[nextpage title=”Introduction”]
This 650 W power supply from StarTech.com looks like a very high-end unit, as it is bigger than traditional power supplies and uses a dual-transformer design, feature we’ve only seen on power supplies on the 1,000 W range like Enermax Galaxy 1000 W and Tagan TurboJet TG1100-U95 1,100 W. It is also cheaper than competing products from better known brands. Is it a good product? That is what we are going to find out.
Figure 1: StarTech.com WattSmart 650 W power supply.
Figure 2: StarTech.com WattSmart 650 W power supply.
As mentioned, this power supply is bigger than regular power supplies, being 7 31/64” (19 cm) deep, while regular power supplies are 5 33/64” (140 mm) deep. So the first thing you need to check before buying this unit is if it fits your case. This shouldn’t be a problem if you are using a good case.
This power supply uses two 80 mm fan, one on the front side and the other on the rear side. We prefer power supplies with big 120 mm or 140 mm on the bottom, as this configuration provides a better airflow and lower noise level, but maybe this kind of fan wouldn’t fit this power supply because of the size of the heatsinks inside the unit.
Like all high-end units today, this power supply features active PFC and high efficiency (at least 80%, according to the manufacturer). Active PFC provides a better utilization of the power grid and allowing the manufacturer to sell this product in Europe (you can read more about PFC on our Power Supply Tutorial).
The higher the efficiency the better – an 80% efficiency means that 80% of the power pulled from the power grid will be converted in power on the power supply outputs and only 20% will be wasted. This translates into less consumption from the power grid (as less power needs to be pulled in order to generate the same amount of power on its outputs), meaning lower electricity bills – compare to less than 70% on regular power supplies.
This power supply comes with six peripheral power cables: two auxiliary power cables for video cards with 6-pin connectors, two cables containing three standard peripheral power connectors and one floppy disk drive power connector each and two cables containing three SATA power connectors each.
The number of connectors is good enough even for high-end users with a big RAID array and two video cards in SLI or CrossFire configuration.
The main motherboard cable uses a 20/24-pin connector and this power supply comes with both ATX12V and EPS12V connectors.
On the aesthetic side the manufacturer made a great job. All wires have a nylon sleeving coming from inside the power supply housing and going all the way to the end of the cable using a top notch rubber to make the finishing, as you can see in Figure 3.
Figure 3: Nylon sleevings go all the way to the end of the cables.
All wires used on this power supply are 18 AWG, which is correct for a power supply from this power range.
This power supply is manufactured by ATNG, which also manufactures power supplies sold in the USA as Rosewill.
Now let’s take a look inside this power supply.
[nextpage title=”A Look Inside The WattSmart 650 W”]
We decided to disassemble this power supply to see what it looks like inside, how it is designed, and what components are used. Please read our Anatomy of Switching Power Supplies tutorial to understand how a power supply works and to compare this power supply to others.
In this page, we will have an overall look, while in the next page we will discuss in details the quality and rating of the components used.
We can point out several differences between this power supply and a low-end (a.k.a. “generic”) one: the construction quality of the printed circuit board (PCB); the use of more components on the transient filtering stage; the active PFC circuitry; the power rating of all components; the design; etcetera.
As we said before, this power supply uses a dual-transformer design, and you can clearly see the two transformers on the pictures below.
[nextpage title=”Transient Filtering Stage”]
As we have mentioned in other articles and reviews, the first place we look when opening a power supply for a hint about its quality, is its filtering stage. The recommended components for this stage are two ferrite coils, two ceramic capacitors (Y capacitors, usually blue), one metalized polyester capacitor (X capacitor), and one MOV (Metal-Oxide Varistor). Very low-end power supplies use fewer components than that, usually removing the MOV, which is essential for cutting spikes coming from the power grid, and the first coil.
On this section this power supply is flawless, as it has more components than the necessary – one extra X capacitor, two extra Y capacitors and a ferrite bead on the main AC cable. It also provides a traditional fuse holder for its fuse, a feature hard to see nowadays, as most manufacturers solder the fuse directly to the printed circuit board.
