by Oleg Artamonov
11/22/2006 | 03:10 PM
The more brands come to the PSU market and the tougher the competition becomes, the wider various marketing inventions are employed besides just technical advances and innovations.
Unfortunately, besides experimenting with colorful box designs and accessories (well, it’s hard to expand a PSU’s accessories set beyond the customary set of a cable, a couple of braces, and a handful of stickers), the marketing department comes up with a technical lingo to bewilder the customer with mysterious terms and abbreviations. Every box and instruction shows a long list of employed technologies, the point of some of which may be distorted almost to the opposite.
That’s why I’m going to walk you through some of the technologies (or what the PSU manufacturers regard as such) most frequently mentioned on boxes with modern power supplies. And then I’ll put PSUs with such technologies to practical tests.
You can find the details on the latest PSU technologies as well as the detailed description of the 6 PSU models from Antec, CoolerMaster, Corsair and High Power in our previous article called ATX Power Supply Units Roundup. Part V.
Let's continue!
Power supplies from Hiper can be easily identified by their meshed cases. In PSUs from other manufacturers there is usually a vent grid in the rear panel plus a couple of small slits opposite to the hottest components. Here, the case is all meshed. Well, I think this has little to do with cooling. That’s rather a matter of aesthetics.
Hiper promotes its Type M series as a cheap version of the Type R series, without detachable cables.
It is all ordinary inside. An active PFC device is located on a separate card. In my previous article I complained that that card was soldered untidily, but it is all very neat here.
Otherwise, the PSU circuit design doesn’t differ much visually from the Type R series. In spite of its active PFC, the PSU is intended for an input voltage of 220V only. Another model, intended for an input voltage of 115V, is produced for the US market.
The PSU is declared to comply with the ATX12V 2.2 standard. It has two +12V lines (separated “virtually”, as usual) with a combined load current up to 30A. The load capacity of the +3.3V and +5V lines is not low, either.
A peak output power of 580W is specified, too, but the conditions under which the PSU can yield it are not declared (the maximum duration of operation under the peak output power is usually specified). Generally speaking, this parameter is rather a marketing trick. Surely, each high-quality power supply is made with some reserve so that it can yield a short-term output power higher than the maximum sustained one without breaking down.
The PSU is equipped with the following cables and connectors:
An adapter from two Molex into one graphics card connector is enclosed with the PSU. It is meant for use with SLI or CrossFire configurations.
When the PSU was working in pair with an APC SmartUPS SC 620, the UPS’ overload indicator would wake up at a load of over 315W. Switching to the batteries was performed without problems.
There’s very low ripple on the PSU output. Its amplitude is smaller than 15 millivolts on the +5V rail (50 millivolts is the acceptable maximum), and there’s almost no ripple at all on the +12V.
The PSU’s cross-load diagram looks good. This unit is operable in the entire range of declared loads, including high load on the +5V and +3.3V rails. When working in that mode, the voltages deflect quite far from the nominal values, but exceed the allowable limits only in the critical case when there’s almost no load at all on the +12V rail. In that part of the diagram which corresponds to the typical power consumption of modern computers, all the three voltages are in a green zone.
The PSU uses a Young Lin Tech DFS122512H fan. The manufacturer declares a fan speed from 700 to 1800rpm depending on the temperature inside the PSU (the temperature thresholds are not specified).
In my tests, the fan speed was 910rpm at a load of 50W and an ambient temperature of 22°C (the PSU was just lying on a desk). This is a good, but not record-breaking, performance. At max load the fan speed grew only to 1800rpm, rising linearly from a load of 150W. So, this power supply should be considered very quiet.
After you turn the computer off, the fan is rotating for a while at a low speed, powered from the standby +5V source. This helps quickly cool the PSU when its work is over.
Alas, the efficiency of the Hiper is low. It’s only 75% at the maximum, which is a very modest result today when all the manufacturers are boasting they have hit the 80% mark. I had even suspected some problem with our testbed, but it produced the same number after calibration. Looking up in the PSU manual I found its specified efficiency, only 76%. So that’s indeed how efficient this PSU is.
