by Oleg Artamonov
05/16/2006 | 12:31 PM
It’s not a secret that PC components are constantly getting hungrier for power. Despite periodic remissions due to transitions to new semiconductor manufacturing processes (for example AMD’s processors began to need considerably less power on a transition from 130nm to 90nm) or to changes in the micro-architecture (compare the consumption of the Intel Core Duo with that of the Pentium D which has roughly the same performance), the overall tendency is clear: the tough market competition forces the manufacturers to seek for more speed before they develop a new tech process, not to mention a new micro-architecture. And in a developed micro-architecture, there are only extensive ways to improve performance: by increasing either the clock rate or the number of units working in parallel (pipelines, ALUs/FPUs, the amount of cache memory, etc). Obviously, either way leads to an increase in power consumption.
As a result, there has never been any long period of time throughout the entire history of personal computers that power consumption would have been lowering and I think this tendency isn’t going to change in near future. Perhaps a transition to molecular bio-computers will effect the change, but from today’s perspective it looks more like science fiction rather than anything to come anytime soon. A few years ago hardware reviewers were testing the then new AMD Athlon 1.4GHz with a TDP of 72W and were all wondering at its fantastically high heat dissipation. Today you can’t surprise anyone with such a number – the TDP of a modern CPU is long over 100 watts.
The same is true for the graphics card market, too. The 3dfx Voodoo2 used to get along without additional cooling whatsoever (only overclockers would put small heatsinks on its chips). Early GeForce 256 cards consumed 25-30W and were quite satisfied with small fans for cooling. Today’s graphics cards may draw over 100 watts of juice and their coolers have long transformed into monstrous contraptions, two slots in height.
The requirements to the power supply have also been growing up. A 145W model would be quite enough at the times of the Pentium M. Today, the declared wattage isn’t simply impressive but downright shocking – the maximum is 1.1kW whereas 400W models are the mainstream. Although many users have a somewhat vague notion of how powerful a PSU they need (they are confused by low-quality products whose real wattage is much lower than the real one as well as by the marketing departments of PSU manufacturers who use wattage as the most intuitively comprehensible characteristic to promote their products), you still can’t deny that the requirements to the computer power supply are ever increasing.
PC enthusiasts have long been ruminating the idea of using two relatively low-wattage power supplies. In the simplest implementation one PSU is responsible for the graphics card, processor and mainboard while the other for various peripherals (hard drives, optical drives, etc). This is simple to do because you don’t have to modify the PSUs much. It’s only necessary to connect their ground wires and PS_ON contacts (the main regulator is turned on at a signal on this contact). On the other hand, there’s not much sense in powering the peripherals from a separate PSU because their share in the overall consumption is too small compared with the combined consumption of the graphics card and processor.
Another approach is to separate the different supply voltages between the different PSUs. E.g., one PSU only provides +12V, and the other, the rest of the voltages. Both the PSUs should be modified so that the regulation in the former was bound to the +12V voltage and in the latter to the +5V (ATX units usually regulate both these voltages simultaneously, basing on some average value). The PSUs can be left as they are, but the modification is advisable – the output voltage may turn to be not very stable because only one of the PSU’s power rails is under load.
So, in this case the two power supplies can give you as stable voltages as you can possibly get, but there are two considerable drawbacks. First, you must have the skills necessary to modify the PSUs on your own. Second, the most voracious components of a modern computer are supplied from the +12V source, so one of the units is going to bear a much higher load than the other.
Of course all these improvements and modifications are for hardcore enthusiasts because ordinary users would find it simpler (and even cheaper, considering the time spent) to buy a new, high-wattage power supply than to take the trouble of fitting two PSUs to their computers. And you have to think about where to put the second PSU in the system case, too! Well, the industry has already made a step towards the enthusiasts: the Cooler Master Stacker case has a place for a second power supply and an adapter for sending it a turn-on signal in parallel with the main PSU.
The same industry has involuntarily made another gift to people who’d like to use two power supplies. Every high-consumption component is now powered from the +12V source and the +12V power rail is split in several PSU outputs. As a result, the graphics card and processor have acquired their own external power connectors that are not electrically connected to each other (the CPU connector is located on the mainboard – it’s the well-known 4-pin ATX12V).
What does it mean? In old times when the ATX12V connector already existed, but PSUs had only one +12V rail, there was no reason why this connector shouldn’t be connected to the appropriate contact (the +12V yellow wire) of the mainboard’s main 20-pin connector. More contacts minimize resistance and power loss. However, this meant that if you tried to power the CPU from a separate power supply whose output voltage differed from the main PSU’s voltage (and they are never going to be exactly the same, of course), you got a short circuit through the mainboard with all the consequences.
