ATX Power Supply Units Roundup: 9 Powerful Models Tested

We are going to introduce to you 9 power supply units in our detailed roundup today. We will discuss in great detail the solutions from Enermax, FSP Group, MGE, SilverStone, SinTek, Zalman and Foxconn. These models are all very powerful ones with over 400W capacity that are available in retail stores. Find out who the winner is now!

by Oleg Artamonov
02/28/2006 | 10:46 PM

In this article I am going to test eight new power supplies that have one thing in common. They are all high-wattage boxed models, 400W and higher, that are meant to be sold in retail right into the hands of the end-users. To be exact, there are tests of nine units in this review because I retested the Foxconn WinFast FA-550A model covered in a previous article. The reasons will be explained below.

Enermax Noisetaker EG495AX-VE(W) (485W)


Lacking any decorations like LEDs, intricately-shaped fan grids, etc., the Noisetaker EG495AX-VE still looks pretty enough, mainly due to the coloring of the case. The photograph can’t convey it in full, but the PSU is not merely violet. Tiny spangles were added into the paint and they shimmer in light very beautifully.

The design of the PSU is standard otherwise. Well, Enermax is not counted among experimenting firms and prefers to focus on the electrical parameters of a PSU rather than on its outward appearance. Take note, however, of the fan speed control knob near the On/Off switch and the filter on the PSU’s output:

It is an ordinary ferrite ring inside a plastic casing. This filter doesn’t play any crucial part in minimizing the voltage ripple on the PSU’s output, yet it does suppress some high-frequency noise. You may have seen this solution already. For example, a ferrite ring is also put on the output cables of the FSP FSP460-60PFN unit. The graphics card cable of some power supplies from OCZ wears such a ring with a couple of capacitors, too.

The internal design is standard, following the classic circuit schematic with group voltage regulation and active power factor correction. There is nothing here I could find any fault with. Everything is neat and tidy.

This model has the typical characteristics of an ATX12V 2.0 power supply (this industry standard doesn’t describe models with an output power of over 400W, so you have to extrapolate). The +12V rail is “virtually” divided in two as is done in the absolute majority of new PC power supplies. In other words, there is only one power rail inside the power supply and its load current is up to 32A, but there are two separate current limiters, 18A each, set up on the PSU’s output. I already wrote in my reviews about the reasons why the sum of the max currents on the 12V1 and 12V2 rails is bigger than the total current on the +12V rail: this power rail is split in two only to comply with the EN-60950 safety regulation which states that the maximum output power on user-accessible contacts must not exceed 240VA (that is, the current must not be higher than 20A at 12V voltage; the protection is set at 18A for a small safety margin). Of course, it is logical to set the current limit higher for each of the 12V rails, i.e. at 18A, to provide more flexibility in load distribution across the PSU’s outputs.

This power supply offers you the following cables and connectors:

The +12V rails are wired in a rather odd fashion. According to the industry standard, the 12V2 rail is to feed the CPU only (through a 4-pin ATX12V connector) and the 12V1 feeds all the rest. In the Noisetaker, however, the 12V2 line also powers the mainboard (through the 24-pin connector) whereas the 12V1 is responsible for the rest of the PSU’s peripheral connectors, i.e. the connectors for hard, optical and floppy drives. The main practical consequence of this slightly different wiring may arise if there is an additional and powerful consumer on the 12V2 rail, a graphics card that is permitted to receive up to 75W of power from the PCI Express slot. The latter is located on the mainboard and is powered from the 12V2 rail in this case. As a result, the load on the 12V2 rail may get very close to the limit (18A) if you’ve got a powerful CPU and a PCI Express graphics card in your system, while the 12V1 line will bear but a small load.

The cross-load characteristic of this PSU looks superb. Only the +12V voltage violated the acceptable limits, but it did so at very unlikely load distributions (a very high load on the +5V rail and an almost-zero load on the +12V). In fact, the EG495AX-VE offers as stable voltages as PSUs with independent voltage regulation normally do.

Under full load (485W) the voltage ripple on the +5V, +12V and +3.3V rails was 15, 30 and 20 millivolts, respectively. The pulsation is high-frequency one, with neither a low-frequency constituent (at 100Hz) nor short spikes at the moments the inverter’s transistors are switched over.

The power supply is cooled with two fans: an 80mm one on the rear panel and a 90mm one on the top panel. Both the fans use ball bearings which gives a peculiar tone to the noise the PSU produces. I can’t call this Noisetaker absolutely silent, but it is very quiet indeed and most users are going to be satisfied with its noise characteristics.

A curious feature of this PSU is that the fans keep on working at their min speed after the PC has been shut down until the temperature is below 40°C. It usually takes about 10 minutes for the temperature to reach this mark.

The velocity sensor of the 80mm fan is placed outside and I used it to measure the rotation speed (the sensor gives out precisely two impulses per each rotation of the fan). Unfortunately, I couldn’t measure the speed of the 90mm fan as its velocity sensor is not outside the PSU while the transparent plastic of the blades makes it impossible to use an optical tachometer. Well, these measurements wouldn’t brings too much new info since the speed of the fans is adjusted simultaneously (in Enermax’s earlier power supplies the automatic adjustment only changed the speed of one fan and the other was controlled manually; here, the adjustment systems are joined together and affect both the fans at once). The graph below shows you the dependence of the fan speed on the PSU load for the two extreme positions of the manual control knob.

The objective measurements prove the PSU’s quietness at work, especially when the manual controller is set to the minimum position. Moreover, the speed of the fan doesn’t almost depend on the PSU load at loads below 250-280W.

The efficiency of this PSU is good at 80%. Regrettably, the design of our testbed doesn’t guarantee high efficiency measurement accuracy at high load power which you should be aware of when you compare these numbers to those you get from other sources, especially if those sources don’t employ high-precision measurement techniques, either. These numbers, however, suit fine for comparing this PSU with other PSUs tested on our own testbed, too.