To all appearances, Hiper prefers to base its new products on older ones without bringing too many changes into them. This supposition is confirmed by the position of the active PFC on a separate card (most other manufacturers have long moved it to the main PSU card) and by its very moderate efficiency. On the other hand, apart from that low efficiency, the Hiper Type M 530W doesn’t have serious defects. It yields low-ripple and stable voltages and has a quiet fan.
Mushkin is yet another manufacturer of overclocker-friendly memory modules that has entered the PSU market, too. Mushkin’s OEM supplier is Topower whose name should be well known to our readers.
The XP-650 has two coaxial 80mm fans, which is not a very frequent cooling solution. Most manufacturers prefer to use a single fan, either 80mm or 120mm in diameter.
Topower’s traditional huge black heatsinks with numerous small ribs occupy almost all of the free space inside the PSU. A fan speed controller card is fastened on one of them.
Otherwise, this is an ordinary model designed according to the time-tested half-bridge push-pull topology and without PFC. What is curious, the company’s website shows a photo of the PSU with a sticker that reads “100 – 240VAC” that should imply active PFC, but our sample has neither the sticker nor PFC. It has an ordinary red-colored switch to select the input voltage with.
The PSU has two connectors for detachable graphics card cables and six connectors for cables of hard and optical drives. These groups of connectors differ only with their color, so you should be careful when attaching the cables. I don’t understand why they didn’t make such a connector configuration that the user wouldn’t have a chance to make a mistake.
The PSU offers the following cables and connectors:
The following is enclosed with the power supply:
The detachable cables are wrapped into vinyl chloride tubes and are screened. The screen is connected to the common wire. As a result, the cables are thick and stiff and it’s hard to bend them at a sharp angle, which may prove inconvenient in computers with several hard and optical drives placed close to each other.
The PSU has a separate cable with a ground clamp. The user is supposed to fasten it under a screw in the system case. Well, I guess the PSU’s own four fastening screws are quite enough for a proper electrical contact with the system case.
The XP-650 is declared to support Rail Fusion technology which means joining several +12V outputs into one in case of overload. I wrote about this technology in the introduction to the article and in the section about the Corsair CMPSU-620HX. It means that there is no splitting of the +12V rail into several lines in this PSU. The PSU manufacturers’ marketing departments have already convinced quite a lot of users that there must be several +12V rail, so they now have to declare new technologies which come down to nothing else but the lack of other technologies. The matter of the splitting the +12V power rail into several output lines is all about ensuring user’s safety and has nothing to do with stability or output power of a power supply.
So, the PSU has four +12V outputs with a current of 20A each. The combined current cannot exceed 44A. As you may have guessed, there is actually only one +12V 44A power rail inside the PSU while the rest of the text on the label has nothing to do with the real technical characteristics of the product.
At a load of 630W the output voltage ripple amounted to 26 millivolts on the +5V rail, to 39 millivolts on the +12V rail and 25 millivolts on the +3.3V rail. There’s only high-frequency pulsation here.
The cross-load characteristic of this power supply looks beautiful. The output voltages do not even approach the 5% deflection anywhere in the allowable load range.
The XP-650 has two Globe Fan RL4G S0802512FHD fans, one in a translucent case with emerald highlighting and another made of ordinary black plastic. The fans are connected in parallel, so their speeds are regulated synchronously.
At loads below 300W the fans are rotating at a speed of 1800rpm. Although this speed is rather low for an 80mm, the fans in this PSU are quite distinctly audible at that. From a load of 300W onward (and sooner in a closed system case), the speed of the fans is growing up linearly until it reaches 2600rpm. The noise of the air passing through the PSU becomes rather irritating. I guess it is due to the densely ribbed heatsinks that obstruct the airflow.
The PSU is only 79% efficient. Its power factor is 0.66 at the maximum.
The Mushkin XP-650 power supply might be a good PSU to buy if it were not for its price. Priced at about $200, it is considerably more expensive than a majority of PSUs of similar wattage, including the above-described Corsair CMPSU-620HX. The Mushkin PSU doesn’t offer anything exceptional. It has a deplorably low efficiency for its product class and unhandy thick cables. That’s why I can’t recommend it to you for purchase, at least until the company have seen to these drawbacks and reduced the price to a competitive level.