Figure 7: Transient filtering stage (part 1).
Figure 8: Transient filtering stage (part 2).
In the next page we will have a more detailed discussion about the components used in the Sta
rTech.com WattSmart 650 W.
[nextpage title=”Primary Analysis”]
We were very curious to check what components were chosen for the power section of this power supply and also how they were set together, i.e., the design used. We were willing to see if the components could really deliver the power announced by StarTech.com.
From all the specs provided on the databook of each component, we are more interested on the maximum continuous current parameter, given in ampères or amps for short. To find the maximum theoretical power capacity of the component in watts we need just to use the formula P = V x I, where P is power in watts, V is the voltage in volts and I is the current in ampères.
We also need to know under which temperature the component manufacturer measured the component maximum current (this piece of information is also found on the component databook). The higher the temperature, the lower current semiconductors can deliver. Currents given at temperatures lower than 50° C are no good, as temperatures below that don’t reflect the power supply real working conditions.
Keep in mind that this doesn’t mean that the power supply will deliver the maximum current rated for each component as the maximum power the power supply can deliver depends on other components used – like the transformer, coils, the PCB layout, the wire gauge and even the width of the printed circuit board traces – not only on the specs of the main components we are going to analyze.
For a better understanding of what we are talking here, please read our Anatomy of Switching Power Supplies tutorial.
This power supply uses a KBU10J rectifying bridge with a heatsink attached to it. This bridge can deliver up to 10 A at 75° C. This is more than adequate rating for a 650 W power supply. The reason why is that at 115 V this unit would be able to pull up to 1,150 W from the power grid; assuming 80% efficiency, the bridge would allow this unit to deliver up to 920 W without burning this component. Of course we are only talking about this component and the real limit will depend on all other components from the power supply.
The manufacturer, however, scratched all other main components, making it almost impossible to identify them. This is the first time we’ve see such thing.
This power supply uses two power MOSFET transistors on its active PFC circuit and two other power MOSFET transistors on the switching section, which uses the traditional two-transistor forward configuration. From our experience we guess this power supply uses four 20N60C3 power MOSFET transistors, which are able to deliver up to 45 A at 25° C or 20 A at 110° C in continuous mode, or up to 300 A at 25° C in pulse mode. These transistors and the active PFC diode are located on the same heatsink.
Figure 10: Active PFC and switching transistors.
The integrated circuit in charge of controlling the active PFC and PWM circuits was scratched as well, but we believe that it is a CM6800, as this is the most popular integrated circuit for these functions.
Figure 11: Active PFC and PWM combo controller.
[nextpage title=”Secondary Analysis”]
This power supply uses four Schottky rectifiers on its secondary, but they were scratched as well, so we couldn’t identify them.
Figure 12: Two of the four Schottky rectifiers used on the secondary.
Figure 13: The other two rectifiers used on the secondary.
We, however, were more curious to find out what configuration was used by the two transformers. There are several ways to use two transformers on a PC power supply. On both of them all main outputs (+12 V, +5 V and +3.3 V) have independent transformer outputs and independent rectifiers.
And here we have one of the advantages of the dual-transformer architecture: the +5 V and +3.3 V outputs are completely independent. On power supplies using just one transformer, the +3.3 V and +5 V outputs share the same output from the transformer, even when they have independent rectifiers, so the maximum current these outputs can pull at the same time will depend on the transformer. With two transformers they are not only completely independent; they are obtained from different transformers: +5 V output comes from the first transformer while +3.3 V output comes from the second transformer.
The two main differences between the available dual-transformer architectures are how the primary and the secondary are obtained. On the primary side, the power supply can have just one switching section driving both transformers (which is cheaper) or one switching section for each transformer (which is better but more expensive). On the secondary side, the power supply can simply join the outputs of the two +12 V rectifiers, using just one filtering section (which is cheaper) or the power supply can have two independent +12 V outputs, each one with its own filtering stage (which is better but more expensive).