Power factor correction is implemented well in the EG495AX-VE: the power factor is never below 0.9 and is even higher than 0.995 at maximum load (the program we use to process the results rounds the number up to two places behind the dot, so it just gave out the ideal coefficient of “1.00” for the first time in my tests!).

So, the Enermax Noisetaker EG495AX-VE(W) is a very good power supply all around and in its every single feature. It is a high-wattage, quiet and very neatly-made PSU with excellent characteristics. It can easily meet the demands of most PC users.

Enermax Liberty ELT620AWT (620W)

The second power supply from Enermax in this review belongs to the Liberty series which features detachable power cables. The steel case of the PSU is painted black with tiny spangles, just like the Noisetaker’s. Besides that, Enermax installs 12cm fans in PSUs of this series and the speed of the fans is controlled automatically – the manual adjustment knob is missing on the case.

The mainboard and CPU power cables are quite reasonably soldered into the power supply because they are used in any system and are also expected to drive high currents and each extra connector would worsen the stability of the voltages (the resistance of a connector pins is counted among the biggest hindrances that may occur on the path of electric current).

The rest of the cables are attached to the PSU via 6-pin color-coded connectors: the brown ones are for graphics cards (there are two such connectors here, so you can easily power up your SLI or CrossFire configuration without any adapters or splitters and the black ones for your peripherals (the connectors are universal and each has all the necessary voltages; that’s why you can attach both PATA and SATA devices to any of them).

There’s not much of a protection against wrong connection of the cables: you cannot plug a HDD cable into a graphics card connector, but you can do the opposite. The comprehensible color coding should help you, so you can only do it wrongly if you are attaching the cables blindly in the assembled computer. But even in this case there can be no harm – your trying to turn the PSU on will just trigger off the short-circuit protection.

This model is internally alike to the EG495AX-VE (see above), except that the components ratings and the size of the heatsinks are increased in proportion to the wattage. The PSU is assembled neatly. The output connectors are soldered on a single card fastened to a side panel. There are three non-detachable cables in this PSU:

The PSU comes with a few detachable cables neatly placed in this cute purse:

All in all, you will get a power supply with an amazing twenty (!) of connectors for your drives if you attach all the cables: 10 plugs for PATA and 10 plugs for SATA devices. And one socket will also be left unoccupied on the PSU.

Besides the cables, the PSU comes with two “Enermax” stickers (black and white) and a strap for wearing something on the neck. The latter can be used for your mobile phone or mp3 player but there’s nothing inside the PSU box that could be worn that way. This is the first time I see accessories supplied with a PSU that do not have any relation whatsoever to powering something.

Despite the considerable increase of the overall wattage of the unit over the EG495XA-VE (by nearly 30%), the allowable total load on the +12V rails has only increased from 32A to 36A. What’s strange, the declared max load on the 12V1 and 12V2 lines is 22A, although this violates the requirements of the EN-60950 standard the very division into 12V1 and 12V2 rails was implemented for. The allowable load on the low-voltage rails has even decreased, from 32A to 28A for the +3.3V. It doesn’t matter much for today’s computer systems, though.

The cross-load diagram for this power supply is nearly an accurate rectangle. There is no beveled corner in the right part of the diagram where the tested PSU usually reaches its maximum output power because with the Liberty ELT620AWT the sum of the maximum loads on each rail is a little bigger than the maximum PSU load overall. The voltages are so stable that they only go out of the required limits in the left part of the graph when there is a great load misbalance towards the +5V rail. Like the Noisetaker, this PSU ensures an excellent stability of the output voltages as you might have expected considering the similarity of their internal design.

I also checked the Liberty out for its compatibility with uninterruptible power supplies. Many users have reported that some PC power supplies with the full range of input voltages supported (i.e. the entire range from 90 to 265V without your having to manually switch between 115 and 230V) cannot normally work with UPSes. The moment the PC switches over to the UPS batteries, the overload protection in the power supply shuts the whole system down.

So, I checked the Liberty ELT620AWT with two UPSes: APC BackUPS CS350 and APC SmartUPS SC620. The maximum load power of the PSU was 200W and 175W as it worked from the electric mains and from the batteries, respectively, when the former UPS indicated overload. With the latter UPS, the numbers were 360W and 330W, respectively. No problems were observed at the moments the system switched over from the electric mains to the UPS batteries.

One 120mm ball-bearings fan is employed to cool this power supply. Unlike with the Noisetaker, there’s no manual speed controller in the Liberty and the speed is growing up linearly starting from a 150W load (the Noisetaker’s fan speed was constant until near 300W). I wouldn’t say the Liberty is louder than the Noisetaker, though. On one hand, the hiss of the air the fan is pumping through becomes perceptible at high loads, but on the other hand, the bearing of one 120mm fan produces much less noise than the bearings of two fans in the Noisetaker. Keep it in mind too that the load of 620W is excessive for a majority of today’s PC systems. The maximum power consumption of an average gaming computer varies within 200-250W and even SLI configurations feel well with a 400W PSU. In other words, there’s a rather small chance of the Liberty’s fan to speed up to its maximum in your particular computer.

Enermax declares for its Liberty series PSUs an efficiency of 80% in a load range of 30-100% of the maximum, i.e. from about 200W and higher in this case, and this is really so. The efficiency is 80% at loads of 150-160W and never goes below this level even though it diminishes somewhat at loads of 450W and higher.

So, the Enermax Liberty, like the above-described Noisetaker, leaves a very positive impression both subjectively and objectively, i.e. with its performance in tests. This is a top-quality, high-wattage and quiet power supply that can easily power up almost any PC configuration imaginable.

FSP Epsilon FX700-GLN (700W)

The Epsilon FX700-GLN is a top-end power supply model offered by FSP Group, but I would first like to give you a brief summary of the company’s current model range since I’ve been noticing on many forums that users find themselves confused in the names even though FSP uses convenient and logical letter-digit combinations in its product nomenclature.

