
−12V 0.1A −12.54V 41.2mV
5Vsb 0.99A 4.93V 36.9mV
AC Power 248.5W
Efficiency 82.89%
Power Factor 0.97

 

Test 3 (258.30W Load – Cold)

Rail Load Voltage Ripple
12V 9.76A 11.97V 49.4mV
5V 19.88A 5.12V 47.5mV
3.3V 10.18A 3.36V 58.1mV
−12V 0.11A −12.63V 42.5mV
5Vsb 0.99A 4.94V 36.3mV
AC Power 309.2W
Efficiency 83.54%
Power Factor 0.97

 

Test 4 (314.87W Load – Cold)

Rail Load Voltage Ripple
12V 14.48A 12.07V 56.3mV
5V 19.96A 4.99V 48.1mV
3.3V 10.18A 3.36V 50.0mV
−12V 0.11A −12.89V 59.4mV
5Vsb 0.99A 4.96V 38.8mV
AC Power 379.2W
Efficiency 83.03%
Power Factor 0.97

 

Test 5 (342.67W Load – Cold)

Rail Load Voltage Ripple
12V 14.6A 12.17V 51.3mV
5V 24.85A 4.97V 53.1mV
3.3V 10.18A 3.36V 66.9mV
−12V 0.11A −12.88V 49.4mV
5Vsb 0.98A 4.92V 35.6mV
AC Power 412.5W
Efficiency 82.83%
Power Factor 0.98

 

Test 6 (394.24W Load – Hot)

Rail Load Voltage Ripple
12V 19.14A 11.96V 40.0mV
5V 25A 5V 36.3mV
3.3V 10.15A 3.35V 51.3mV
−12V 0.11A 13.15V 34.4mV
5Vsb 0.99A 4.96V 27.5mV
AC Power 476.0W
Efficiency 82.82%
Power Factor 0.98

 

Test 7 (446.74W Load – Hot)

Rail Load Voltage Ripple
12V 23.68A 11.84V 55.6mV
5V 25.1A 5.02V 26.9mV
3.3V 10.15A 3.35V 66.9mV
−12V 0.11A −13.17V 35.6mV
5Vsb 0.99A 4.95V 28.1mV
AC Power 557W
Efficiency 80.21%
Power Factor 0.98

 

Overload Test (499.84W Load – Hot)

Rail Load Voltage Ripple
12V 28.17A 11.78V 72.5mV
5V 25.15A 5.03V 40mV
3.3V 10.15A 3.35V 68.1mV
−12V 0.11A −13.16V 64.4mV
5Vsb 0.98A 4.92V 47.5mV
AC Power 626.2W
Efficiency 79.82%
Power Factor 0.99

 

At first, it appeared that this unit was going to pass with flying colours, based on the voltage regulation. Not only were the voltages in spec, but they were in spec by a large margin. For the most part, they were within 2% of their nominal values, other than the −12v rail, where the ATX specification allows for it to be off by a bit more. The unit also remained practically silent, even at high loads. But then we got to the ripple testing, and that’s where things got a bit more ugly. The ripple was out of spec on the 3.3v rail for all of the tests except for tests 1 and 4, but even then, it was only just in spec. The 5v was a little high at times as well. It looks like we either have poor ripple suppression capabilities or bad capacitors. Personally, I’m hoping it’s just the latter.

 

Rail Test 7 (446.74W) Overload Test (499.84W)
12V Ripple on the 12V rail at 446.74W load Ripple on the 12V rail at 499.84W load
5V Ripple on the 5V Rail at 446.74W Load Ripple on the 5v rail at 499.84W load
3.3V Ripple on the 3.3v rail at 446.74W load Ripple on the 3.3V rail at 499.84W load
−12V Ripple on the -12V rail at 446.74W load Ripple on the -12V rail at 499.84W load
5Vsb Ripple on the 5Vsb rail at 446.74W load Ripple on the 5Vsb rail at 499.84W load

 

A look Inside

_

 

It looks like Topower did an alright job here. It appears well built, although the more I look at this unit, the more dated it seems, since it uses an old school half-bridge design. Starting at the input filtering, the SAF450 has two differential-mode chokes, one common-mode choke, two X capacitors and five Y capacitors (including the one after the rectifier). The two primary capacitors are 1000μF parts made by Koshin. 1000μF is a lot for a 450 Watter, so the hold-up time is excellent. In fact, it actually lasts several seconds with no power coming from the mains when under no load or very light load. The two primary switching transistors are Unisonic Technologies Co (UTC) 2SC3320s rated for 15A at 25°C, although Unisonic does not specify what they can handle at higher temperatures. Another impressive feature is the size of those heat sinks, although they need to be big to allow the PSU to remain as quiet as it did without overheating.

The APFC components and the bridge rectifier are located on a smaller PCB (see left picture below) which is screwed onto the two main heat sinks. The two PFC transistors are International Rectifier IRFP460A MOSFETs rated at 20A at 25°C and 13A at 100°C each, with an on resistance of 0.27Ω. The bridge rectifier is rated at 10A, so if the other components could handle it, the PSU could pull up to 1.2kW from a 120V mains supply and 2.4kW from a 240V mains supply. Assuming 80% efficiency, this would equate to maximum output wattages of 960W and 1.92kW respectively.

_

 

The 12V rail is fed by two NXP Semiconductors BYV32E-200 ultra-fast recovery rectifiers rated at 20A each. While 40A is enough current, it would be preferable to use Schottky rectifiers rather than fast recoveries, since they have a lower voltage drop and thus offer better efficiency. The 5v rail uses two MOSPEC S30D40C Schottky rectifiers rated at 30A each and the 3.3v rail uses a single MOSPEC S20C40C Schottky rectifier rated at 20A. Since the 3.3v rail claims to be able to deliver 28A, a 30A or higher rated part really should have been used here. Fuhjyyu capacitors are used throughout the secondary side. Why anyone would build a good unit and then let it down with such low quality capacitors is beyond me. If they are starting to go bad in this unit, it could explain why the ripple on the 3.3V rail was too high.

 

The fans are both Power Cooler brand. The odd thing is, that the rear one (with the Top-silent sticker) is actually a ball bearing, unlike the front one which is only a sleeve. While sleeve bearing fans are not generally as reliable as ball bearing ones, it probably won’t be too much of a problem here given how slow they run.

Specifications

Real Wattage 500W
OEM Topower
PFC Active
Price Unknown, no longer available
ATX Connector type 20 Pin
CPU Connector ATX12V (4 Pin)
PCI-E Connectors None
Molex (Peripheral) Connectors 6
FDD Power connectors 2
SATA Power connectors 2
Fan only connectors 6 (3× 3 pin and 3× Molex)

 

Conclusions

Pros: Active PFC, Efficient, Good voltage regulation, Very quiet, Looks awesome, Good hold-up time

Cons: Low quality Fuhjyyu capacitors

Bottom line: It would have been an excellent unit, if only it wasn’t for those capacitors.

Score: 7/10

Discuss this review in our forums here!

Review sample source: Purchased second hand