by Oleg Artamonov
05/02/2007 | 08:46 PM
Traditionally, the reaching of a next round number in the specification of any PC device is considered a great achievement. The first 1GHz processor, the first 1GB memory module, the first 1TB hard disk. Alas, this is not the same with power supplies. While the memory amount and the CPU frequency have their own practical worth because they allow a faster and better execution of ever more sophisticated computing tasks, the PSU is only needed to service the other system components. It cannot do anything on its own. Thus, the release of 1000W power supplies does not mean your games will be running faster or anything, but only that the air exhausted from your system case will be warmer now.
There is one rule that describes the difference between the goal and the means in a succinct and clear way: you don’t want a hammer, but you want the nail in the wall. The hammer is just a means to get the nail into the wall, acceptable as far as there is no other, better, means. Yet the goal is the nail rather than the process of hammering it in. A 10kg hammer wouldn’t be an achievement of the hammer maker. It would only be a trouble if you tried to bang nails in with it.
The same goes for power supplies. A high-wattage PSU is only a means rather than a goal. It is the means to power the PC up. So, a 1000W PSU should not be considered an achievement. It is rather an inevitable evil – nails have become so big these days that you do need a 10kg hammer to hang a picture on the wall.
A natural question, are such high-wattage PSUs really necessary? If yes, in what particular conditions? To answer that question we measured power consumption of two very advanced PC configurations.
AMD-based system was configured like follows:
We installed Windows XP SP2 on these systems and ran Stress Prime 2004 / Orthos for the CPU and 3DMark 2006 for the graphics card; these two programs were running simultaneously in the third test mode. Here are the PSU power consumption numbers (using a Tagan TurboJet TG1100-U96; we measured its power draw from the wall outlet and multiplied the result by this PSU’s efficiency factor, about 0.83):
What do we see? A powerful gaming system with an overclocked CPU and running two applications that put a heavy load on the CPU and both graphics cards simultaneously consumes less than 450W from the power supply! Even the dual-core 4x4 platform from AMD, which is not even a viable solution for home use but rather a concept so as to respond somehow to Intel’s quad-core CPUs, has a peak power draw of just over 600W!
So, it is all clear here: a power supply with wattage higher than 600-700W is not necessary for a powerful gaming system unless you are into such things as Peltier coolers, etc.
However, we’ve got more frightening numbers when measured power consumption of the systems in idle mode. Even the Core 2 Quad platform consumed as much as 300W at that! Alas, while the manufacturers of CPUs have taken to reducing power consumption of their products, the makers of chipsets and graphics cards have, on the contrary, relaxed. One GeForce 8800 GTX consumes as much as 70W when idle and we’ve got two of them in our system. 140 watts for the graphics subsystem alone that is displaying an empty Windows Desktop – that’s too much, to my mind.
And finally, to measure power consumption of a typical home PC we assembled a system with a Core 2 Duo X6800 (running at its default clock rate) and with one GeForce 8800 GTX. This PC was easily powered by a 420W PSU from HEC Group and our measurements showed that it consumed only 280W under 3DMark06. This is not a big number, considering that 300W PSUs are currently regarded as low-wattage products.
It is a widespread opinion that you should buy a PSU with higher wattage than you actually need as a kind of reserve for the future. This opinion is questionable. To begin with, such a PSU may prove much noisier at work than models with lower wattage. And second, when you do need its full wattage, if ever, the selection of connectors may have changed once again, and you’ll have to use a lot of adapters in your PSU. I can remind you the recent innovations: a 24-pin mainboard power connector, 8-pin power connectors for CPUs, SATA power connectors, 6-pin connectors for graphics cards. Yes, you can power all these components with an older PSU via adapters, but adapters are no good in terms of reliability and the ease of assembly.
Anyway, now that the manufacturers have got to producing 1000W and higher power supplies, our job is to test them and report the results to you. And it’s up to you to decide if you need such a PSU at all.
To load the PSUs fully we had to revise our testbed and install larger heatsinks and more powerful fans (Jamicon, 120x120x38mm, 2800rpm) on the load transistors of the +12V channels:
In the bottom part of the photograph you can see a small (80x80x25mm) fan on the heatsink placed on the load for the +5V and +3.3V channels (up to 250W combined). At the top there are two giants capable of loading the +12V power rail of any PSU up to 880W.