What distinguishes FSP from other companies is that low-end models are not a separate branch of the company’s product development, but are renamed previous-generation models. We have already reviewed FSPxxx-60THN units on our site (“xxx” stands for the wattage of a particular model) that belonged to the first generation of ATX12V 2.0 power supplies. Besides the THN series, there exist THN-P, THA and THA-P series that differ from THN in the type of the fan and/or the availability of passive PFC. All these power supplies have an identical circuit design. The BlueStorm AX500-A model may also be viewed as belonging to the THN-P series. It could be marked as “FSP460-60THN-P” because it doesn’t differ in circuit design from the other THN-P series models.

With the arrival of new products the THN and THA series are moving now to the low-end sector where they acquire new names: ATX-xxxPN (previously FSPxxx-60THN), ATX-xxxPNF (previously FSPxxx-60THN-P), ATX-xxxPA (previously FSPxxx-60THA) and ATX-xxxPAF (previously FSPxxx-60THA-P). These “new” models are absolutely the same as the “old” ones – the change of the name only serves to indicate that they are low-end products now. Curiously enough, THA/THA-P series units were but seldom available in retail shops because their small 80mm fan meant increased noise, but they are now widely available under the new name of ATX-xxxPA/PAF.

There was already a remarking of this kind once when, following the release of first ATX12V 2.0 products, the older models of the FSPxxx-60BTV series were renamed as ATX-xxxGTF. And now the ATX-xxxGTF series leaves the production lines to be replaced with ATX-xxxPN/PNF/PA/PAF series units.

Some confusion results from FSP Group’s power supply nomenclature already including models with the PN suffix; marked as FSPxxx-60PN, these units comply with the transitional ATX12V 1.3 standard (they are listed in the table above for reference purposes). You should be aware that, though with similar-sounding suffixes, “ATX”- and “FSP”-marked models belong to different market segments and the suffixes have different meanings. Power supplies of the FSPxxx-60PN and FSPxxx-60PN(PF) series are currently leaving production. There’s no sense in making them anymore now that two generations of ATX12V 2.0 models are already selling.

The two models of the FSPxxx-60GNF series are fan-less units also known as FSP Zen.

The units of the FSPxxx-60PLN/PFN/PLG series stand somewhat apart from the others in the company’s nomenclature. Strictly speaking, they don’t make up any series; these are mostly single models with a single type of the fan and PFC and with a single output power rating. You should be very cautious when comparing these models by their names alone. For example, the FSP400-60PFN and the FSP460-60PFN not only differ in their rated wattage, but in fact belong to two quite different versions of the ATX12V standard. Their circuit design is absolutely different, too.

These units will be replaced by new models in the market, though. The new FSPxxx-60GLC/GLN/HLC/HLN and FSPxxx-80GLC/GLN series (this is where the Epsilon FX700-GLN belongs to) are power supplies with an absolutely new circuit design (as will be shown below) which are going to fully cover the midrange and top-end price sectors in the product assortment of FSP Group.

The units with the number 60 in the name are “home-oriented” ATX12V 2.0 models. Each of them comes in four flavors varying in the PFC design and fan size. Unlike with the older units, a PFC device is always present in the new ones (this is logical enough as the circuit design of a PSU with active PFC doesn’t permit to take this PFC out easily), but may support different ranges of input voltages: either the universal range of 90-265V (in the units with G as the first letter of the suffix) or the range of 176-265V (in the units with H as the first letter of the suffix). The power supplies may come with 80mm (“C” as the last letter in the suffix) or 120mm (denoted with the letter N) fans. I guess that power supplies with small fans won’t be too widespread in retail.

The units with the number “80” in the name are server-oriented EPS12V models, but this division is not too strict. There are no conspicuous differences between the two PSU standards except that EPS12V demands an 8-pin CPU power connector as opposed to the 4-pin connector described in ATX12V. Thus, these units may be considered as higher-wattage versions of the “60” units – at least, there are no fundamental differences in the circuit design. The 80 series units come not in four but in two versions differing in the type of the fan: FSPxxx-80GLC and FSPxxx-80GLN. The active PFC device supports the full range of input voltages in both the versions.

To summarize what I’ve written above: 1) the ATX-xxxGTF and FSPxxx-60PN(PF) series units complying with the older versions of the PSU standard are not produced anymore; 2) the FSPxxx-60THN/THA series models of the up-to-date ATX12V 2.0 standard are moved into the low-end sector and are renamed as ATX-xxxPN/PNF/PA/PAF (their circuit design remains unchanged); 3) the new models FSPxxx-60GLC/GLN/HLC/HLN and FSPxxx-80GLC/GLN fill in the mainstream and high-end sectors of the market.

The above-mentioned names of power supplies are only applicable to OEM models. Boxed PSU models from FSP Group as well as from its partner brands like Zalman are based on the OEM models and the parameters and sometimes even the name of a boxed PSU may indicate quite clearly what OEM model it is based on. Thus, the table above makes it clear that the Epsilon FX700-GLN is a retail version of the FSP700-80GLN power supply (700W wattage, active PFC with an input voltage range of 90 to 265V, and a 120mm fan).

However, it would be wrong to extrapolate all the data on some boxed or OEM version of a power supply on all the other versions (obviously, one and the same model FSP700-80GLN may be ordered from FSP by some other vendor and be produced under the vendor’s own brand). These versions would have the same circuit design and, accordingly, the basic parameters, but all the external features like the coloring, length of the cables, the number and type of the power connectors are up to the particular vendor. For example, the OEM model FSP600-80GLN is sold here, in Russia, in two boxed versions: as FSP Epsilon FX600-GLN and as FSP Optima Pro OPS600-80GLN (the latter is shipped into Russia exclusively, although is represented on the global FSP Group website). These two versions are absolutely identical inside but do differ a lot externally. The Optima Pro is cheaper than the Epsilon and has fewer and shorter cables.

