Ten Kilowatts of Power: 1000W+ PSU Roundup

The manufacturers are getting ever higher with their PSUs, reaching the 1500W milestone already. Today we are going to introduce to you 10 units with the capacity of 1000W or well beyond that. We will discuss the solutions from Antec, Enermax, Floston Energetix, FSP Group, Silverstone, Tagan, Thermaltake and Ultra.

by Oleg Artamonov
01/18/2008 | 09:40 AM

In my last article about 1000W power supplies I wrote that a modern PC could be actually satisfied with much less power. A configuration with a quad-core CPU and a couple of top-end graphics cards would consume no more than 500W from the power supply. It is only hardcore overclockers that utilize Peltier elements or Freon compressors that may indeed require twice or thrice that amount of juice.


However, PSU manufacturers go on pumping up the capacity of their products and I now have the test results of models with a wattage rating up to 1500W. My job is to test, and it’s up to you to decide if your PC really needs that much power. I can only tell you one funny thing: the combined wattage of all the PSUs tested for this review exceeds 10 thousand watts!

Testing Methodology

Check out our article called X-bit Labs Presents: Power Supply Unit Testing Methodology In-Depth for a description of our testing methodology, the equipment we use, and a brief explanation of what the specified and tested parameters of power supplies mean. If you feel overwhelmed with the numbers and terms this article abounds in, check out an appropriate section of the mentioned Description for an explanation.

Antec TruePower Quattro TPQ-1000 (1000W)

The 1000W PSUs covered in my previous report were mostly cooled with 80mm fans and were quite noisy as the consequence. So I only expected quietness from models with 135mm fans.

Anyway, the 1000W model that opens up this review employs an 80mm fan again although the manufacturer promises quiet operation.

The PSU would be no different externally from any other regular PSU if it were not for its peculiar coloring. Using a pair of white stripes Antec has produced a hint of the racing car of recent past. The word Quattro in the model name provokes car-related associations as well but it means four +12V lines here.

There are five connectors for detachable cables on the rear panel. Note that the connectors are mechanically identical despite the different colors and labels. Each of them has keys (slanted corners) in the last two contacts of the top row and in the first one of the bottom row. In other words, nothing prevents you from plugging any cable into any connector.

Moreover, the electric wiring of the connectors is identical. Each of them has two ground wires, two +12V, one +5V and one +3.3V pin. I think that graphics card connectors should have been made differently, with three +12V pins and three ground pins; the other two voltages are not necessary for graphics cards altogether. This would make the PSU more reliable because it is the pins that are the weak point of every power cable (that’s why the number of pins is ever increasing along with the power consumption of PCs: graphics cards have moved from Molex connectors to 6- and now even 8-pin plugs, and mainboards have progressed from 20-pin to 20+4 and 24+4-pin connectors). So, this engineering solution seems questionable to me.

Most of the interior is occupied by well-ribbed heatsinks, which is logical considering that the TPQ-1000 is cooled with only one 80mm fan. Take note that the PSU is based on one power transformer (there are quite a lot of dual-transformer designs among competing products). It was fitted into the necessary dimensions by increasing the operating frequency of the PWM regulator.

The additional card on one of the heatsinks is not only a fan speed controller but also a circuit that limits the current in the “virtual” +12V lines. The card’s input receives +12V from the PSU’s rectifier and then this voltage goes to three resistors with a very low resistance. The PSU’s 12V cables are connected to the resistors at the other side. Thus, the voltage drop on the specific resistor is proportional to the load current on the cable attached to it. If this voltage reaches the threshold corresponding to 20A, the protection circuit wakes up to shut the PSU down. There are three resistors on the additional card, responsible for the 12V2, 12V3 and 12V4 lines. The 12V1 line is taken from the main PCB. This kind of separation – when there is actually one common +12V power rail inside the PSU but the output cables are divided into a few groups, each of which has a certain limitation of the maximum load current – is employed in a majority of PSUs for which multiple +12V output lines are declared. That’s why I call such lines “virtual”.

Four +12V output lines are declared for the TPQ-1000, actually. Each has a max load of 18A but the combined load is 70A rather than 72A (18*4). In other words, the load capacity of the common internal +12V power rail is 70A.

The PSU has the following cables and connectors:

Included with the PSU are:

All of the cables are sleeved. The sleeves are fastened by means of rubber rings at their ends.

The PSU worked normally together with an APC SmartUPS SC 620 at loads up to 379W (from the mains) and 340W (from the batteries). It had no problems switching to the batteries.