Before getting to the PSUs proper, I want to tell you about the new connector that is meant to replace or add to the customary 6-pin graphics card connector:
This is a 12V power connector as before, but with two more pins. It resembles the 8-pin CPU connector installed on some mainboards, but is not compatible with it electrically as well as mechanically (in other words, you cannot plug one into another). Anyway, the PSU manufacturers often label this connector as “PCI-E” to avoid confusion.
The first thing striking your eye in this PSU is not some design feature, but a yellow-and-black sticker that warns you against plugging the PSU into wall sockets rated for a current below 10A (or below 15A for the 110V power grid). I guess the old joke that you’ll need a separate wire from the power distribution unit for the computer, just like as for an electric oven, is getting closer to the truth now. Of course, these requirements to the wall outlet are only applicable when your computer indeed consumes 1000W from the PSU. Otherwise, the Real Power Pro can be safely plugged into wall sockets rated for lower currents.
Although this PSU doesn’t look large, it is actually longer than a standard ATX unit by 40 millimeters. The appearance proves misleading due to the 135mm fan. If there was a 120mm fan inside, the PSU would look larger. The length of 180 millimeters is not a record, though. For example, the PSU from PC Power & Cooling is 230mm long, having the same wattage.
The components are rather densely packed inside. You can see two absolutely identical power transformers here. The PSU implements a rather original solution when the transformers are working in sync, each of them bearing half the total load. I think the main reason for this solution is that a 1000W power transformer is too large and wouldn’t fit into the PSU. The maximum height of the transformer is limited by the need to leave some room at the top for a fan. Two transformers, each with half the necessary capacity, take more length, but less width and height.
The heatsinks are designed in an original way: their long narrow ribs have a square section instead of being thin and rectangular. I guess this helps to distribute heat more uniformly along the entire length of the ribs. You can see black rings on two ribs: these are fastenings of the thermal resistors of the fan speed management system.
The declared output power of 1000W is continuous rather than a peak one, and the PSU can yield most of it via its +12V power rail. The latter has six output lines with a current limiter of 18A on each. This division is “virtual”, meaning that there is only one +12V power rail inside the PSU and it is split at the output into two lines (there are two 40A filtering chokes instead of a single 80A one) and then into six lines by means of limiting the allowable load current.
Interesting in this PSU is the high allowable load on the +5V standby source, up to 3.5A.
Judging by the UL certificate number (the third one from the left in the bottom row of symbols on the PSU label), the actual manufacturer of this PSU is Enhance Electronics, a well-known firm with a good reputation. To remind you, Cooler Master’s PSUs of lower wattages are produced by AcBel, but the latter probably does not produce 1000W units.
The PSU offers the following cables and connectors:
So, this PSU can be connected without adapters to any home PC configuration you can think of, including configurations with two GeForce 8800 GTX and a RAID array consisting of a dozen hard disk drives.
The cross-load characteristics of this PSU look good, except that the +12V voltage is set higher than necessary by about 3%. The +5V voltage is stable throughout the entire load range while the +3.3V voltage deflects much from the nominal value only when there is a very high load on both the +3.3V and +12V rails, which is a virtually impossible situation for a modern computer.
The output voltage ripple on the +5V and +12V rails is 30 and 65 millivolts, respectively. This is within the normal range.
The PSU employs a Young Lin Tech Co. DFS132512H fan. Its size of 135x135x25mm is larger than the common 120x120x25mm.
The fan speed management is implemented perfectly: the speed is kept at 830rpm at loads below 500W, which makes the PSU very quiet. The speed is growing up at higher loads, but only to 1500rpm at the max output power. This is indeed good. And the temperature of the exhausted air is only 10°C higher than that of the incoming air.
I also want to note that this diagram coincides with the one you can see at the Cooler Master website. Many manufacturers (even renowned brands as you will see shortly) prefer to specify the fan speed for an ambient temperature of +18°C, which is obviously lower not only than the temperature inside the system case, but even than a normal room temperature (e.g. the temperature in our test lab is air-conditioned to be 22-24°C).