I reviewed power supplies from OCZ twice in my articles. Those were ModStream and PowerStream series, both manufactured for OCZ by Topower. With the release of its new GameXStream series OCZ decided to change its supplier. Both these PSUs are manufactured by FSP Group. This is obvious even from the label. There’s only a different manufacturer name on it in comparison with FSP’s own Epsilon that we have tested in our labs, too.
Below I will describe the two models, 600W and 700W, together since they are based on the same platform and do not differ much from each other.
The internal design of the PSUs is typical of FSP’s new series and is first of all remarkable for its small heatsinks. As I wrote earlier, it had become possible due to increased efficiency of the PSU. Particularly, they make uncommon (in comparison with other PSUs) use of diode packs. One 30A pack is usually installed to provide a 30A current, but in new PSUs from FSP they connect two packs in parallel. Why if each of them can ensure the required 30A? Because the voltage drop on diodes depends non-linearly on the passing current. So, two packs through each of which a current of 15A is passing generate less heat than one pack with a 30A current. This is the reason why there are so many components on the heatsinks of this GameXStream PSU. Three heatsinks even had to be used because the engineers couldn’t have accommodated all the components on only two of them.
On one hand, this approach increases the cost of the semiconductor PSU components, but the cost of the heatsinks is lowered. FSP’s engineers must have achieved a positive overall effect with that solution.
Switching power supplies – any switching power supplies – cannot work under zero load because the energy accumulated in the inductances of the PSU’s output circuits (in the transformer or choke, depending on the topology employed) doesn’t go into the load, which results in an uncontrollable growth of voltage on the PSU output. The ATX standard, however, requires that the PSU be able to start up without external load, and the manufacturers used to install resistors that provided the necessary minimum of load. Such PSUs worked normally and yielded stable voltages even if nothing at all was connected to them.
Lately, those resistors have been abandoned as the manufacturers are trying to improve the efficiency of their products. Of course, the resistors dissipate little power, but this is still considerable when the PSU is under minimum load. Another problem rose up as a consequence. New PSUs would not start up with some mainboards, those that didn’t provide some minimum load right after the start. As a result, the PSU that didn’t have its own internal load would not operate correctly or would not send the Power OK signal to the mainboard.
To avoid this problem, FSP engineers implemented a “smart” internal load that works only for the first few seconds on your turning the PSU on. It is located on a separate small card (see the photograph above) and consists of two clusters of pull-up resistors, two transistors that connect those clusters to the +5V and +12V circuits, and an RC circuit that keeps the transistors open for a short interval after the turning on of the PSU. As soon as the RC circuit capacitor is charged up, the transistors close and turn off the pull-up resistors. By this time, the mainboard should have started up successfully, providing an external load on the PSU. This is how simple it is. It’s even strange that PSU manufacturers didn’t come to this solution long ago. The problem of not-starting-up mainboards wouldn’t have occurred then (this problem was reported not only with regards to FSP power supplies, by the way).
However, there can be another problem. The PSU may refuse to start up if you turn it on right after having shut it down. If the capacitor of the RC circuit hasn’t had enough time to get uncharged, the transistors will be open initially. I don’t think that’s much of problem, though. You can just wait for a few minutes and try to turn the PC on again.
The PCBs of both power supplies have a marking that reads “FSP700-80GLC(NEW)” which seems to imply a new revision of the model. Unfortunately, I have no information about the changes (besides the above-described card, but it is separate from the main PCB). Visually, the group voltage regulation choke seems to have become larger. A more detailed examination of the interior of the PSU would be beyond the scope of this review.
The PSU is equipped with the following cables and connectors:
The cables are all sleeved.
Here’s the label of the senior, 700W unit. The junior model differs from it in having a 100W lower load power both on the +12V rail and in general.
The +12V power rail is “virtually” divided into 18A pieces just as the standard demands. It’s good that the label tells clearly which connector is powered by which line. The wires of the different lines are color-coded in the connectors: yellow for the 12V1 and a combination of yellow with black, blue and green for the other three lines.
The power supply worked normally with an APC SmartUPS SC 620 under loads up to 370W irrespective of the power source (electric mains or batteries). The UPS switched to its batteries without problems.
The cross-load characteristics of the 600W unit don’t look too pretty, but are acceptable. The +5V voltage is not very stable, but fits into the green zone within the load range typical of modern computers, i.e. 20-50W. The +12V voltage is set higher than necessary and is going to be 2-3% above the nominal value in a real computer. It deflects by 5% from the nominal – the maximum permissible deflection – only when there’s high load on the +5V and +3.3V rails, which is insignificant for today’s PCs.