The output voltage ripple is very low on the +12V rail even at full load. It is no higher than 40 milliseconds, which is only one third of the permissible maximum. The +5V rail meets the requirements too, but without any reserve, while the pulsation on the +3.3V rail exceeds the permissible limit even though by just a few millivolts. The cause of this problem is the pulsation of a mysterious origin at a frequency of about 1.5 kHz. Usually, there are only two types of pulsation in a PSU: at the frequency of the PWM-regulator (tens of kilohertz) and at the double frequency of the power mains (100Hz in our region).

The PSU boasts superb stability of the output voltages none of which violates the limits or even reaches a 4% deflection from the nominal value (a 5% deflection being the permissible limit). The +12V voltage, which is loaded the most in the modern PC, is ideal deflecting no more than 1% from the nominal value!

The PSU employs an 80x80x25mm ADDA AD0812UB-A70GL fan with a rated speed of 3400rpm (at 12V voltage).

The fan speed is constant at 1600rpm until a load of 500W. Then it starts to grow, reaching over 2500rpm at the maximum. So, the PSU is not really silent but it is quiet in comparison with the models I tested in my previous review whose 80mm fans would easily speed up to 3500rpm and more. Alas, the tradeoff for the quietness is high temperature: the difference between the air temperatures at the input and output amounts to 20°C, which is quite a lot.

You may wonder why the fan speed could not be set lower for loads of 200-300W to make the PSU truly silent? The problem is that PSUs still use fans whose speed is regulated by the supply voltage rather than by an independent signal (like in the new CPU coolers with a 4-pin connector). With this regulation, the speed can only be reduced to 50% of the rated one or the fan may not start up at all. You can expect a wider regulation range when fans with a separate speed-control contact appear in PSUs.

This model was developed to comply with the 80 Plus requirements and it is indeed 80% and more efficient through the load range from 20% (200W) to maximum. Of course, the efficiency is going to be lower in an 110V power grid, but the reserve is large enough for the PSU to exceed the 80% barrier then, too.

So, the Antec TruePower Quattro TPQ-1000 is a very high-quality and high-wattage power supply featuring superb stability of the output voltages through the entire load range and a relatively quiet operation. The latter thing is, however, accompanied with a high temperature at high loads. Moreover, this PSU is still obviously noisier than lower-wattage models like the Antec NeoHE, for example. If you want to have a really quiet PSU, you should realistically estimate the power needs of your PC and choose a quieter, perhaps lower-wattage, model. But if you are set upon buying a 1000W power source, the TruePower Quattro is going to be a good choice.

Enermax Galaxy DXX EGX1000EWL (1000W)

As opposed to the above-discussed Antec TruePower Quattro, the PSU from Enermax is cooled with two fans. The cooling system is rather non-standard with one 135mm and one 80mm fan.

It’s not the two fans but the huge size of the PSU that you notice first, though. Its length is 220 millimeters (the length of the standard ATX power supply is 145mm). So, if you decide on buying a Galaxy DXX, make sure beforehand that it will fit into your system case. You’ll need 270-300mm of free space between the rear panel and the optical drives in order to install this PSU with all its cables and connectors.

There are eight connectors for detachable cables on the rear panel: six for peripherals and two for graphics cards. The connectors have differently positioned keys, making it impossible to plug a wrong cable in.

The PSU features a dual-transformer design. To remind you, this design is employed only to keep within the desired form-factor. In many cases it is easier to accommodate two smaller transformers, half the total wattage each, instead of one large transformer for the full wattage. The dual-transformer design has neither advantages nor drawbacks in comparison with the traditional single-transformer one for the end-user.

The PSU features active PFC and dedicated voltage regulation. Both have become standard features for this class of PSUs.

The PSU has five “virtual” +12V lines with a combined current of 75A. The load capacity of the +5V standby source is high. It is usually no higher than 3A but the Galaxy DXX can provide a current of 6A across this rail.

The PSU is equipped with the following cables and connectors:

Included with the PSU are the following cables:

This abundance of cables is impressive indeed. You instantly recall the picture from the PSU box that explains what and how many devices you can attach to it:

Sixteen CPU cores (4 by 4), five graphics cards, 24 hard disk drives. I guess the marketing folk from Enermax have gone too excited about this PSU as to imagine PC systems that just do not exist. That’s what I wrote about in my previous report concerning 1000W power supplies – such PSUs are not yet really necessary even for top-end gaming configurations that require much less power.

Anyway, 1000W PSUs do exist and my job is to test them using an artificial 1000W load.

The PSU worked with an APC SmartUPS SC 620 at loads up to 335W. The switching to the batteries was performed normally.

I can also tell you that the PSU wouldn’t start up unless there was a load of 90W or higher on its +12V rail.