The manufacturer promises a typical efficiency of higher than 85% and compliance with the 80 Plus requirements (an efficiency of higher than 80% at loads from 20% to maximum). When working in a 220V power grid, the PSU efficiency was 87-88% at the top part of the diagram. The efficiency is going to be lower by 2-3% in an 110V power grid due to a bigger loss in the input circuits.
So, the Cooler Master Real Power Pro RS-A00-EMBA is an excellent 1000W power supply that easily yields its full output power and remains quiet under high loads. If the load is less than half the maximum, this PSU is almost silent. Quite a rare case, I have found virtually no drawbacks in this product. If you need a PSU with such a huge wattage, the model from Cooler Master is going to be an excellent choice.
P.S. Besides the standard selection of screws, cords and braces, PSUs usually come with pretty-looking but useless stickers. Cooler Master surprised me, having included a much more useful accessory into the box – a bottle opener.
This device proved to be perfectly compatible with all sorts of beer preferred by the author of this article. :)
The OCZ ProXStream 1000W has unusual dimensions for its class. It is a perfectly standard power supply of the ATX form-factor at 150x86x140mm (for comparison, the dimensions of above-described PSU from Cooler Master are 150x86x180mm and the PSU from PC Power & Cooling I will talk about below is 150x86x230mm). On one hand, it is good the PSU has a small length, but on the other hand, there are hardly many people who have a PC with a peak power draw of over 300-400W and a cramped system case the increased-length PSU wouldn’t fit into.
The smaller dimensions mean higher component density.
The PSU’s electronics is assembled on two PCBs placed one above the other. I have met this design in power supplies from FSP Group before (this company is the actual manufacturer of the OCZ1000PXS). Those were entry-level server-oriented models like FSP460-60PFN.
The top PCB heatsinks press against the PSU cover via an additional plate which you can see in the first photograph. This can hardly have a big effect on cooling efficiency because the cover is made from steel and lacks any ribbing. However, if you buy an OCZ1000PXS for you PC, don’t worry about its case getting very hot at work. This is normal.
An active PFC device and the main regulator’s switching transistors are located on the bottom PCB. The top PCB carries a power transformer, input rectifiers with accompanying chokes and capacitors, and a small card with overcurrent protection circuits.
The PSU is based on the ML4800CP chip which combines PFC and main regulator controllers. The quality of assembly and soldering is excellent. This PSU can be considered exemplary in this respect.
The OCZ1000PXS offers the following cables and connectors:
Unlike the above-described power supply from Cooler Master and unlike many other PSUs with similar wattage, the OCZ model doesn’t have 8-pin power connectors for graphics cards.
The PSU has four “virtual” +12V output lines with a total current of 70A (840W). Take note of the warning: if the input voltage is lower than 100V, the load power should not be higher than 900W. That’s new high-wattage PSUs for you: we’ve got to make sure our wall outlets can sustain 15A (see the previous section) or do not load the PSU much if the mains voltage is below the norm…
The cross-load diagram looks good: the +5V and +3.3V voltages deflect by no more than 2% from the nominal value at any permissible load while the +12V only deflects by 3-4% (to remind you, a deflection of 5% is considered allowable) at nearly maximum loads.
The output voltage ripple is low at max load, barely reaching 50 millivolts even on the +12V rail, the allowable maximum being 120 millivolts.
The PSU employs a Protechnic Electric MGT8012UB-R25 fan. The maker’s website only reveals that the letter U stands for Ultra High Speed. The latter could be deduced from its consumption current of 0.66A, which is very high for an 80mm fan.
Such powerful fans are usually designed with fewer blades (even with three blades only) and in thicker cases (80x80x38mm instead of 80x80x25mm), but this one is designed like an ordinary 7-blade 80x80x25mm fan. This sounds somewhat disturbing even before you turn the PSU on. Seven blades on an 8W motor can hardly be quiet.