It’s worse with the 700W model. Its +3.3 voltage fluctuates wildly through the diagram, making it rather untidy, with ragged borders and a lot of red color. Well, it’s all normal within the area that is the most important for a modern computer system, but anyway. This cross-load diagram is acceptable and nothing more. Moreover, there’s quite a lot of PSUs with group voltage regulation (like in the GameXCtream) that show better results in this test.
The two models don’t differ much in terms of output voltage ripple. The oscillogram above is for the senior model: 25 millivolts on the +5V rail, 81 millivolts on the +12V rail, and 28 millivolts on the +3.3V rail. That’s all within the norm. Only high-frequency pulsation is present.
Protechnic Electric MGA12012HB-O25 fans are installed in both models. These are the same fans as in the original PSUs from FSP. They have gained a reputation of average-quality fans among the users. Some users have reported that the fan blades are buzzing audibly in their samples of the PSU.
I was very pleased with very quiet operation of FSP Epsilon power supplies, but the OCZ GameXStream is just the opposite. Even under minimum load the fan speed is over 1300rpm, and it grows up to 2200rpm at the maximum. The fan speed graphs of the 600W and 700W models almost coincide so I only publish the graph of the senior model above.
The efficiency of the PSUs is high. It reaches 87% at one point with the 700W model, the average being about 86%. The 600W model has an efficiency of 84-84%.
So, while the FSP Epsilon units left very good impressions on me, I am rather disappointed with the new PSUs from OCZ. Although they do not in fact differ from the Epsilons, and even have absolutely the same fans (for example, Zalman replaces the original fans in its PSUs, which are too manufactured by FSP, with quieter ones), they are rather noisy at work and do not have high voltage stability. I can’t say if all other PSUs produced for OCZ are like that or if only the Epsilon series is (and if yes, then it may be just models produced at a certain time period) because FSP Group’s representatives didn’t answer to my inquiries. So, be careful when shopping. Well, if you can choose between these two PSUs and other models described in this review, you should instead consider the products under the brands of Seasonic, Corsair or Zalman, even though the latter company buys its PSUs from FSP, too. If you are choosing between FSP Epsilon and OCZ GameXStream PSUs, you should be aware that these two series differ only in the name on the label. So, you may want to choose what is cheaper.
I have mentioned Seasonic above. Products from that company come under the brands of Antec and Corsair, but here’s a power supply with a native label. It is the last version of the Seasonic S-12 model that is RoHS compliant and has sleeved cables. I guess the users will be more attracted by the latter than by the former thing.
It doesn’t immediately strike one’s eye, but the circuit design of this PSU is identical to the above-described Corsair CMPSU-620HX. The S-12 just has larger heatsinks that have two times more turned-back “petals” and cover completely the PCB underneath. Those petals were alternating in the 620HX.
Otherwise the two PSU models are identical except that the Corsair one has higher wattage and the S-12 lacks detachable cables. Seasonic offers the latter option in its M-12 model that also features a second fan on the rear panel.
The PSU complies with the ATX12V 2.2 standard, providing a max current of 33A on the +12V power rail which is “virtually” divided into two outputs. It offers the following cables and connectors:
The cables are all sleeved. A couple of adapters are included with the PSU for powering floppy drives if you’ve still got them in your system.
When the PSU was working with an APC SmartUPS SC 620, a load of 360W and 345W for the mains and batteries, respectively, was the highest allowable for the PSU. The switching to the batteries was performed without problems.
At loads higher than 300W a choke would begin to buzz quietly in the PSU. I had to put my ear next to the PSU case to hear it, though.
Apart from the short-term spikes at the moments of switching the transistors, which are going to be effectively filtered out by the mainboard, the voltage ripple at a load of 470W was 30 millivolts on the +5V rail, 50 millivolts on the +12V rail and 23 millivolts on the +3.3V rail. The PWM controller works at a frequency of about 130kHz.
The cross-load diagram of this power supply looks perfect. It’s all green except for the +3.3V voltage going into yellow a little. All the voltages are going to be rock-solid under the conditions of a real computer system.