The high-frequency ripple of the output voltages is low even at full load: there is almost no pulsation on the +5V rail altogether. On the other two rails the pulsation is far smaller than the permissible maximums.

Besides the high-frequency ripple, the Galaxy DDX has a pulsation with a frequency of about 1.5 kHz. Like with the above-discussed Antec TruePower Quattro, the cause of this pulsation is unclear. Well, even so, it’s all right with the +12V and +5V rails while the +3.3V barely but meets the requirements to the output voltage ripple, too.

The +12V and +3.3V voltages are very stable but the +5V voltage is worse, sinking below the nominal value as the load grows up and even violates the permissible limit. However, this violation occurs when the combined load on the +5V and +3.3V rails is over 100 watts, which is unrealistic for a modern PC, which puts the biggest load on the +12V rail. So, the voltage regulation is overall good in the Galaxy DXX.

The main fan is a 135mm RL4T B1352512MB-3M from Globe Fan. Although the impeller is made from translucent plastic, the fan is not highlighted.

The main fan is accompanied with a 80mm SuperRed CHA8012DB-OA(E) from Cheng Home Electronic. Two different and differently oriented fans located close to each other often produce a rather annoying noise of air flow but this effect is barely perceivable in the Galaxy DDX because the fans are at a big distance from each other thanks to the large size of the housing.

The fan speed is almost constant at loads up to 600W and then begins to grow up. Although the 80mm fan is rotating at a lower speed than in the above-discussed Antec TruePower Quattro, the total amount of noise produced by the Enermax PSU is higher due to the second fan but the internal temperature is much lower. The difference between the input and output air temperature is just over 10°C even at full load.

Alas, the PSU is not very efficient. It can hardly meet the 80 Plus requirements as its efficiency is below 80% even in the 220V power grid.

The Enermax Galaxy DXX is overall a high-quality power supply for people who need to connect very, very many devices. It offers much more cables and connectors than most other available PSUs. On the downside are its very large size, not-very-high efficiency, and a noise level, which is low for a 1000W PSU but not really quiet. I think the PSU would be much better if the developer managed to increase its efficiency to 85% and get rid of the auxiliary 80mm fan.

Floston Energetix ENFP-1050W (1050W)

I have tested Floston’s products before – that was the ENFP-750W model, and now I’ve got the senior model of the Energetix series with an output power of 1050W. Like the 750W unit, it comes in a neat trunk that can be later used for storing your tools, for example.

The ENFP-1050W has a black matte case like most other models of the same output power. It is longer than the standard ATX unit and is not actually meant for cramped system cases.

Most of its cables are detachable. Their connectors can be found on the rear panel: four for graphics cards and eight for other peripherals. It is impossible to plug a cable into a wrong connector.

The PSU follows the dual-transformer design that is often employed in today’s high-wattage models because a 1000W power transformer is either too large or demands much higher operating frequencies. The first thing would make it impossible to fit the transformer into the PSU considering that some 27 millimeters of space must be left above the electronics for a fan. The second thing would call for the use of different materials for magnetic paths, of expensive transistors, etc. Thus the use of two transformers, half the total output power each, is justifiable both technically and commercially.

Some manufacturers claim that this design ensures a more stable operation of the PSU, but generally it is not so. Single- and dual-transformer designs provide identical characteristics and the choice of the particular design is determined by engineering and technological reasons.

The PSU also features active PFC and dedicated voltage regulation. Both have become almost obligatory for PSUs of such a high wattage.

The card with the output connectors is placed near the rear panel of the case. It carries ceramic capacitors, which is good. Although their capacitance is low, they work normally at high frequencies, smoothing out the ripple at the PSU’s output effectively as opposed to electrolytic capacitors whose efficiency degenerates with the growth of the operating frequency.

The PSU has four +12V output lines: two of them have a max current of 20A as demanded by the EN-60950 safety standard but the other two have a max current of 36A. The reason behind that is clear: the EN-60950 standard is not obligatory while the more relaxed limitation gives the user more freedom in connecting the PSU to the consumers. The PSU can provide a total of 72A across all of its 12V lines. Note also the high allowable load of the standby 5V source: it is 6A, which is two times that of many other PSUs.

The PSU is equipped with the following cables and connectors:

Included with the PSU are:

All of the cables are sleeved.

When I set out to test the PSU, I found that it was only the 12V4 line that had a max current of 36A. The rest of the +12V lines, including the 12V3 one, had a limitation of 20A. There are seats for two shunts near the 12V3 output on the PCB but only one of them, for a current of 20A, is installed. I don’t view this as a drawback of the PSU, though.

The PSU coped with full load successfully.