As I have said above, the Epsilon FX700-GLN obviously corresponds to the OEM model FSP700-80GLN. It draws on the design tradition of the BlueStorm series in its appearance, but with certain differences. The color of the case is of a somewhat different hue and looks prettier, in my opinion. The fan has changed from blue to lusterless transparent, but still hasn’t acquired highlighting.

A glance at the internals of this power supplies usually provokes a cry of surprise from a specialist as the heatsinks look far too small for the declared wattage of 700W. The heatsink with the output diode packs and the cross metal bar that is pressed – and not wholly so – to the steel side panel of the case are the only heat-spreaders to have any ribbing at all. The heatsinks of the active PFC device (the leftmost one in the picture) and of the inverter’s transistors (in the middle) are bare aluminum bars.

By the way, take note of the text on the PCB that reads: “FSP700-80GLC Main Board”. I said above that the Epsilon is a GNL model, having a 12mm fan, but there’s no inconsistency here: the GLC and GLN series use the same PCBs but differ in the configuration of the heatsinks optimized for the particular type of the fan.

Running a little ahead, I’d like to tell you this power supply had no problems working under full load continuously. Before describing the PSU’s components, I’d want to give you a couple more snapshots of its internals to better illustrate its design features. Here’s a side view:

And here’s the PCB taken out of the case:

In most “ordinary” power supplies, high-voltage output rails use large diode packs in TO-247 packaging, but the first thing to draw my attention here is that the packs on the appropriate heatsink are all Fairchild’s YM3045N (MBRP3045N) in small TO-220 packages, yet there are as many as eight in total! Closer examination shows that the packs are connected in parallel in pairs, two of them belonging to the +5V line, two more to the +3.3V line, and the remaining four to the +12V. So, there are two diode packs, each with a max current of 30A, on the +5V channel whose allowable current is 30A, too. Why this double reserve? The fact is the temperature of a diode pack depends on two factors: the passing current and the voltage drop on it (Power = I*U), but these parameters are not independent because the voltage is higher when the current is higher. Thus, the amount of power dissipated by a diode pack grows up at a faster rate than the current passing through this pack, and two packs, each of which passes a current of 15A through, dissipate less heat than one 30A pack. Talking about the specific characteristics of the YM3045N packs, it is easy to calculate that the difference in heat dissipation between these two designs is more than 30% if the current is 30A.

Of course, two diode packs allow using a simpler and cheaper heatsink, but cost more by themselves than a single diode pack. Thus, the manufacturer always chooses between two options: they can put more expensive diodes and save on the heatsink or they can save on the diodes but put down a more expensive heatsink. The latter way is more common, but FSP’s engineers took the former approach. You might say it would be well to have both expensive diodes and a good heatsink all at once, but this would inevitably affect the end cost of the power supply. For example, FSP’s fan-less power supply called Zen uses the same YM3045N diodes and the heatsinks occupy almost the entire free space inside the case, too, but that PSU is priced appropriately high, even though it is one of the cheapest fan-less solutions available.

The engineers followed the same approach with the rest of the semiconductors. Take note of the tidy row of three transistors on the PFC heatsink or of the two diode bridges on both sides of the central heatsink.

So, it’s now clear with the diodes, but the smaller heatsinks are not only the consequence of the developer’s using more diode packs. Their size alone may improve the cooling efficiency. A smaller heatsink offers less resistance to the incoming stream of air. The speed of the air stream is higher then and its temperature gets lower. Of course, this effect is not very big, but you shouldn’t dismiss it altogether.

Another interesting feature of this power supply is the increased frequency of the PWM regulator. It is usually near 30kHz (this looks like fluctuations at 60kHz in oscillograms of a PSU’s output voltages since the push-pull inverter doubles the frequency). In the Epsilon this frequency is 55kHz or almost two times higher than usual. The higher PWM frequency permits to use smaller ferrites (coils and transformers) keeping the same maximum output power or, on the contrary, to increase the power while keeping their size the same. This explains the rather small (for a 700W power supply) dimensions of the power transformer and of the group regulation coil.

The main PWM regulator and the active PFC device are assembled on one chip, Champion Microelectronics CM6800G, which is installed on a small separate card to the right of the PFC coil and next to the central heatsink of the PSU. The standby voltage source is based on a Fairchild FSDM311 chip which is a full-featured PWM controller for low-power sources. And finally, there is a supervisor of the output voltages and currents on yet another small card installed near the spot where the output cables are soldered to the PSU (you can identify it in the snapshots by the long white connector on its edge). The PSU follows the traditional design with group voltage regulation.

The power supply offers you the following cables and connectors:

You can’t really complain that this is an insufficient selection of cables. Almost any modern configuration can be attached to the Epsilon without any adapters or splitters. The cables are all sleeved.

The declared characteristics are truly impressive. The FX700-GLN has four +12V lines with a current of 15A each, and the only limitation to their combined load is the overall wattage of the unit, i.e. 680W for the main rails. In other words, the manufacturer promises you can load this power supply through one 12V power rail only.

The PSU label also shows you how the power channels are wired to the output connectors: one channel powers up the first processor (meaning one 4-pin half of the 8-pin EPS12V connector; you will plug in only this half on your ATX12V system which only has one 4-pin power connector on the mainboard); the second channel powers one of the graphics cards and the second processor; the third channel is for the mainboard and all the peripherals; the fourth is for the second graphics card. It is easy to identify the connectors and tell which is for the first graphics card and which is for the second one because different channels are coded with strips of different colors on the wires (a black, blue or green strip goes against the yellow background of the 12V cable).

The division of the +12V power rail into four channels is traditionally virtual, i.e. there is actually one high-power 12V rail inside the PSU which is split in four by means of four shunts and an over-current protection circuit.

The cross-load diagram looks well. The only thing I can cavil at is the +5V voltage which goes down considerably at high loads. This is not going to be a big problem for a modern computer system where the consumption from the +5V and +3.3V rails is never higher than a few dozen of watts. Otherwise the cross-load diagram proves that the power supply can really yield its full power along one 12V channel only.