The PSU makes use of an Adda AD1212MB-A71GL fan whose speed is kept at about 750rpm until a load of 250W. Then the speed begins to grow, reaching a mere 1150rpm at 470W. That’s a recording-breaking result, by the way. The fan speed is going to be somewhat higher in a hot system case, yet it anyway won’t be as high as in the PSUs described earlier in this review. The S-12 is one of the quietest power supplies ever tested in our labs!
The SS-500HT also boasts a very high efficiency: 86% at high load and a little lower at low loads. It is going to be about 83% under real-life conditions. The S-12 doesn’t comply with the 80 Plus program because its efficiency is lower than 80% at a 150W load. Well, Seasonic has a special version of this PSU that is declared to meet that requirement. Perhaps its efficiency graph would look better.
So, the S-12 has gone through our test program with ease. It offers all the connectors you may need, its output voltages are rock solid, its fan is very quiet, and its efficiency is always higher than 80%. The only thing I may want to improve in this PSU is to lower its voltage ripple, especially the short spikes at the moment of switching the inverter’s transistors, but this is hardly a serious defect. This won’t affect the stability of the system in any way.
I once tested a Zeus in our lab. It was the 650W ST65ZF model, but it’s clear even visually that the ST75ZF and the ST85ZF differ from it not only in wattage: the 80mm fan has moved to another panel of the case and it can be seen through the vent grid that the electronics resides on two PCBs whereas the ST65ZF was a traditional single-card model.
The 750W and 850W models are very similar, so I will discuss them both together. According to the certificate number (UL# E166947), the actual manufacturer of the PSU is Enhance Electronics. One of the PCBs bears the marking of ETASIS Electronics.
The PSU is cooled with a single 80mm fan located on its internal panel (it is internal when you install the PSU into the system case). Using an 80mm fan in such high-wattage models usually results in noisiness. Well, I’ll check this out soon.
The cover removed, you can see a very crowded design with two main PCBs densely populated with heatsinks and auxiliary cards.
The top PCB carries an active PFC device (based on an UCC3818AD chip), line filters and a couple of high-voltage capacitors.
Most of the PSU’s electronics are located on the bottom PCB: a power transformer (in the far left corner), two heatsinks with the inverter’s transistors and diode packs, and three small additional cards. One of them accommodates the main PWM controller of the PSU (a UCC3895DW chip), but the other two are quite a surprise:
Each card carries a choke, an L6730 chip (it is a PWM controller and a synchronous rectifier controller in a single case), and a few transistors in planar cases covered with a small heatsink. These are independent regulators of the +5V and +3.3V voltages.
What’s so strange about that? I have often referred to power supplies with independent voltage regulation in this review, let alone my previous articles. But the point is that other such PSUs do not have truly independent regulation. They use the so-called regulators on magnetic amplifiers, also known as saturable-core chokes. Such a regulator has very simple design and superb efficiency, but cannot work independently from a direct voltage like a full-featured PWM controller does. It is fed a pulsating, unsmoothed voltage from the secondary winding of the main transformer and its point is in narrowing the width of the impulses thus reducing the average voltage we get after the rectification. That is, applying voltage from the transformer’s 12V winding to this regulator and setting it up to reduce the impulse width in 2.5 times, we get 5V at the output. In fact, magnetic amplifiers are used in a majority of power supplies, except for the cheapest models, to get +3.3V voltage from the transformer’s 5V winding, but I only regard a PSU as having independent voltage regulation if both +5V and +3.3V voltages are provided by independent magamp regulators. It’s only in this case that none of the PSU’s output voltages is influenced strongly by the others.
Here, each card is a complete and independent PWM regulator that receives a direct voltage on its input. The regulator’s operating frequency is 400kHz, which helps greatly reduce the size of the choke and the capacitance of the smoothing capacitors on the output. A synchronous rectifier is used to increase the efficiency. I don’t have anything against this solution, yet I would much like to know the reasons why the developers decided to go for such a complex design.
The PSUs comply with the EPS12V and ATX12V 2.2 standards. The load capacity of the +5V and +3.3V rails isn’t high for their wattage, but the modern standards do not demand that. Instead, the PSU can yield almost of all its output power on the +12V rail alone. Well, there’s only a 50W increase of the allowable load power on the +12V rail with the senior model whereas its overall wattage is higher by 100W.