The high-frequency pulsation at the PSU’s output was within acceptable limits at full load although nearly reached them on the +3.3V rail.

Besides, there is low-frequency ripple (at a double frequency of the mains, i.e. 100Hz) on two power rails. Although this ripple is within the permissible limits I still regard the very existence of it as a certain drawback.

The cross-load diagram looks good. The voltages remain within the allowable 5% deflection at any load distribution including a great load misbalance towards the 5V or 12V rail. The +12V voltage is higher than necessary at low loads but becomes normal as the load increases.

The PSU is cooled with a 135mm fan from Young Lin Tech (the DFB132512H model).

The fan speed is constant at low loads but grows up linearly starting from a load of 200W and reaches the ceiling at a load of about 800W. The PSU does not overheat even at full load but its speed regulation might be more effective: the fan could have a lower speed in the load range of 400-500W. With this regulation, the PSU is average in terms of noisiness.

The efficiency reaches 85% at one point of the graph only. Anyway, the PSU is quite efficient by today’s standards, over 80% through most of the load range. The power factor is generally over 0.95 as is typical of PSUs with active power factor correction.

Thus, the Floston Energetix ENFP-1050W is a good product overall. It is free from serious defects, delivers stable voltages, copes easily with full load, and produces a reasonable amount of noise (for a 1000W unit). The ENFP-1050W also has a lower retail price than many other models with detachable cables and such original packaging.

FSP Epsilon 1010 (1010W)

I introduced to you the new FSP Epsilon series that starts with an 800W model in an earlier review (the older series included models with a wattage of 600W and 700W). Alas, it didn’t impress me much, being noisy and providing unstable voltages. Let’s see if the senior model of the Epsilon series can change my opinion about it.

The main distinguishing feature of the FSP Epsilon series, externally at least, is the small dimensions of the housing which are exactly those of the standard ATX unit: 145 millimeters of length. It means this PSU can be fitted into a very cramped system case if you want to do that for some reason.

Unlike the two previous models, the Epsilon’s cables are not detachable, but FSP also offers the Everest series, which is based on the same platform but features detachable cables.

The internal design is the same as you can see in the latest products from FSP. It is very roomy inside: the developer avoided the need to use large heatsinks by installing a redundant amount of load-bearing semiconductor components (for example, there are parallel-connected pairs or quads of diode packs rather than single packs in the output rectifier). These are special features of FSP power supplies I have wrote about a number of times.

The PSU follows the already classic single-ended circuit design with one power transformer. It has active PFC but lacks dedicated voltage regulation.

Note that the heatsink with output diode packs (the rightmost in the photo above) is fastened to the back panel of the case. This panel gets very hot at work as a consequence, but you shouldn’t worry about that. It is only colder in other PSUs because has no thermal contact with the heatsinks.

The PSU has four “virtual” +12V output lines capable of providing a combined current of 75A (900W). The number 75 is given in small print on the label and the large text “985W” right below the 12V lines may be somewhat confusing. The label also shows which exactly line is connected to which cables. It is good when you suspect your system does not start up due to overload of a specific power rail. Unfortunately, many manufacturers forget to provide such a table not only on the label but even in the manual.

The output power of 1010W is given for an ambient air temperature of 25°C. For 40°C the PSU’s output power is not higher than 850W. According to ATX12V Power Supply Design Guide, the power supply must provide full output power at an ambient air temperature up to 50°C, which is true for the above-discussed Antec TruePower Quattro and Enermax Galaxy DXX. So, it is a question if the Epsilon 1010 should be viewed as a 1010W power supply at all. Perhaps its proper place is in the 850W category?

The PSU has the following cables and connectors:

All of the cables are sleeved. Cables of different types differ in the color of the sleeves.

The PSU worked normally at full load notwithstanding the modest size of its heatsinks.

Alas, the very first test shows that the PSU is not quite all right. The output voltage ripple at full load is slightly above the norm on every power rail. The pulsation is high on the +12V rail due to short-term spikes meaning that the manufacturer should have used better electrolytic capacitors that would filter the high-frequency noise more effectively.

By the way, the FSP Group website shows 140 millivolts for the 12V rail and 75 millivolts for the other rails as the allowable maximums in the PSU’s specs but I want to remind you that ATX12V Power Supply Design Guide suggests other maximums: 120 millivolts for the +12V rail and 50 millivolts for the other rails. So, the FSP Epsilon 1010 does not even formally comply with the industry standard.

The cross-load characteristics are no better: the +12V voltage is too high at minimum loads while the +5V voltage sags at high loads. The Epsilon 1010 doesn’t look good even in comparison with many PSUs with joint voltage regulation, let alone models with dedicated voltage regulation which are a majority of 1000W PSUs.