Usually power supplies with separated secondaries will also have separated switchers, while power supplies that join the +12 V outputs will also join the switchers in order to cut costs.
The advantage of the more expensive implementation is that you have two really independent +12 V power supplies, each one feeding different components. This implementation provides a higher current (and thus power) limit and also a “cleaner” output, i.e., less noise. This happens because noise produced at one +12 V rail won’t be propagated to the other rail.
In Figure 14 we show the difference between these two designs. StarTech.com WattSmart 650 W uses the cheapest approach. Another power supply that uses the same design is Tagan TurboJet TG1100-U95 1,100 W. Enermax Galaxy 1000 W uses the more expensive design.
Figure 14: The two different ways to implement two transformers.
This power supply thermal sensor is located on the secondary heatsink, as you can see in Figure 15. This sensor is used to control the fan speed according to the power supply internal temperature.
On this power supply all the electrolytic capacitors are manufactured by Teapo (a Taiwanese company). The active PFC capacitor is rated at 85° C, while the electrolytic capacitors from the secondary are rated at 105° C.
[nextpage title=”Power Distribution”]
This power supply has the following specs:
- +3.3 V: 24 A
- +5 V: 30 A
- +12V1: 18 A
- +12V2: 18 A
- +12V3: 18 A
- +12V4: 18 A
- -12 V: 0.5 A
- +5VSB: 3 A
As you can see this power supply has four virtual rails. Inside the power supply these rails are connected to the same place but each one uses an independent over current protection (OCP) circuit, usually set at a value a little bit higher than what is printed on the label.
This power supply +12V virtual rails are distributed like this:
- +12V1 (yellow with black stripe wire): ATX12V and EPS12V connectors.
- +12V2 (solid yellow wire): PCI Express auxiliary connector.
- +12V3 (yellow with blue stripe wire): PCI Express auxiliary connector.
- +12V4 (yellow with green stripe wire): Main motherboard cable and all peripheral cables.
We think this distribution is simply perfect, as the cables that really demand power (video cards and CPU – i.e., EPS12V/ATX12V) are independent and aren’t shared with the main motherboard cable or with any peripheral cable.
Now let’s see if this power supply can really deliver 650 W of power.
[nextpage title=”Load Tests”]
We conducted several tests with this power supply, as described in the article Hardware Secrets Power Supply Test Methodology. All the tests described below were taken with a room temperature between 44° C and 48° C. During our tests the power supply temperature was between 49° C and 53° C.
First we tested this power supply with five different load patterns, trying to pull around 20%, 40%, 60%, 80%, and 100% of its labeled maximum capacity (actual percentage used listed under “% Max Load”), watching how the reviewed unit behaved under each load. In the table below we list the load patterns we used and the results for each load.
+12V2 is the second +12V input of our load tester and on this test it was connected to the power supply EPS12V connector.
If you add all the power listed for each test, you may find a different value than what is posted under “Total” below. Since each output can vary slightly (e.g., the +5 V output working at +5.10 V), the actual total amount of power being delivered is slightly different than the calculated value. On the “Total” row we are using the real amount of power being delivered, as measured by our load tester.