The ST75ZF worked normally with an APC SmartUPS SC 620 at loads below 360W from the mains and at loads below 320W from the batteries. The numbers are both 5W lower with the ST85ZF.
The PSUs come with the following cables and connectors:
The cables are all sleeved, except for the cables you connect to your optical and hard drives.
At a load of 840W the voltage ripple was 25 millivolts on the +5V rail, 39 millivolts on the +12V rail, and 28 millivolts on the +3.3V rail. I drew two markers in the oscillogram, at the maximums of pulsations on the +5V and +12V rails. It’s clear that the pulsations are not synchronized. This is a consequence of the use of fully independent PWM regulators on the different power rails of the PSU.
Both PSU models have excellent cross-load characteristics – the diagrams are almost all green. The certain spottiness is due to the influence of the high-frequency (400 kHz – we’ve never tested PSUs with a frequency of over 130 kHz) pulsation of the PSUs’ output voltage. We’ll add additional filters to our testbed for future tests.
Identical San Cooler 80 fans, the 9A0812S413 model, are installed in both PSUs, but in the junior model the fan would emit an irritating hiss at a speed of over 2400rpm while the senior model was surprisingly quiet, for its class. Perhaps there is just a defective sample of the fan in our ST75ZF.
The fan speed grows steadily up within a load range of 200 to 600W and then becomes constant at 3350rpm. Of course, there’s no talking about silence here. The whisper of the air passing through the PSU is perfectly audible, yet I had feared an even worse result. The ST85ZF is quite acceptable with its noise characteristics. The ST75ZF may be improved by replacing its native fan with another sample.
The efficiency of the PSUs got over 87% at a load of 200W and reached a record-breaking 80% in the end (the difference between the two models is less than 1%). At low loads, however, the efficiency degenerates catastrophically, to 50% and 51% at a load of 50W. Well, you will hardly buy an 850W power supply costing a few hundred US dollars to install it into a low-end computer while a serious gaming station with top-end graphics cards consuming 30-40W each even in idle mode can hardly have a power draw of less than 100W. At loads of about 150W, the Seasonic PSUs are a mere 5% behind the FSP models.
So, the two new power supplies from the Zeus series leave a nice impression on me, even though I can hardly imagine a computer that would require as much juice. If your computer doesn’t need that much, you may want to buy a simpler and quieter PSU and save some money at that. But in their class both the Zeus models are very good, with superb voltage stability, high efficiency, rather quiet operation, high quality of manufacture and the availability of all the connectors you may need.
I described the Zalman ZM460-APS in one of my earlier reviews, but then the model with the B suffix appeared and I have been asked to check out the difference.
This PSU is made by FSP Group. There can’t be any mistaking it although there is Zalman’s marking even on the PCB. This is the characteristic design of the FSPxxx-60GLN series. Moreover, there are markings on the PCB for wattages ranging from 250W to 500W, which can be assembled on the same platform, whereas Zalman is offering only one 460W model.
I have already described the circuit design of the new power supplies from FSP, also in this article, so I will only say that it is used in the ZM460B-APS, too. The only noticeable difference from the original PSUs is the considerably larger heatsink with diode packs (the leftmost in the photograph) and the slightly larger group regulation choke (the large coil to the right of the mentioned heatsink). The other two heatsinks have remained just plain aluminum bars.
The PSU complies with the ATX12V 2.2 standard and this seems to be its main difference from the ZM460-APS – the previous model was designed in compliance with the ATX12V 2.0 standard. The load capacity of the +12V rail has been increased, so it now can carry a current of 34A as opposed to the previous model’s 30A. The +12V rail is divided into two lines in the “virtual” way, of course.
The PSU is equipped with the following cables and connectors:
When the PSU was working with an APC SmartUPS SC 620, the UPS would report overload at a load of over 340W from the mains as well as batteries. I had no problems with using the power supply jointly with that UPS.
At a load of 438W the voltage ripple amounts to 23 millivolts on the +5V rail, to 25 millivolts on the +12V rail, and to 15 millivolts on the +3.3V rail. There’s only high-frequency pulsation here.