The PSU is cooled with a Protechnic Electric fan (120x120x25mm, MGA12012YB-O25). It is not highlighted despite the translucent impeller. The On/Off switch is the only shining detail in the whole PSU.

The fan speed is adjusted linearly depending on load, reaching 1500rpm at 350W. The fan accelerates to 2900rpm at full load, which is a very high – and noisy – speed for a 120mm impeller.

So, the Epsilon 1010 is noisier than average although it doesn’t heat up much: the temperature difference is no bigger than 10°C.

The efficiency is traditionally superb. Although it lowers towards full load, its maximum is as high as 87%. The Epsilon series and other models based on the same platform have proved their high efficiency many times already.

The previous three PSUs could be said to have minor drawbacks, but the FSP Epsilon 1010 can only be viewed as having very few advantages such as its small size. However, these do not make up for its shortcomings: noisy operation, poor voltage stability, and too-high output voltage ripple.

SilverPower GuardianX SP-1000E (1000W)

Here is a 1000W power supply with one 80mm fan. Who said this cooling design is obsolete?

The SP-1000E doesn't stand out from the outside: a massive matte black case without any decorations except for a Power LED. This exterior is quite typical of 1000W PSUs. Not long ago, the very number 1000W was enough to allure the customer.

Like the FSP Epsilon, the SP-1000E doesn’t have detachable cables. The actual manufacturer of this PSU is Enhance Electronics as is indicated by the UL certificate number on the label: E166947.

The internal design is interesting. The first thing to catch the observer’s eye is that the heatsinks are different. Each group of components has an individual heatsink. The heatsinks are rather simple with longitudinal ribs: the fan in this PSU drives the air lengthwise rather than across as in the PSUs with 120mm and 135mm fans.

Then, the small upright cards are a curious thing, too. They carry fully independent additional PWM controllers. This seems to be all right since many PSUs have dedicated voltage regulation but it is usually implemented by means of a saturated-core regulator that is not quite an independent device but can only work together with the PSU’s main regulator, receiving a pulsating, not constant, voltage from it. Here, we have two full-featured PWM-controllers working from a constant +12V voltage and providing +5V and +3.3V voltages at the output.

I can’t tell you why this solution was chosen – the simpler saturated-core regulator does its job well enough in other PSUs. Perhaps the developer had his reasons I don’t know about.

Otherwise, the PSU is just like any other model. It uses a single-transformer design (and the exterior of the transformer such as the complex shape of the core and the multiple-strand winding indicate that it is meant for high operating frequencies to fit into the required dimensions) and has active PFC.

The PSU has as many as six “virtual” +12V lines but it doesn’t mean that their combined load capacity is higher than that of the above-discussed models. Although each of the lines has a max current of 18A, and two of them even have a max current of 28A, they are all limited to 80A in total.

The PSU is equipped with the following cables and connectors:

All of the cables are sleeved.

The PSUs worked together with an APC SmartUPS SC 620 at loads up to 372W (from the mains) and 340W (from the batteries). At higher loads the UPS would report overload after switching to the batteries but continued to operate.

Alas, the output voltage ripple is near the permissible maximums on all the power rails, and short-terms spikes even exceed them.

The cross-load characteristics aren’t good, either. The +5V and +12V voltages sag heavily under load. That’s not a critical, yet unpleasant, situation.

The PSU is cooled by an 80x80x25mm ADDA AD0812UB-A70GL fan, the same as in the above-discussed Antec TruePower Quattro TPQ-1000.

But unlike with the Antec PSU, there is a linear dependence between the speed and load, without a flat stretch at low loads. Moreover, the fan is as fast as 3600rpm at full load. So, there is no silence here but the temperature of the output air is much lower than with the Antec. This PSU won’t die from overheat.

The efficiency is high at about 85% through most of the load range. It goes down somewhat at the end of the graph, yet still remains higher than 80%.

The result of the last test cannot change the overall impression, though. The Silver Power GuardianX SP-1000E is not an interesting product if compared with the available alternatives. It is noisy, its voltages are not very stable, and its output voltage ripple violates the permissible limits.

Tagan TG1300-U6 and Tagan ITZ1300 (1300W)

You may have been wondering so far where are those over-1000W units I promised you at the beginning of the article. 1010W and 1050W are no different from 1000W, really. Well, here they are – two 1300W power supplies from Tagan! The ITZ series PSU was provided to us by ABS Computer Technologies Inc.

Although the two models go under different names, I’ll be discussing them both together as I could find no fundamental difference in their parameters and internal design. As is usual for PSUs selling under the Tagan brand, these two are manufactured by Topower.