Input | Test 1 | Test 2 | Test 3 | Test 4 | Test 5 |
+12V1 | 5 A (60 W) | 10 A (120 W) | 14.5 A (174 W) | 20 A (240 W) | 28 A (336 W) |
+12V2 | 4.5 A (54 W) | 9.5 A (114 W) | 14 A (168 W) | 18.5 A (222 W) | 20 A (240 W) |
+5V | 1 A (5 W) | 2 A (10 W) | 4 A (20 W) | 5 A (25 W) | 6 A (30 W) |
+3.3 V | 1 A (3.3 W) | 2 A (6.6 W) | 4 A (13.2 W) | 5 A (16.5 W) | 6 A (19.8 W) |
+5VSB | 1 A (5 W) | 1.5 A (7.5 W) | 2 A (10 W) | 2.5 A (12.5) | 3 A (15 W) |
-12 V | 0.5 A (6 W) | 0.5 A (6 W) | 0.5 A (6 W) | 0.5 A (6 W) | 0.5 A (6 W) |
Total | 135.3 W | 266.3 W | 394 W | 522.9 W | 644.9 W |
% Max Load | 20.8% | 41.0% | 60.6% | 80.4% | 99.2% |
Result | Pass | Pass | Pass | Pass | Pass |
Voltage Regulation | Pass | Pass | Pass | Pass | Pass |
Ripple and Noise | Pass | Pass | Pass | Pass | Pass |
AC Power | 162 W | 309 W | 459 W | 621 W | 785 W |
Efficiency | 83.5% | 86.2% | 85.8% | 84.2% | 82.2% |
StarTech.com could deliver its rated power, which is great. Voltage regulation and efficiency were the best features of this unit. All voltages were really stable, being inside a 3% limit from the nominal voltage in all our tests – which is really great, as the limit is 5% –, except the +3.3 V output during test number one, which was at +3.41 V – yet inside the 5% limit.
We were really impressed by this power supply efficiency, always above 82% peaking 86.2% at test number two.
However the main problem with WattSmart 650 W is the level of electrical noise it generates. Even though it was inside ATX specs (i.e., below 120 mV for +12 V outputs and below 50 mV for +5 V and +3.3 V outputs) it was almost reproved on test number five, where noise level on +12V2 peaked 115.4 mV. On all tests noise level on +12 V inputs from our load tester was very high, as you can see in the table below. Noise level on +5 V and +3.3 V inputs was fine.
Input | Test 1 | Test 2 | Test 3 | Test 4 | Test 5 |
+12V1 | 53.2 mV | 68.8 mV | 80.4 mV | 90.2 mV | 109 mV |
+12V2 | 55.4 mV | 72 mV | 83.8 mV | 96 mV | 115.4 mV |
+5V | 15 mV | 19 mV | 20.2 mV | 22.8 mV | 24 mV |
+3.3 V | 11.8 mV | 17.2 mV | 14.2 mV | 14.2 mV | 17 mV |
Below you can see screenshots from our oscilloscope for test number five.
Figure 16: Noise level at +12V1 input of the load tester.
Figure 17: Noise level at +12V2 input of the load tester.
Figure 18: Noise level at +5V input of the load tester.
Figure 19: Noise level at +3.3V input of the load tester.
[nextpage title=”Overload Tests”]
After these tests we tried to pull even more power from StarTech.com WattSmart 650 W, but the main problem was noise. Pulling 650 W we were already too close to the 120 mV limit set by the ATX standard and by pulling just a little bit more noise surpassed this value.
So we have two results for maximum power. The first, shown below, is with the power supply working within ATX specs, i.e., with noise level below 120 mV at +12 V (during this test room temperature was 50.7° C and the power supply housing was at 48.8° C). Then the second result is the maximum power we could pull but with noise level outside specs.
Input | Maximum |
+12V1 | 30 A (360 W) |
+12V2 | 20 A (240 W) |
+5V | 6 A (30 W) |
+3.3 V | 6 A (19.8 W) |
+5VSB | 3 A (15 W) |
-12 V | 0.5 A (6 W) |
Total | 666 W |
% Max Load | 102.7% |
AC Power | 816 W |
Efficiency | 81.6% |
With the power supply running under this configuration noise level 115 mV at +12V1 and 120 mV at +12V.
Figure 20: Noise level at +12V1 input of the load tester.
Figure 21: Noise level at +12V2 input of the load tester.
The problem was that we could pull more power from this unit, but the noise level was above the maximum admissible and a user doesn’t have a way to know that this is happening. We could pull up to 793 W with this power supply by pulling 33 A from +12V1 and 27 A from +12V2, but under this scenario we saw a 149.4 mV noise at +12V1 and 152 mV at +12V2. We were pulling 995 W from the wall, so efficiency was 79.7%. Room temperature was at 50° C and the power supply housing was at 49° C.