The PSU proved to work normally under a sustained load of 680W and the temperature of the heatsink with the diode packs, which is the hottest one, was not higher than 70°C at that.

The voltage ripple on the +5V, +12V and +3.3V rails was 42, 78 and 43 millivolts, the allowable maximums being 50, 120 and 50 millivolts, respectively. Most of this ripple is caused by short-term spikes at the moments the inverter’s transistors are switched over. Low-frequency pulsations (at the double mains frequency, which is 100Hz in our case) are missing altogether.

The Epsilon is cooled with a Protechnic Electric MGA12012HB-O25 fan whose speed control circuit is located on the main PCB (in previous PSU models from FSP it used to be placed on a separate small card fastened to a heatsink).

The fan speed depends linearly on the PSU load (on its temperature, to be exact) across all the measurement range. This PSU is among the quietest ones ever tested in our labs at loads lower than 300-400W and is definitely quieter than any earlier produced power supply from FSP, except for fan-less ones, of course. THN series PSUs with 120mm fans brought FSP the reputation of a manufacturer of quiet, but not exactly silent, PSUs. The GLN series is now among the best as concerns its noise characteristics.

When you turn the power supply on, the fan may be altogether idle for the first 2-3 minutes, but as the PSU gets hotter and the voltage on the fan rises, the blades begin to twitch a little (the fan’s electronics are trying to spin up the fan with such jerks) and then get to rotate constantly.

The manufacturer declares an efficiency of 85% and is honest about that, as you can see. The power factor is not so good, but the difference between the Epsilon’s 0.97-0.98 and many other PSUs’ 0.99 is really negligible.

So, the FSP Epsilon FX700-GLN power supply is a very well-made model and is a worthy replacement for the previous THN series. It is a very quiet and high-wattage unit that yields stable voltages and is well equipped with various connectors. It is also not very expensive at $180 which is not too much for a PSU with characteristics like these. The Epsilon series comprises only two models, FX600-GLN and FX700-GLN, but the GLN series of OEM power supplies includes a number of models from 250W and higher, so it is quite possible that more models, also in the entry-level sector, will be available to users in retail shops.

MGE Magnum 500 (400W)

Well, this power supply kept me surprised from the very beginning. The name of this fan-less-like model (alas, it only pretends to lack a fan!) is “Magnum 500”, which seems to imply a maximum output power of 500W. The PSU’s label, however, declares an output power of 400W, while its box tells you three different numbers: the wattage of this PSU is 500W at an ambient temperature of 25°C, 450W at 40°C and 300W at 50°C. Since Paragraph 5.1 of ATX12V Power Supply Design Guide states that the power supply must be operable under the maximum load within a temperature range of +10 to +50°C, I have to consider the Magnum 500 as a 300W power supply.

This PSU has an aluminum case with some ribbing on the sides. This proved to be a pure decoration, however, because none of the heating-up elements of the PSU has thermal contact with the case. A heatsink consisting of thin copper plates is placed outside and is covered under a grid aluminum casing.

The most unexpected element of the PSU’s rear panel is the large LCD display located right in the center of it. When working, it shows the power consumption level and temperature of the PSU, and the voltage on the +12V rail:

The practical value of the display is rather dubious as you will probably have to use a mirror to see it.

The connectors for the power cables are on the other side of the PSU (all the cables are detachable, including the mainboard one). There is another diagnostics element here – a row of five LEDs:

According to the user manual, these LEDs are to indicate what voltages are being used by the computer. But according to my own eyes, these LEDs are simply connected to the PSU’s output lines and report that there really is voltage on them. All the voltages are always present in a working power supply, so all the LEDs are always shining. That’s why the diagnostic value of this illumination is rather dubious, too.

There are rubber plugs for all the output connectors of the PSU – the three connectors for drives are closed with them in the snapshot above.

When open, the power supply exposes its superb thin-ribbed copper heatsinks (resembling CPU heatsinks in 1U servers) as well as an 80mm fan that is blowing at them. The fan is fully hidden in the case and cannot be seen from the outside. The LCD display card is on the right; the flat red-colored thermal sensor that sticks out from it measures the air temperature inside the PSU.

Two heat pipes are fastened on the smaller heatsink; they transfer heat on to the outside heat-spreader.

Well, if you remove the pretty-looking top parts of the heatsinks, the view is not at all handsome. You see dirty, slightly oxidized plates without a trace of thermal paste. The heat-transfer efficiency can hardly be high here.

Moreover, the efficiency of the fan itself is questionable. There’s an almost blank wall behind it, without any vent openings. The openings above the fan and heatsinks let a large portion of the air get out of the case immediately instead of making it cool the heatsinks. In the external panel of the case, on the contrary, there are very few vent openings. They are in fact only in the external heatsink, while most of that panel is occupied by the LCD display. Thus, the PSU’s cooling system hinders the movement of the air stream greatly. It could have been made more efficient, I guess.

This power supply comes with the following cables:

I can’t be silent about the way the cables are designed here. The idea to pack each cable into a separate pipe for easier laying-out inside the PC case has been developed to the point of ultimate absurdity in the Magnum 500. All the cables are double-screened (aluminum foil and wire braiding above it; this screening doesn’t have any big effect on the PSU’s parameters, by the way); a thick transparent plastic insulates everything on top. As a result, the cables are so stiff that it may become a serious problem to lay them out in a small-size PC case.

Moreover, the braiding sticks out a few millimeters from under the insulating pipe. I guess the PSU should have come with some insulating tape for the user to finish the cables off to a normal condition, but I couldn’t find anything like that in the PSU box.

And the last touch, the developers decided to use individual colors for each contact of a connector. In other PSUs, all +12V wires are colored yellow, +5V are red, “ground” are black, etc., but here you just can’t find two same-color wires. There are some twenty pins in the mainboard power connector, and wires of twenty different colors go to it. I can’t find any practical sense in such decorations and this may make it difficult to repair and diagnose the power supply for failures. This is also simply confusing: a CPU power connector with a red wire may perplex any advanced user because red denotes +5V in other PSUs and this voltage is not supplied to the CPU connector.