All difference between the PSUs seems to boil down to the exterior design: the ordinary TG1300-U6 has a plain matte black case whereas the ITZ is blue and covered all over with Certified by ABS Gaming Lab stickers.

The PSUs feature a dual-transformer design. As you can see, the transformers are very large even for half the total wattage each, which must be the reason for choosing that design solution. One 1300W transformer just wouldn’t fit into the same housing.

The PSUs have queerly shaped heatsinks – they are round and hollow inside. This doesn’t tell anything about cooling efficiency, though. I could name you a number of very quiet PSUs whose heatsinks are just simple molded bars.

The ITZ series model is identical inside except that it has a differently shaped heatsink on the diode bridge. Both PSUs feature active PFC and dedicated voltage regulation.

Providing a total output power of 1300W, the PSUs can yield up to 1104W (92A) via the +12V power rail divided into six virtual output lines. The load capacity of the standby source is 6A like that of the PSUs from Enermax and Floston.

The PSUs have the following cables and connectors:

Interestingly, if you want to connect more PATA drives than there are appropriate connectors, you can use four adapters from SATA connectors included with the PSU:

I think it is a very right solution. PSUs must support the most popular PC configurations (i.e. SATA drives) while backward compatibility should be maintained by means of adapters.

Well, let’s get to the tests now. Fortunately, our testbed has the same load capacity of the 12V channels as the TG1300-U6 supports (92A).

Together with an APC SmartUPS SC 620 the PSU worked normally at loads up to 385W (from the mains) and 270W (from the batteries). When the load was higher, the UPS would work for a few seconds and shut down reporting overload. The first number is indicative of a high efficiency of the PSU while the second tells that its active PFC is not quite compatible with UPSes.

The output voltage ripple is higher than the permissible maximum on the +5V rail of the TG1300-U6 (due to short-term spikes). The ITZ1300 has poor results on all the three rails. I’d say Tagan should pay more attention to the quality of the capacitors than to pretty-looking promo stickers.

It’s all right about the cross-load characteristics: the PSUs can work under any load remaining within the 5% voltage deflection. The main voltage, 12V, is almost ideal.

The PSUs are cooled with fans from Top Motor, an 80x80x25mm DF128025BU together with an 80x80x15mm DF128015BU. The consumption current of .45A means that these fans are not going to be quiet.

Rotating at a reasonable speed up to a load of 400W, the fans then quickly reached 3700-4200rpm. They do cool the PSU well – the temperature difference is no higher than 12°C – but the 1300W Tagan is not a quiet PSU. It produces an audible noise even at minimum load.

The PSUs have a high efficiency at medium loads but it worsens towards full load, dropping below 80%.

Thus, the Tagan TG1300-U6 and ITZ1300 are rather ambiguous products. On one hand, they feature neat assembly, a good selection of connectors and excellent stability of the output voltages, but on the other hand, their output voltage ripple is above the limit (especially that of the certified ITZ1300) and they produce too much noise. The wattage of 1300W is about two times as much as a modern PC with a quad-core CPU and a couple of graphics cards needs, so I don’t see any reason why you may want to sacrifice silence for the sake of wattage in this particular case.

Thermaltake Toughpower W0133RE (1200W) and W0171RE (1500W)

The wattage rating of the 1300W model discussed in the previous section was impressive enough, but now I’m going to talk about the absolute leader among all PSUs I have tested so far. It is the 1500W Toughpower W0171RE. In my imaginary rating list, it is slightly superior to my home vacuum-cleaner, yet still far inferior to the electric iron and electric kettle.

The 1500W model came along with a 1200W W0133RE. Both being based on the same platform, I will discuss them together.

These PSUs are almost as large as the Enermax Galaxy DXX, but have much higher wattage than the latter. The exterior is traditional for the latest generation of PSUs from Thermaltake: a dark-beige matte case, aluminum labels on the sides (which make it hard to dismantle the unit: there are two screws under the label, and aluminum is not a simple paper sticker “Warranty Void…”), and a punched-out but pretty-looking fan grid.

The rear panels of the PSUs offer four connectors for peripherals and as many as six for graphics cards. It is indicated what +12V lines power what connectors, and there are appropriate labels on the cables, too.

The photo of the internal design is quite an extraordinary view: each PSU consists of two halves each of which is a full-featured 750W power supply! I haven’t seen this kind of design before. Yes, there are PSUs with multiple regulators or even two power transformers, but here we have two PSUs, each with its own high-voltage section even, packed into a single housing!

One of these sub-PSUs is responsible for the +5V rail and three +12V lines. The other provides the +3.3V rail and three more +12V lines. Each of the sub-PSUs is absolutely independent. Each has its own active PFC device even.