And really, the PSU could keep the fan speed at 3200rpm until a load of 500W (this was not quiet, yet acceptable), but then the speed grew up to a fantastic 5500rpm and the sound of the OCZ1000PXS could be easily distinguished against the roar of the four fans our testbed was equipped with (which were two powerful 120x120x38mm fans with a rated speed of 2800rpm and two 80x80x25mm fans at 3000rpm). And I can’t even say the PSU ensures exceptional cooling. The air temperature grew by about 12°C in it at the maximum, which is rather a typical value. It seems that most of the fan’s power is spent pushing the necessary amount of air through the densely packed components of the PSU that produce a considerable resistance to the air flow.
The PSU is just a little less efficient than the above-described Cooler Master Real Power Pro, confidently fitting into the 80 Plus requirements and notching 86% at one point (at a load of 470W).
Summing it up, the OCZ ProXStream 1000W is good in everything except its noise. With its superb quality of manufacture, excellent voltage stability, low voltage ripple, and standard ATX form-factor, this power supply is, unfortunately, not suitable for home use. It will make a perfect PSU for a server that is located in a dedicated room where nobody would care about its noise. Alas, replacing the fan with a quieter one is not possible. The use of such a powerful fan is necessitated by the PSU’s high component density.
Note that OCZ Technologies has a 1000W GameXStream model with a 120mm fan. This model has standard dimensions of 150x86x140mm, too, but I can’t tell you anything about its noise until we have it in our labs. These two models are surely designed in different ways: the ProXStream just wouldn’t accommodate a 120mm fan.
PC Power & Cooling (PCP&C) is a renowned manufacturer of high-quality power supplies, from quiet models of the Silencer series with wattage ratings beginning at 310W and to the Turbo-Cool 1KW-SR, which I will test right now.
The Turbo-Cool 1KW-SR is one of the largest PSUs we’ve ever tested in our labs. It is 230mm long, which is 90mm longer than a standard ATX power supply. So, if you want to buy this PSU, make sure beforehand that you system case offers at least 270mm of free room (the length of the PSU proper plus at least 40mm for its cables).
The internals of this PSU can surprise a specialist even. There are actually three almost independent power supplies in one case here. In an ordinary PSU there is one power source and a common high-voltage section, the power transformer providing the necessary voltages (even in the above-described Cooler Master the two power transformers work strictly synchronously, thus making up one power source). The only common thing in the Turbo-Cool is the active PFC device. On the vertically standing cards there are three absolutely independent sources, each with its own high-voltage section and its own power transformer. I haven’t seen this kind of PSU design before.
Common too are the cooling heatsinks. These are ribbed aluminum bars placed across the PSU, and across the cards with independent sources, which are fastened to small and rather thin individual heatsinks that transfer heat from semiconductor components. As opposed to the OCZ power supply, these heatsinks have no contact with the PSU cover.
The PSU can sustain a continuous load of 1000W, yielding 864W from its +12V rail (the PSU doesn’t split its +12V power rail into multiple outputs). There are no limitations on the input voltage and ambient air temperature: the 1KW-SR can work at full load with 90VAC voltage and at an up to 50°C temperature.
Peak loads are specified, too. The period of time the PSU can work under such loads is not specified, but it’s usually one minute.
The PSU offers the following cables and connectors:
Two adapters for 8-pin graphics card connectors are included with the PSU. They look somewhat funny at first because they have 8-pin connectors on both ends, making an impression that they adapt nothing.
The cables for peripheral devices have different lengths so that you could connect your HDDs and optical drives in a most optimal way.
Although the ATX12V and EPS12V standards allow a 5% deflection of the output voltages from the nominal values, the PSU manufacturer promises a deflection of no more than 2% for the three main voltages: +5V, +12V, and +3.3V. The PSU actually violates these limits, but only at extreme loads. The +12V voltage is more than 2% above the nominal value when there is a high load on the +5V rail and a low load on the +12V rail. The +3.3V voltage is about 3% below its nominal at the max allowable load on that rail. The voltages are overall very stable. At load distributions typical of modern computes – this is the bottom third of the diagram – the three voltages are all green.
There’s low voltage ripple on the PSU output, with single high-frequency spikes shooting up and down randomly.