The cross-load diagram may not look as perfect as those of PSUs with independent voltage regulation, yet it is good anyway. The +5V is somewhat unstable, but it is going to fit into the “green zone” in a modern computer where there’s usually a load of 50W at maximum on this power rail.
FSP install fans from Protechnic Electric into its power supplies whereas the Zalman features a high-quality and low-speed Minebea-Matsushita 4710KL-04W-B20 fan.
The fan speed is regulated linearly from 634 to 1350rpm. Although there is no flat stretch in the graph in the area of small loads, the PSU is very quiet. The fan only reaches 800rpm at a load of 150W. Comparing the Zalman ZM460B-APS and the Seasonic S-12 SS-500HT, the latter is somewhat quieter at full load, but the difference is barely perceivable, especially considering the other noises in a computer that consumes so much power. So, both these PSUs are very quiet, almost silent.
Contrary to my expectations, the efficiency of this PSU is good, although not record-breaking, reaching 83% at the maximum. It is 80% at full load. The ZM460-APS had similar results, though.
So, there is in fact no big difference between the ZM460-APS and the ZM460B-APS. The second model has a slightly higher load capacity of the +12V rail, has a second graphics card connector and two more SATA power connectors. One of its heatsinks is larger, but this doesn’t affect the speed of the fan, which is the same as with the previous model. The PSU has good parameters and stands out among its competitors with its very quiet operation. The Seasonic S-12 is the only other model that can challenge it in this respect, so the ZM460B-APS will be a good choice if you need a quiet power supply. If you’ve already bought a ZM460-APS, you should not regret your having not the latest model because the B version is just a minor improvement upon an already superb power supply.
The two power supplies from Zippy, a well-known brand on the server market that has recently taken a dash into the home market too, are opponents to the Silverstone Zeus series described above. They have the same wattage and price. I will describe both models together as they are based around the same platform and don’t differ much from each other.
The PSUs are very large, longer than a standard ATX power supply. With all their impressive wattage they are definitely not to be installed into a compact system case. The components are densely packed inside the PSU case. There are four heatsinks in there with an additional 40mm fan on one of them.
Three out of the four heatsinks have contact with the top panel of the case. Two are fastened to it with screws and one more is pressed to it through a heat-conducting pad.
The fan’s airflow goes somewhat aside of the main PSU components. Anyway, the air stream from this fan is exhausted though the main fan grid, so the smaller fan indeed contributes to cooling the PSU.
Most of the “smaller” electronics is located on the cards standing upright along the other side of the PSU case. You can see a large transformer tied up with steel clamps. The choke peeping from behind it is large, too. Some reviewers pay special attention to the size of components, but a power supply is such a complex device with a lot of interconnected subunits and it would be wrong to examine those subunits independently of others. The dimensions of the transformer and chokes, for example, depend not only on the wattage of a PSU but also on its topology and the operating frequency of the PWM regulator. Here, this regulator is based on a ML4800CP chip and works at a frequency of 75 kHz, which is not high by today’s standards.
Besides the high current, up to 52A, on the +12V rail (by the way, the PSU doesn’t have any splitting of this power rail into several output lines), the PSL-6720P allows putting a high load on the +5V rail, up to 45A, which is higher than any other modern PSU can offer. The question is if anyone needs so high a current today.
The senior model has an 8A (96W) higher current on the +12V rail. Its overall wattage is higher by 130W, too.
The PSUs both offer the same set of cables and connectors:
In comparison with the 850W Silverstone, there are only two graphics card connectors instead of four. This matters for Quad-SLI systems (you can learn more about them in the Graphics Cards section of our site) as well as for ordinary SLI systems built on two graphics cards based on Nvidia’s new G80 GPU that has tremendous power consumption.
The senior model worked with an APC SmartUPS SC 620 at loads up to 350W irrespective of the power source. The junior model worked at 340W and 330W from the mains and battery, respectively. The switching to the batteries was performed without problems.
At a load of 850W the voltage ripple amounted to 21 millivolts on the +5V rail, to 70 millivolts on the +12V rail, and to 43 millivolts on the +3.3. Both high- and low-frequency pulsation can be observed here. The pulsation amplitude was 1-2-millivolt lower with the junior model at a 720W load.
The cross-load diagram looks very good. Both PSUs work without problems through the specified load range.