The junior model can yield all its output power, 1200W, via the +12V power rail. To be more exact, there are two +12V power rails (distributed among the sub-PSUs as described above) with a load capacity of 600W each.

The senior model can provide 1440W via the +12V rail, which is slightly less than the full output power. The difference is small, though, especially for such big numbers.

The PSUs are equipped with the following cables and connectors:

Included with the PSUs are:

Thus, the senior model offers two more graphics card connectors (8 as opposed to the junior model’s 6) by means of the non-detachable cable.

I had to test the senior model under a load that was somewhat below its maximum because our testbed could only support PSUs with an output power of 1350W out of which 1100W could be taken from the +12V rail.

The PSUs worked together with an APC SmartUPS SC 620 at loads up to 375W (from the mains) and 340W (from the batteries). At higher loads the UPS would shut down some 5 seconds after switching to the batteries.

Alas, the output voltage ripple on the +3.3V rail exceeds the permissible limits. This voltage isn’t quite right in many high-wattage PSUs as you may have noticed. The +5V rail doesn’t meet the requirements as well if you note the individual spikes whose peaks are above the required 50 millivolts.

It’s almost the same with the 1500W model: the pulsation on the +3.3V rail violates the permissible limits while the pulsation on the +5V rail is as about as high as the limit.

But like with the above-discussed FSP Epsilon, the W0171RE specification at Thermaltake’s website says that the maximum allowable output voltage ripple is 240 millivolts for the 12V rail and 100 millivolts for the other rails, which is two times higher than the limits proposed in ATX12V Power Supply Design Guide. For the W0133RE model the maximum allowable limits comply with the standard. I don’t know if it is just a mistake or a deliberate attempt to “legalize” a power supply that does not comply with the industry standard.

Besides, there is low-frequency ripple, too. It is within the allowable limits, but I still regard the very existence of such pulsation as a drawback, even though not a crucial one.

It is the stability of the output voltages that is really a drawback of these PSUs. The +3.3V voltage sags heavily under load, not only when the +3.3V rail but also when the +12V rail, located in the same sub-PSU, is loaded. As a result, the maximum output power you can get from the PSU without exceeding the allowable 5% deflection is only some 950W. When the combined load on the +5V and +3.3V rails is 60W, the total output power of the PSU is only 600W.

It’s somewhat better with the senior model: the +3.3V voltage is still sagging, but the 5% border of the diagram has shifted to the right, the PSU providing over 1000W of power. On the other hand, there is no talking about 1500W here, and even these 1000W was obtained with voltages being on the verge of the allowable deflection from the nominal values.

The PSUs are cooled with 140x140x25mm Yate Loon D14BH-12 fans which have a rated speed of 2800rpm according to their manufacturer and 2300rpm according to Thermaltake.

The junior PSU model kept the fan speed constant until a load of 750W. However, that constant speed was not very low: the large 14cm fan was perfectly audible at 1300rpm.

The senior PSU model had a somewhat lower initial speed of the fan but started to increase it sooner: at a load of 600W. So, both PSUs are average in terms of noisiness and can stand a comparison with other high-wattage PSUs in this respect but they won’t suit people who want to have a silent PC.

Both PSUs have a normal efficiency, 80-83%. The power factor is over 0.95 as is typical of PSUs with active power factor correction.

The question I am always asking when testing such PSUs – what PC configuration might require that much power? – is complimented with the cross-load diagrams when it comes to the two PSUs from Thermaltake. Alas, none of them could provide its full output power in my tests just because the +3.3V voltage exceeded the permissible 5% deflection from the nominal value long before that moment. As a result, we cannot say too many really good things about the Toughpower W0133RE and W0177RE: high price, not very good stability of the output voltages, obviously overstated specified output power, rather noisy fans…

Ultra Products X3 ULT-HE1000X (1000W)

Having taken a look at models of a record-breaking wattage, I return now to a 1000W PSU again. This wattage has almost become ordinary already. So, it is a 1000W power supply from Ultra Products.

This PSU resembles the Floston ENFP-1050W I have discussed above, but it’s all right. Today, it is a common thing that identical products sell under different brands or different products sell under the same label. Judging by the UL certificate number, the PSU is manufactured by Andyson International, i.e. a third party for both Ultra and Floston. Let’s see if these two PSUs prove to be different in the tests.

The internal design confirms my supposition about the kinship between the PSUs: this one has two transformers, the same component layout, and same-shape heatsinks as the Floston unit.

That’s a dual-transformer PSU (as opposed to the PSUs from Thermaltake, the transformers are operating synchronously, governed by a common PWM controller) with active PFC and dedicated voltage regulation.