The PSU employs a Delta Electronics EFB0812EH power supply with a rated speed of 5000rpm. Unlike the above-described model from OCZ, it is a three-blade model with lower noise but also lower static pressure. This means that at the same speed and aerodynamic resistance the EFB0812EH is pumping a smaller amount of air than the MGT8012UB-R25. It is clear, though, even externally that the aerodynamic resistance of the components of the Turbo-Cool PSU is lower than in the OCZ ProXStream.
The fan speed proved to be lower, too, varying from 2400 to 4670rpm depending on load. Alas, this PSU is not quiet, either, although it is more agreeable to the ear than the OCZ model.
The Turbo-Cool 1KW-SR does not meet the 80 Plus requirements since its efficiency is below 80% at a load of 200W (20% from the maximum). The average efficiency is 83% just as promised by the manufacturer.
The PSU is no record-breaker in terms of efficiency under low loads, its efficiency degenerating to 50% at 50W. Well, these numbers are rather theoretical ones. If your PC can load a 1000W power supply, its power consumption is going to be higher not only than 50W but even than 150W even in idle mode!
Thus, the PC Power & Cooling Turbo-Cool 1KW-SR is a superbly made product with very good electrical parameters, but noisy and very large. Although it is considerably quieter than the OCZ model, this PSU may still prove too noisy for home use. It is an appealing option as a PSU for a powerful workstation or server with an appropriate system case, though.
We have already tested high-wattage PSUs from SilverStone in our labs. That was the Zeus series limited to 750W. And now we’ve got a 1KW unit from the Olympia series. Like with OCZ’ product range, this is not the single model with that wattage available from SilverStone. The company offers a 1000W Strider ST1000 with a 135mm fan and detachable cables.
Talking about the market positioning of such PSUs, the models with an 80mm fan are intended for servers and powerful workstations in the first place due to their high noise level (the manufacturers can’t also miss the opportunity to mention that these power supplies suit for gaming systems as well). The quieter models with 135mm fans are for home computers. We haven’t yet got an ST1000 for us to test, so we can’t tell you how it compares with the very quiet PSU from Cooler Master.
With a design similar to the model from PC Power & Cooling, the PSU from SilverStone is shorter by 1 centimeter: 150x86x220mm.
The PSU has a non-standard input connector:
The connector is a logical consequence of the increase in current consumption to 15-16A, but it is going to be a problem for owners of UPSes whose cables wouldn’t plug into this power supply.
The PSU has an ordinary internal design. Its component layout resembles Zippy’s models I tested in an earlier article.
The power components are distributed among the four separate heatsinks. An additional plate, covered with black insulating film from above, is fastened on the rear pair of heatsinks.
There is a fan speed controller at one side. The corresponding thermal resistor is fastened nearby on the heatsink. Below on the same heatsink are two SBL4060PT diode packs of the input rectifier.
The PSU’s continuous output power is 1000 watts. The manufacturer’s website also mentions a peak output power of 1100W as well. The PSU can work under full load at an input voltage of 90V and an ambient temperature of 50°C. The +12V rail is not split up into multiple outputs here. The PSU allows to put a long-term load of 80A (960W) on this rail, and even up to 88A for a short period of time.
The PSU offers the following cables and connectors:
The PSU has 6-pin as well as 8-pin connectors for graphics cards. The latter are labeled as “PCI Express” for you to not confuse them with the 8-pin CPU power connector. The CPU power cable has no such label.
The PSU yields stable voltages under any allowable load, but the +12V voltage is considerably higher than necessary (it remains within the allowable limits, though).
Strictly speaking, the level of output voltages can be described with two parameters. One shows how far the average value of a voltage deflects from the nominal value (this parameter is often referred to as Voltage Range or just Range). The other, usually called Line Regulation, shows how far the minimum and maximum of a voltage stand from each other. So, if the PSU yields a minimum of 12.25V, an average of 12.3V and a maximum of 12.35V, the Voltage Range (the average deflection from the nominal value) is 2.5%, yet the voltage fluctuates little around that average value, by less than 0.5% (Line Regulation).
To cut it brief, it is better to have large same-colored areas on the cross-load diagram, even if their color is not green, than narrow bands from red on one side of the diagram to green in the middle and then to red again on the other side. We’ve got the former case with the SilverStone Olympia OP1000.