As mentioned above, the PSU’s label promises a maximum output power of 400W. The PSU complies with the ATX12V 2.0 standard, meaning that the load power on the +12V rail is two times higher than the load on the +5V and +3.3V rails.

The cross-load characteristic of the PSU is surprisingly good. It works without problems across the entire range of allowable loads. The +3.3V voltage is not very stable, yet it never violates the required limits.

At a load of 385W, the voltage ripple on the PSU’s output is 16 millivolts on the +5V rail, 15 millivolts on the +12V rail and 14 millivolts on the +3.3V rail. The pulsations are all high-frequency ones, without a 100Hz constituent.

It is next to impossible to measure the rotation speed of the fan in the Magnum 500, so the diagram above shows you the dependence of the fan voltage on the PSU load. The diagram shows that the voltage grows up linearly along with the load (along with the PSU temperature, to be exact). The noise from the fan is perceptible at loads higher than 300W.

The efficiency and power factors of this PSU are not very low, but are definitely worse than those of the Enermax and FSP models (see above). This PSU just fits the requirements of the standard (an efficiency of no less than 60% at a load of 100W and no less than 70% at a load of 200W and higher).

My overall impression about the Magnum 500 is rather negative. Yes, the technical parameters of this PSU are good s it yields stable voltages with a low voltage ripple, but some design solutions employed in it are rather questionable. It has two supposedly diagnostic devices (an LCD display and a row of LEDs), one of which is absolutely useless and another would have been better replaced with vent holes. Its ventilation was not well thought-out. The shabby-looking cables with their insanely excessive screening are not very easy to deal with, too.

MGE Vigor 500 (500W)

This power supply from MGE (do not confuse it with MGE UPS System, the renowned manufacturer of UPSes and line filters, which is quite another firm) has a more traditional appearance and doesn’t try to look like a fan-less PSU. Two cooling fans are installed inside the shiny chromium-plated steel case, 120mm and 80mm in diameter (this is rather an unusual design for a major brand, although MGE has used it before in a slightly different way). There is a fan speed control on the case and blue plastic windows in the side panels.

The actual manufacturer of this power supply is Wintech Electronics Corp.

The power supply is ordinary enough inside: the regulator is based on the KA7500B chip; the heatsinks are 5 millimeters thick; there are two 820µF capacitors on the input; and there’s only one section in the line filter instead of the required two.

The PSU offers the following cables and connectors:

The PSU comes with two PATA → SATA power adapters (one connector in each adapter).

The specified characteristics of the PSU make it clear that it complies with the ATX12V 1.2 standard, i.e. it is an out-dated product. Note also that the total load power cannot be higher than 486W (or 448W if you count in the three basic voltages only) according to the specified currents, although the manufacturer claims 500W.

The cross-load characteristic of this PSU is far from impressive. Too-high +5V voltages limit the diagram from below and too-high +12V voltages do the same from above. The narrow strip remaining can’t be considered normal, either, because the +5V voltage is 4% higher than normal.

At a load of 440W, the voltage ripple on the +5V rail was 47 millivolts (18 millivolts of high-frequency pulsation and the rest is 100Hz pulsation). The voltage ripple on the +12V rail was 65 millivolts (with high-frequency pulsation accounting for only 17 millivolts) and on the +3.3V rail, 32 millivolts.

As you see, the speed of the fans depends but slightly on the load and is mostly determined by the position of the manual control knob (there are two lines in each graph, for the two extreme positions of the controller). The PSU becomes rather hot under maximum load when the fan speed is set at the minimum. I wouldn’t recommend you to use it in this mode.

The PSU efficiency is good, but the power factor is far inferior to that of PSUs with active PFC; the Vigor lacks any PFC whatsoever.

All in all, the MGE Vigor 500 model can hardly be interesting for today’s users. It is compliant with an obsolete standard, has poor voltage stability, and offers inefficient automatic fan speed management. So, it just can’t stand a chance in a competition against the numerous and better opponents.

SilverStone Zeus ST65ZF (ENS-0565) (650W)

The PSU from SilverStone – this manufacturer makes a debut in our tests, by the way – is impressive with its very appearance. It is simply big. The standard ATX unit has a length of 140 millimeters, but this one, 180 millimeters.

This is also the single power supply in this review to be cooled with an 80mm fan:

The huge and massive heatsinks dominate the inside of the case, but the circuit design is rather standard: active PFC, group voltage regulation, and a +12V rail “virtually” split in four lines with varying allowable currents.

The components are placed densely, some on separate cards (you can see the voltages and currents supervisor card in the snapshot above). I have no complaints about the assembly quality – everything is very neat here.

The declared parameters of this model are typical of a high-wattage ATX12V 2.0 model. However, if you compare it with the FSP Epsilon, you will note the lack of an output power reserve on particular lines. In order to get all the 650 watts from this PSU you have to put the maximum load on all the available power rails, while the Epsilon gives enough freedom in re-distributing the load between the 5V and 12V lines even at the maximum output power.

The Zeus ST65ZF offers the following cables and connectors:

The cables are hidden in braided screens, except for the cables for hard disk drives.

The cross-load characteristic graph of the power supply looks good, but not quite impressive: the +12V voltage goes through all the range from the minimum to the maximum limit. The +5V and +3.3V voltages aren’t very stable, either. The PSU yields the declared output power, though.

The voltage ripple on the +5V, +12V and +3.3V rails was 35, 42 and 28 millivolts, respectively, under a load of 630W.

The Zeus ST65ZF uses an 80mm ADDA AD0812UB-A70GL fan. Its speed is varied under loads ranging from 200 to 450W. Outside this range, the fan speed is constant. The min speed is over 2500rpm, so the fan can’t be called quiet even at small loads on the PSU; the maximum speed is over 3600rpm.