The single visible difference of the Floston unit is the selection of connectors: all of the Ultra PSU’s cables are detachable and there are six types of connectors for different cables. Here, I should express my comprehensions about the connectors again: the connectors for PATA drives (the top row) have open contacts, not separated from each other, and there will be a short circuit if any metallic thing is accidentally inserted into the connector.

The cables are flat and black, which looks very original. I described an Ultra ULT-XF500 power supply in an earlier review, which had similar cables, and complained that its cables were stiff and hard to lay out in the system case whereas the benefits of better ventilation were not so obvious. Fortunately, the cables of the ULT-HE1000X are softer, provoking no such problems. Still, I think that the material of the cables doesn’t have a big effect on the ventilation of the system case and it is an aesthetic rather than technical solution.


The total output power being 1000W, the PSU allows a load up to 70A (840W) on its +12V rail. The latter is not split into multiple output lines. To remind you, such splitting is implemented only to comply with the EN-60950 safety standard and has no effect on other technical properties of the PSU. Thus, the ULT-HE1000X is no worse in this respect than PSUs with four, five or six +12V lines. The non-split +12V power rail may even be viewed as an advantage as it gives the user full freedom in connecting the cables.

The PSU has the following connectors:

Included with the PSU are:

On one hand, this list is very long indeed. The manufacturer put a lot of cables of different lengths into the box so that you could pick up exactly what your system needs. There is even a short 18cm cable for an optical drive that usually resides opposite the PSU.

But on the other hand, there are only two connectors for SATA power cables and two types of such cables those that plug into the PSU’s SATA connectors and those that plug into the PSU’s PATA connectors. I think a 1000W power supply should be oriented at modern systems in which there are almost no devices with PATA power supply.

Comparing this PSU to the Floston ENFP-1050W again, the latter has universal connectors for power cables of peripheral devices, and the manufacturer can include as many cables of each type as deems proper at the given moment.

Together with an APC SmartUPS SC 620 this PSU worked normally at loads up to 370W (from the mains) and 310W (from the batteries). If the load was higher than these values, the UPS reported overload when switching to the batteries, yet went on working.

The output voltage ripple is overall normal, yet the individual spikes provoke some apprehensions.

Alas, there are also low-frequency pulsations at the PSU output, and the combined ripple on the +5V rail is slightly above the maximum permissible value.

The cross-load diagram looks good. Even though some portion of it is red, this red section falls on the range of very high loads, not only on the +12V rail but also on the +5V and +3.3V rails, which is not usual for a modern system.

The PSU employs a 135x135x25mm YLTC DFB132512H fan.

The speed of the fan is varied from 1000 to 1950rpm, making the PSU average in terms of noise. It does look good among 1000W model in this respect, especially in comparison with the PSUs from Thermaltake, let alone from FSP Group.

The efficiency is not as high as you might wish. It is barely above 80% at the maximum and drops to 77% towards the full load.

So, the Ultra Products ULT-HE1000X falls into the same category as the above-discussed PSU from Floston. It is a good mainstream model that provides reasonable characteristics and is free from obvious defects. Interestingly, this PSU behaved differently in my tests than the ENFP-1050W despite the external resemblance between them. It did not pass the output voltage ripple test but only slightly exceeded the permissible limit. The good thing about this PSU is that it comes with a broad selection of cables of various lengths, which facilitate the process of connecting the devices of a particular PC system.


The manufacturers are getting ever higher with their PSUs, reaching the 1500W milestone already. I still cannot grasp the purpose of such a large reserve of power – because even a very top-end gaming configuration needs only half as much at best – but 1000W power supplies have in fact become ordinary products while the manufacturers are rushing towards new goals.

Alas, not all of them reach the goals they set. There are four PSUs in this review (grouped in pairs according to the same platform) with wattage ratings much higher than 1000W, two from Tagan and two from Thermaltake, but none of them would be a very good purchase as they have high price, noisy fans, too-high output voltage ripple, and (the Thermaltake models) not very good stability of the output voltages.

The 1010W FSP Epsilon is no good, either. My earlier tests indicated that this platform was only good for 400-500W models and not higher. The Epsilon 1010 proves the point: high voltage ripple, a noisy fan, and very unstable voltages. All of this makes it impossible for me to write a positive review of it.

So if you do want a 1000W power supply, you should consider the models from Antec and Enermax as well as from Floston and Ultra Products, but I want to note that none of them has proved to be better overall than the Cooler Master Real Power Pro (tested in my previous review) that boasted stable operation at a surprisingly low (for such a high wattage) level of noise.