The output voltage ripple at max load is on the verge of the allowable: 50 millivolts on the +5V rail (50 millivolts is the allowable maximum, actually) and about 100 millivolts on the +12V rail (120 millivolts is allowable).
The PSU employs a Sunon PMD1208PTB1-A fan that can reach a speed of 4700rpm at 12V voltage and 4.8W output power. The fan has standard dimensions of 80x80x25mm and seven blades.
Although the fan speed is lower than that of the Turbo-Cool 1KW-SR, the Olympia OP1000 is only really quieter at loads from 300 to 600W because the 1KW-SR increases its fan speed continuously while the OP1000 begins to do that only from a load of 500W.
The PSU meets the 80 Plus requirements, at least in a 220V power grid (its efficiency is going to be a little lower in an 110V grid). The average efficiency is about 83% with minor fluctuations depending on load.
So, the SilverStone Olympia OP1000 performed not much worse in my tests than the Turbo-Cool 1KW-SR despite the use of a traditional circuit design. The latter PSU has somewhat more stable voltages and lower output voltage ripple at full load. The OP1000 easily meets the requirements of the ATX12V and EPS12V standards, though, and won’t provoke any problems at work. It may prove somewhat too noisy for home use, though.
While low-wattage power supplies aren’t very variegated when it comes to their exterior design, there is even less variation among 1000W models. They are all black bricks, perhaps with varying length. The TG1100-U96 is in between the PSUs from OCZ and PC P&C in this respect: 150x86x170mm which is a mere 3 centimeters longer than a standard ATX power supply.
But looking at the back panel we can see one more significant difference: this PSU features a push-pull cooling design with two fans. One fan is sucking air into the PSU case and the other exhausts it.
Like the PSU from Cooler Master, the TG1100-U96 has two power transformers that work in sync. The manufacturer says this improves stability of the output voltages, but this is not exactly so. As I wrote above, the two-transformer circuit is simpler in terms of component placement and is also cheaper than one transformer of double capacity. The PSU marking (TOP-Txxx) reveals the actual manufacturer of the PSU, which is Topower.
The heatsinks are shaped in a fanciful way here: they have a central groove along their entire length and small dense fan-like ribbing. Intricately shaped heatsinks are not always good, though. More often than not they create additional aerodynamic resistance, slowing air flow and calling for a more powerful fan.
The max output power of this PSU is 1000W and it can yield 960W of it through its +12V rail. The +12V rail is split into six “virtual” output lines. The manufacturer’s website doesn’t say at which ambient temperature and input voltage these numbers are correct.
The PSU offers the following cables and connectors:
Topower, the actual manufacturer of this PSU, is known to be predisposed towards screened cables. Unfortunately, this screening brings no real benefits. It has no effect on stability of the PC and on the output voltage ripple. Screened cables are thick and stiff and it is hard to lay them out in the system case in such a way that they don’t take too much space and press on the connectors too much. This can be seen with the 8-pin CPU power connector which is actually composed out of two cables with 4 wires in each (but the connector itself is not splittable).
Fortunately, the Tagan PSU uses screening on the graphics card and CPU cables only (the VGA/HDD cable shown in the picture above is a strange reminder of the past when graphics cards used to be powered from the same connector as HDDs). HDD power cables in original Topower PSUs can be designed in the same way, too, which is downright unacceptable.
An interesting feature of this PSU is that it comes with adapters from SATA power connectors (the PSU offers 10 of them) into PATA or Molex power connectors (the PSU offers only three such connectors). I think this is a reasonable approach. If the PSU is expected to be used in modern powerful systems, its selection of native connectors should meet the requirements of such systems, not the other way around.
The cross-load characteristics are rather stable, but the +5V and +12V voltages are set too high, by 3-4%.
The oscillogram of high-frequency pulsations (one division of the horizontal axis means 10 microseconds) shows that the PSU barely but meets the requirements of the standard: the voltage ripple of the +5V rail is 50 millivolts, which is exactly the maximum allowable value.