The PSU efficiency is good at near 79%. The power factor is what you can expect from a regular power supply with active power factor correction.

The Zeus ST65ZF is positioned by the manufacturer as a power supply for workstations, although the term “workstation” has lost much of its original meaning nowadays. Some twenty years ago it used to denote high-performance PCs installed at work places and intended for graphics processing and other computation-intensive tasks, but ordinary home PCs have long acquired the same technical characteristics and the term “workstation” is now used by PC suppliers for marketing purposes mainly.

The Zeus has one significant drawback for home use (and probably for office use, too) – its 80mm fan is rather too noisy. The competitor companies (for example, Enermax and FSP) and SilverStone itself offer power supplies that work quietly but have the same or even higher wattage (see the FSP Epsilon above).

SinTek WIN550XSPX-X (550W)

This powers supply from an ominously named manufacturer bears a strong resemblance to the MGE Vortec models I have reviewed in my previous articles, see for instance the article called ATX Power Supply Units Roundup: Part II . It uses the same aluminum case (painted a different color and with aluminum wire instead of transparent windows) and the same placement of the fans, and there is the same knob to control the fan speed. A short investigation proved that the actual manufacturer of these PSUs is Wintech which has already been mentioned in this review (see the sections about the PSUs from MGE).

The PSU features detachable power cables for the peripherals; the mainboard and CPU cables are fixed. The different connectors are only labeled and not color-coded (as in the Enermax Liberty, for example), so you should be very careful when attaching the cables.

There are two control knobs near the connectors, labeled as “Memory” and “PCI Express”. I supposed that the former controls the +3.3V voltage, although the memory slots are powered from their own regulator on any modern mainboard and don’t care a bit about the voltage on the +3.3V rail (some enthusiasts do bind the memory voltage regulator to the input voltage as to an etalon, but this redesign requires some soldering skills and worsens the system stability and reliability).

The neighboring knob, “PCI Express”, should have controlled the +12V voltage, but I couldn’t spot any correlation between its position and the voltage on any of the PSU’s outputs.

The thermometer on the side of the PSU shows the temperature of the group regulation coil and the heatsink with diode packs (the thermal sensor is placed in between them). This thermometer will probably be obscured from view after installation unless your system case has a very large side window or there is no side panel in your case at all. Fortunately, the manufacturer made use of a functionally complete thermometer unit which can be easily taken out (you’ll need a cross-tipped screwdriver to dismantle the PSU and a flat-tipped screwdriver to release the plastic feet of the thermometer and to extract it) and placed on the front panel of the system case, for example. The thermometer is fed through a standard Molex connector (it can be seen in the snapshot of the internal design of the PSU, see below) and the thermal sensor has a long enough wire for you to place it anywhere in the system case. The thermometer lacks highlighting.

The internal design of the PSU is standard enough; this is a typical circuit design with group voltage regulation. The PSU doesn’t have any power factor correction whatsoever. The golden color of the heatsinks resembles Enermax, but this is the only common feature between the two companies: Wintech has nothing to do with Enermax as yet. The PSU is assembled neatly without visible flaws.

A maximum load of 34A is declared for the +12V rail and it is not divided into two 18A channels. As I’ve written in my reviews, this division is made only to comply with the EN-60950 safety regulation and does not affect the voltage stability or any other of the PSU’s characteristics (many manufacturers claim that two +12V channels ensure a higher load capacity or a higher stability of the voltages, but this is not true).

Curiously enough, the PSU label only mentions one input voltage (115V/60Hz). In fact, it has an ordinary 115/230V switch and, like all switching power supplies, is indifferent as to 50Hz or 60Hz is the mains frequency.

There are two stationary cables:

And these cables are also included with the PSU:

The cross-load characteristic of the PSU isn’t really good. All the three voltages exhibit but poor stability, the +12V even going through all the range from the minimum to the maximum. The PSU allows easily getting the declared power, but it is not good from the standpoint of modern, +12V-oriented PC systems that the whole cross-load characteristic is shifted up and leftwards into high +5V loads whereas in the bottom right part of the diagram (where there are high loads on the +12V rail) the voltages are out of the acceptable limits.

The voltage ripple under 450W load is 35 millivolts on the +5V rail (most of this ripple is 100Hz low-frequency pulsation, high-frequency pulsation accounting for only 10 millivolts); 62 millivolts on the +12V rail (high-frequency pulsation accounting for only 14 millivolts) and 30 millivolts on the +3.3V rail (high- and low-frequency pulsation each accounting for half that value).

The cooling system this PSU is equipped with is rather strange and I have only seen it in PSUs manufactured by Wintech: one 120mm fan and one smaller fan. In PSUs from other manufacturers there are either two small fans (80mm or 92mm) or a single 120mm fan.

The graph shows the speed of the two fans when the manual controller is set to the minimum position. As you see, the speed depends but slightly on the PSU load, like with the above-described MGE Vigor 500. When the manual controller is set for the maximum speed, the speed doesn’t depend on the load at all and is 3300rpm for the 80mm fan and 2030rpm for the 120mm fan. This speed management can hardly be considered right because the PSU may overheat under high loads when the speed is manually set to the minimum. At the maximum fan speed the PSU is too noisy even at low loads.

The efficiency is quite good at average loads, but degenerates considerably at high ones. The power factor is normal for a PFC-less model.

The SinTek WIN550XSPX-X is neatly assembled and has good parameters, yet I would want to have more stable voltages, especially on the +12V rail, and a more efficient fan speed management system that would adapt for different loads without the user’s intervention.

Zalman ZM460-APS (460W)

We have already reviewed power supplies from Zalman on our site, but those were only ATX12V 1.2 and 1.3 models with 80mm fans (which was quite a strange thing considering Zalman’s overall policy of creating quiet PC components). It is only recently that the company has released a new model, an ATX12V 2.0 power supply with a 120mm fan.