We tried to pull even more power from the unit, but it wouldn’t turn on – showing us the over power protection (OPP) in action, which is great: it allows us to go over above the power supply limit but not high enough to the point where we would burn it.
On this unit over current protection (OCP) is disabled or is set at a value over 33 A. We made a simple test here, we set +12V1 at 3 A and then increased +12V2 to 33 A, and the power supply would work just fine. Since the label states a maximum current of 18 A per rail the power supply should not allow this, as we were pulling 33 A from the power supply +12V1 virtual rail (don’t get confused here, +12V1 and +12V2 mentioned above are the name of the inputs located on our load tester), since we connected the EPS12V connector from the power supply to the +12V2 input from our load tester and kept the ATX12V connector disconnected from the tester.
Short circuit protection (SCP) worked fine for both +5 V and +12 V lines.
During our tests we could see the speed of the power supply fans changing as the power supply temperature increased. Below 30° C it spun slowly, making almost no noise, and after this temperature it started increasing its speed, which also increased noise level.
We were impressed by the cooling system used by this power supply. Its two 80 mm fans were able to keep the power supply housing temperature at room temperature or one degree Celsius below the temperature inside our hot box. Usually during our tests the temperature of the power supply housing is between 2° C to 5° C above the temperature inside our hot box.
[nextpage title=”Main Specifications”]
StarTech.com WattSmart 650 W power supply specs include:
- ATX12V 2.2
- Nominal labeled power: 650 W.
- Measured maximum power: 666 W at 50.7° C (or 793 W at 50° C with +12 V noise level outside specs).
- Labeled efficiency: 80% minimum
- Measured efficiency: Between 82.2% and 86.2% at 115 V.
- Active PFC: Yes.
- Motherboard Connectors: One 20/24-pin connector, one ATX12V connector and one EPS12V connector.
- Peripheral Connectors: two auxiliary power cables for video cards with 6-pin connectors, two cables containing three standard peripheral power connectors and one floppy disk drive power connector each and two cables containing three SATA power connectors each.
- Protections: Information not available. During our tests we could test that this power supply has over load (OLP) and short-circuit (SCP) protections working. Over current protection (OCP) test failed.
- Warranty: 3 years.
- Real manufacturer: ATNG
- More Information: https://www.startech.com
- Average price in the US*: USD 135.00.
* Researched at Shopping.com on the day we published this review.
[nextpage title=”Conclusions”]
StarTech.com WattSmart 650 W can really deliver 650 W at a room temperature of 50° C. This is really good.
The number of cables it has – two 6-pin PCI Express auxiliary connectors for video cards, six SATA power connectors and six peripheral standard power connectors – makes it a good product even to high-end users building a PC with a RAID array, two or more optical units and two video cards in SLI or CrossFire configuration.
Another good thing about this power supply was its efficiency, between 82.2% and 86.2% depending on the load.
Its cooling system showed to be very efficient, keeping the power supply at the same temperature level found inside our hot box (usually during our tests the power supply temperature is between 2° C to 5° C higher than the temperature inside our hot box).
But the main problem with this unit is the level of electrical noise. When pulling 650 W it practically touched the 120 mV limit for the +12 V outputs. By pulling just a little bit more than 650 W from this unit noise level at +12 V outputs surpassed the 120 mV limit, and the problem is that users don’t have a way to know this without reading a review like ours telling them that this happens.
On the other hand, this power supply won’t explode if you try to pull more power that it can handle.
At the end the dual-transformer design didn’t help this power supply. We’ve seen other power supplies in the 600-700 W range that have only one transformer and are able to deliver its labeled power at 50° C and have a far lower noise level.
Unfortunately because of its very high noise level we can’t give this product an award seal.
Leave a Reply