Alas, when we switch into slower scanning, 2 milliseconds per division, we can see low-frequency pulsations on the +12V rail. The combined voltage ripple is thus over 150 millivolts (three divisions of the vertical axis), the allowable maximum being 120 millivolts. This oscillogram was recorded at full load. When the load is lower, the voltage ripple is lower, too.
The PSU employs Top Motor DF128025BU and Tagan S0801512HG fans (the latter one is actually manufactured by Globe Fan). The first fan is 80x80x25mm large and is classified as a high-speed model whereas the second one has a modest size of 80x80x15mm. However, there is only one fan speed curve on the PSU label, showing that the fan speed (I wonder what fan of the two it is?) varies from 1465rpm to 2690rpm, depending on current load.
The real situation is not that bright, though. Quite expectedly, the manufacturer showed the speed of the slower fan (S0801512HG) on the label and, presumably, measured it under an ambient temperature of 18°C because our graph, recorded at a room temperature of 24°C, has the same shape but is shifted upwards by 300-400rpm.
This is not the worst thing, though. The second fan (DF128025BU) indeed proved to be a high-speed one. Having started from 2200rpm, it reached 4200rpm at a load of 1000W. With this fan the diagram doesn’t look at all as pretty as the one on the PSU label.
And I even wonder if the low-speed fan is needed at all here. The push-pull design works well when both fans have identical, or at least similar, parameters. Here, we’ve got a powerful high-performance fan coupled with a rather slow and thin one (15mm instead of 25mm, and the thinner blades create lower static pressure). Moreover, we can consider the example of the above-described SilverStone Olympia OP1000 to see that its fan, with parameters similar to the DF128025BU, could cool the PSU easily all alone. So, I’ve got an impression that the second, low-speed fan of the Tagan TG1100-U96 PSU is mere decoration.
The Tagan TG1100-U96 can’t be called quiet, of course. It’s got its name of TurboJet for a reason. Although the more powerful fan is hidden deep in the system case, not every system case is going to muffle its noise which becomes very high at high loads on the PSU.
Well, noise is not the worst thing here, I think. After all, the competing models with an 80mm fan aren’t quiet, either. The bad thing is that the manufacturer attempts to mislead the customer by not mentioning the PSU’s having a second fan, which is powerful and noisy.
The efficiency of this PSU is quite typical for a 1000W model, a little less than 85% at the maximum. However, it degenerates quickly at loads below 200W, down to 50% at a load of 50W.
I can’t say I’m pleased with the Tagan TurboJet TG1100-U96. And my gripes are not actually about the PSU’s technical characteristics, but about how they are interpreted by the manufacturer. First, the pretty useless and cumbersome screening of the cables is advertised as a means to reduce noise and pulsations (but the PSU couldn’t fit into the requirements of the standard in terms of output pulsations; moreover, it is the single model among those included into this roundup that has a noticeable low-frequency voltage ripple on its output). Second, the PSU label shows a fan speed graph that is very misleading. It shows the speed of the slower of the two fans installed in the PSU. And talking about its parameters, the TG1100-U96 is at best equal to its opponents but is never better than them.
The PowerTrain TOP-1000P9 U14 is a second unit out of the six included into this review that is equipped with a 135mm fan (the Cooler Master Real Power Pro was the first one). Hopefully, it will be noiseless – we’ve already found out that the PSUs with 80mm fans can’t be quiet…
This PSU uses a design with two transformers working in sync, too. It is equipped with active PFC and features independent voltage regulation. These two features are in fact standard for PSUs of this class.
The heatsinks of this PSU aren’t as fanciful as in the Tagan, yet they are remarkable, too, consisting of a large number of small vertical corrugated ribs.
Having a total wattage of 1000 Watts, this PSU can deliver up to 900 Watts (75A) along its 12V power rail, which is divided into six output lines. And like with other new PSUs from Topower, this division is very tenuous.
There are current-measuring shunts next to the output cables on the PCB (the thick U-shaped strap in the photograph above). The signal from the straps should go to the current limitation circuit so that if there is a current of over 18A on one strap, the protection shuts the PSU down. I call this implementation “virtual splitting”: there is one +12V power rail, but the load current is deliberately limited on the output connectors.