Despite the skilful disguise (even the PCB bears the name of Zalman), one can immediately see a product of FSP Group here – the FSP460-60GLN model with active PFC, a full range of input voltages supported, and a 120mm fan. Moreover, the unit’s PCB carries “checkboxes” for the full range of wattages of the GLN series, from 250 to 500W, while Zalman has only released one model.

The power supply is designed alike to the above-described FSP700-80GLN (Epsilon FX700-GLN), but the number of semiconductors has been reduced according to the wattage: there’s only one diode bridge left in the input rectifier; the PFC includes two transistors instead of three; there are six rather than eight diode packs on the PSU’s output. The heatsinks have become simpler, too. The aluminum plate that used to be pressed to a side of the PSU case is missing now. Otherwise, there are no changes: a CM6800C-based PWM controller, active PFC, group voltage regulation, and a standby source on the FSDM311 chip.

Contrary to the FX700-GLN, the maximum power load on the +12V rail is lower than the overall wattage of the PSU. This doesn’t prevent it from complying with the ATX12V 2.0 requirements, though.

This PSU is equipped with the following cables and connectors:

The cables are hidden in braided pipes. The PSU comes with a splitter from one PCI-E power connector to two for SLI configurations, an additional adapter from 1 PATA to 2 SATA power connectors, and a ZM-MC1 adapter to attach more fans.

The PSU easily passed our sustained full load test and the temperature of the heatsink with the diode packs only grew to 68°C (the similar temperature of the PSUs of different wattage is obviously the result of an efficient fan speed management system).

The cross-load diagram looks like the one of the FX700-GLN. The PSU easily works at full power. When there’s a great misbalance of load towards the +12V, the stability of the +12V and +3.3V voltages is good and of the +5V voltage, acceptable.

The voltage ripple at the maximum load was 20 millivolts on the +5V rail, 31 millivolts on the +12V rail and 28 millivolts on the +3.3V rail.

Another point of difference between the ZM460-APS and the power supplies that sell under the FSP brand is that a more expensive and, hopefully, more reliable and quiet fan NMB 4710KL-04W-B20 is installed in it instead of a fan from Protechnic Electric.

The fan speed is adjusted linearly, like in the “native” unit from FSP. The speed is not very high even at the maximum load. Under small loads the fan is near silent.

The efficiency is lower than that of the FX700-GLN, but it meets the specification: an efficiency of 85% was declared for the FX700-GLN whereas the specified efficiency of the ZM460-APS is 80% (the number is taken from the Zalman website; the FSP Group site declares an even lower number, “not less than 75% at full load”, for the FSP460-60GLN model).

Being much alike to the FSP Epsilon, the Zalman ZM460-APS is a very good home-oriented power supply. It features a highest assembly quality, good parameters, very quiet operation and an output power sufficient for a majority of computer systems, also those that include two graphics cards and a top-end central processor. If the Epsilon is too powerful and too expensive for you, the ZM460-APS from Zalman may be just what you need.

Foxconn WinFast FA-550A (550W)

We already tested the WinFast FA-550A power supply on our site and you can refer to our article called Power Supply Units from Foxconn and Hiper Group in Our Lab for details about its interior and exterior design.

The last time, however, the FA-550A had a very poor result in our cross-load characteristic test despite its independent voltage regulation. The +3.3V voltage quickly left the allowable range preventing the PSU from yielding the declared output power even.

That was rather strange for a PSU of that class, especially since other PSUs from Channel Well Technology (this company sells its produce under the Foxconn brand) did much better. I mean the FA-380A unit mentioned in the previous review and the Antec TruePower True430P model which is also manufactured by CWT.

That’s why I suspected I dealt with a defective sample of the PSU, and CWT has been kind to offer us another sample of the FA-550A model. This sample is absolutely the same as the earlier tested one in its specification and internal design, so I won’t describe it once again. I’ll just give you the cross-load diagram for this sample:

As you see, it is all much better now. The cross-load diagram fully covers the allowable range of output power for this PSU, just like it has to for a PSU with independent voltage regulation. Moreover, the diagram nearly coincides with the diagram for the above-mentioned Antec True430P: excellent stability of the +5V voltage (which almost remains constant) and good stability of the +3.3V and +12V voltages.

The voltage ripple at full load was near 20 milliseconds on the +5V rail (it was only high-frequency ripple without a 100Hz constituent), near 30 millivolts on the +3.3V rail (high- and low-frequency alike), and near 45 millivolts on the +12V rail (both low-frequency and high-frequency pulsations).

So, I can remove all the blame from the Foxconn WinFast FA-550A power supply, but of course I can’t say how frequently you can meet a sample with poor +3.3V stability. However, considering the results of the tests of a WinFast FA-380A and an Antec Tru430P, based on the same components, I’m inclined to think this is rather a rare problem.


Without any doubt, the best products in this review are the power supplies from Enermax, FSP Group and Zalman. Top-quality and easy-to-use, these power supplies can meet the requirements of almost any modern computer and be quiet at that. It’s hard to choose among these models as each of them can satisfy the majority of users.

The PSUs from SinTek and MEG are, on the contrary, a bit of a disappointment, each having some or other problems with the real parameters or the manufacturing quality, or offering some questionable design solutions (like the very thick, untidy and unhandy cables included with the MGE Magnum). So, these units are somewhat inferior to the leaders, yet they can satisfy many users, too, with the exception of the out-dated model Vigor 500. It’s a curious fact that both the companies that openly declare their American origin (you can find this web address on the MGE power supplies, while SinTek is said to be the leading U.S. manufacturer of power supplies on the home page of their own site) buy their power supplies from the non-American Wintech.

And finally, here is the power supply from SilverStone, a high-quality and high-wattage but very noisy model. This PSU would do well in a server system I guess. The Zeus might compete with the server models from FSP (FSP460-60PFN and FSP550-60PLN) which have similar noise characteristics, but the newer quiet models from Enermax and FSP leave it no chance as PSUs for home or office use.


Click here to download the cross-load characteristics of the tested PSUs

Click here to download the viewer of the cross-load data