I have mentioned above that the PSUs came with two fans each, a Delta Electronics EFB0812SHF (a nominal speed of 4700rpm!) and a Sanyo SanAce40 109P0412J3023 (12,500rpm). I only measured the speed of the former fan because the second, 40mm fan proved to be too small for our optical tachometer, let alone being situated deep in the case (I test power supplies with a closed cover to avoid interfering with the thermal conditions of the PSU as well as to protect myself from splinters of transistors in case it fails the test).
The fans proved to be far not the best I had met with. The PSUs both produced a loud high-frequency whistle, much more annoying that the ordinary noise of the PSU fan. This noise is not only audible above the common noise of the system case, but even above the noise of my testbed that is equipped with four powerful fans and has never been intended for quiet operation. I have to admit that the Zeus PSUs are much more pleasing to deal with, even the model with a defective fan. Well, the speed graph that closely approaches the 4500rpm mark speaks for itself.
The efficiency of these PSUs (it coincides up to 1% between the two models) is yet another unpleasant surprise. It barely notched the 80% mark, which is very low for a power supply of that class, wattage and price by today’s standards.
Being comparable to the Silverstone Zeus in price (and being much more expensive than typical 600W power supplies), the PSUs from Zippy are inferior to them in other aspects, mainly in their noise characteristics. Their noise is loud and has an irritating spectrum that lies in the area of high frequencies. To put it simple, these PSUs produce a piercing shrill that is downright unacceptable not only for a home computer, but even for a workstation unless it is a server located in a special room where the noise factor is of the least concern.
So, I’ve led you through over a dozen new high-wattage power supplies that are selling under different brands, including such new brands (for this market) as Corsair and Mushkin (for details on the ANtec, Corsair, CoolerMaster and High Power PSUs see our previous article called ATX Power Supply Units Roundup. Part V). These are mostly products from a category of $100 and higher that we can expect good parameters and problems-free operation from, yet we don’t always get it in practice.
First, I have to rebuke Mushkin. The company enters the PSU market with a noisy model that has low efficiency but comes at a 10-20% higher price than its closest opponents that are free from such drawbacks. I guess that’s not a reasonable choice for a successful debut. I hope Mushkin’s upcoming PSU models will either be reasonably priced, according to their characteristics, or, better yet, acquire characteristics that would match their price.
Another name on my no-go list is OCZ, yet another memory module manufacturer that has taken to producing power supplies. Alas, its new PSUs (actually manufactured by FSP Group) have disappointed me with their noisiness and not very stable voltages. As I had already seen in my tests of the FSP Epsilon series, power supplies based on that platform can be very quiet at work.
And the last brand to be criticized here is Zippy. Once again this company has offered a prettily packaged server-oriented PSU as a home/games-oriented one. I have nothing against pretty-looking packages, but I do think that the level of noise typical of a server room is unacceptably high for a home environment. Zippy is opposed by Silverstone whose power supplies have comparable characteristics and price but are much quieter. Of course, I have to declare Silverstone’s products the best in the top-end PSU class.
I haven’t quite cleared things out with the Antec Neo HE 550. I’ve found two serious defects in it, but I tested revision A3.1 and cannot say if revision A4 has the same problems. Probably not, because users who reported poor compatibility with mainboards mentioned the earlier revision of that PSU model. So, don’t forget to check the revision number if you want to purchase this PSU.
The Antec TruePower 2.0 proved to be a very reasonable version of a midrange power supply, with good parameters and quiet operation, although the manufacturer doesn’t tout it as loudly as the Neo HE.
As for my favorites, they are Antec Phantom 500, Corsair CMPSU-620HX, Seasonic S-12 SS-500HT and Zalman ZM460B-APS. These four models all boast excellent operating parameters and are also very quiet (I’d even say silent if the Corsair model was not on the list).
The CoolerMaster iGreen Power might have joined that company, but its developers didn’t manage to cool it quietly despite its very high efficiency. Its fan speed is high at high loads. The High Power HPC-560-A12S has the same problem, too. Of course, it’s not difficult for an enthusiast to replace a fan with a quieter one, but I can’t call a product good if it requires you to make any improvements with your own hands.
And finally, the power supply from Hiper has proved to be good at work, but the company should think about modernizing its circuit design. Its efficiency of 75% is very mediocre by today’s standards.