So, the TOP-1000P9 has those shunts but has no protection. If the current in one shunt is higher than 18A, the “Combined” indicator on the front panel of the PSU will light up. That’s all – nothing else is going to happen. Of course, there is no “automatic combining of the lines” here as you can read in the advertising materials. You can just regard the PSU as having had no separate lines beforehand. This is not a technical drawback. It is rather interesting from the marketing point of view: the manufacturer wants to make its product appealing as having the advantages of models with multiple +12V lines (which are actually imaginary advantages because the division into multiple lines has no effect on stability or cross-load characteristics) as well as those of models with one +12V power rail.
The PSU offers the following cables and connectors:
The screening of the graphics card cables has no practical meaning because it does not affect the level of interference (these are power rather than signal cables, after all), but makes it difficult to lay the cables neatly in the system case. Moreover, these cables have ferrite rings to suppress high-frequency interference. In my opinion, the manufacturer should have limited itself to the rings only – this wouldn’t look so cool and unusual, but wouldn’t hinder the assembling of the system, either.
The PSU delivers very stable voltages. The diagram is mostly green. The +12V voltage is a little higher than necessary, by 2-3%, but this deflection is not critical in practice.
It is worse with the output voltage ripple at the maximum load. The PSU exceeds the allowable limit on the +5V rail (the limit is 50 millivolts or one division of the vertical axis) and is close to the limit on the +12V and +3.3V rails (120 millivolts and 50 millivolts, respectively). But unlike the Tagan unit, the TOP-1000P9 doesn’t have a strong low-frequency pulsation.
The PSU uses a 135x135x25mm Globe Fan RL4B S1352512HB fan with blue LED highlighting.
The fan speed changes but little at loads from zero to 600W and then begins to grow up quickly. Although the Topower is louder than the above-described PSU from Cooler Master (whose fan started out with a speed of only 830rpm), it is overall good through the entire range of loads and is surely quieter than models with 80mm fans. In fact, the TOP-1000P9 is the only serious rival to the Real Power Pro in terms of noise among all the PSUs tested here.
The PSU boasts a good efficiency of 85% which lowers to 82% at high loads. The PSU complies with the 80+Plus requirements at least in a 220V power grid: its efficiency is higher than 80% at 20% from max load (200W).
So, the only real disadvantage of the Topower PowerTrain TOP-1000P9 U14 is that its output voltage ripple is higher than allowable at full load. The ripple is lower at lower loads, though. Otherwise, this is a good product, with high efficiency, stable output voltages, and moderate noise level. The last thing is especially important since there are only two PSUs among the tested ones, namely the Topower PowerTrain and the Cooler Master Real Power Pro, that are quiet enough for a home computer.
It might have been expected that noise would prove the greatest problem with high-wattage power supplies, especially under high load. Neither of the PSUs with an 80mm fan can be considered suitable for a home computer. They are all too loud and the noise from their fans will be audible even through the noise of graphics card and CPU coolers.
This singles out two models, the Cooler Master Real Power Pro RS-A00-EMBA and the Topower PowerTrain TOP-1000P9 U14, as the solutions with much quieter 135mm fans.
The all-around winner of today’s tests is the PSU from Cooler Master that boasts superb parameters together with very quiet operation. Being a good product overall, the Topower model is not as good as the Real Power Pro in its noise characteristics as well as in electrical parameters. Particularly, the TOP-1000P9 does not meet the requirements of the output voltage ripple standard .
If you are choosing a power supply for your server or workstation and do not worry that much about the noise, you should also consider the models with an 80mm fan from OCZ, PC Power & Cooling and SilverStone. Each of these three products is good and the OCZ unit is also the only tested 1000W unit to fit within the standard ATX form-factor dimensions of 150x86x140mm. Its high noise level is the tradeoff for the compactness, though.
The Tagan TurboJet TG1100-U96 disappointed us this time. And it is not only its not quite satisfactory parameters (it doesn’t meet the output voltage ripple requirements) but also the manufacturer’s attempt to intentionally mislead the user. The PSU has two fans that differ greatly in the level of consumed power and generated noise, but the PSU label shows the speed/noise diagram of the less powerful fan of the two without even mentioning the other one.