The connectors had to be separated on the mainboard after the division of the PSU’s output lines had been made: it would be illogical to divide anything in the PSU only to join everything back together on the mainboard. So it has come to be that the 12-volt CPU power connector is not electrically connected to the 12-volt contacts of the mainboard itself. And you can apply different voltage to them, from different PSUs, without fearing any problems.
The same is true for graphics cards after they have acquired dedicated power connectors – first of the 4-pin HDD or floppy variety and then special 6-pin ones (and the power consumption of graphics cards has grown dramatically, too, up to 100 watts and higher).
And now the idea of using multiple power supplies has emerged again. On one hand, there appear graphics card models with dedicated power supplies. For example, the power connector of the ASUS EAX 1800XT TOP is placed on an external bracket and is connected to an external power supply. This solution helps unload the main power supply of the computer, but has two drawbacks. First, there are only a few graphics card models that come with an additional PSU. Second, if you use a SLI or CrossFire system, you have to power one of the cards from the main PSU or to put up with two additional PSUs.
The solution suggests itself: an additional 12V power supply that would connect to any graphics card and would easily support two graphics cards at once (or even four cards because Nvidia is actively promoting its quad SLI technology).
Two companies have come up with such products this spring. Thermaltake was the first with an official announcement, but FSP Group’s product was described in detail on the Web a few days earlier, although without yet being officially released. Anyway, we got a FSP VGA Power from Fortron/Source for out tests first.
As its name suggests, the FSP VGA Power is a power supply for graphics cards. It is to be inserted into a 5.25” bay of your system case. It lacks any controls, not even an On/Off switch. There are only a few vent holes in its front and a decorative transparent panel with the device’s name.
A 220V power connector, a 12-pin input connector and a 4-pin Molex connector are at the device’s back. The latter is an input connector. One of the Molex plugs of the main PSU is connected to it and the VGA Power turns on as soon as it senses voltage in this input. This solution makes the operation of the additional PSU fully automatic (it is turned on and off automatically without the user’s doing anything about it) and is better than connecting, via an adapter, to the mainboard’s connector to get the PS_ON signal which is sent to turn on the main PSU when the computer is started – extra adapters can’t do anything good, but only lead to extra power loss (Thermaltake’s solution uses the signal from the mainboard, by the way).
And this is how you connect the PSU to the electricity mains: the VGA Power comes with a back-panel bracket the power cord is plugged into. You can do without the bracket, though. The connectors are all identical, so you can plug the cord right into the PSU, without the additional adapter. You can only meet some problems if you’ve got a UPS: you’ll either have to make an adapter with your own hands or to replace the plug on the included adapter to connect your UPS to the VGA Power.
A cable with two 6-pin graphics card connectors is included with the VGA Power to connect the load to it.
A load capacity of 300W is declared for the VGA Power. Its single output voltage is +12V (the maximum current is 25A then). The PSU features active PFC and supports a full range of input AC voltages without any switches.
The VGA Power is designed internally just like an ordinary switching power supply, although it differs from the standard ATX power supply, of course. There is only one output voltage, so there is no need for a group regulation choke. A line filter can be seen in the top part of the photograph above (three coils with capacitors in-between) and an active PFC coil (the leftmost one). In the bottom of the photo there is a power transformer lying on a side and a large high-voltage capacitor (it only looks so large from above – they had to use a capacitor with a big-diameter case due to the height limitations).
Diode packs of the output rectifier are installed on the largest heatsink. The other two heatsinks carry switch transistors of the main regulator and the components of the active PFC device. A small heat-spreader, actually a small strip of aluminum, is fastened on the diode bridge on the PSU’s input.
Most of the integral electronics, i.e. electronic chips, are placed on the reverse side of the PCB, so the power supply is not as simple as it looks when you view it from above. The VGA Power has its own standby source. In the ordinary ATX power supply it generates the +5V SB voltage for the mainboard. Here, it is used for the PSU’s own purposes.
Generally speaking, the VGA Power is not something simple. Its complexity isn’t much lower than that of standard computer PSUs.
The PSU’s front panel is highlighted with four blue LEDs at work. The highlighting is bright and I think it would be better if the LEDs were fully hidden in the case to highlight just the name of the device rather than to shine forward like four small electric torches as they do.
I was of course interested in how noisy the VGA Power would prove to be at work. It is cooled with two 40mm fans built into its front panel (there is a dust filter in front of the fans; it cannot be removed, but you can clean it from the outside with a vacuum cleaner when necessary). The speed of the fans is automatically varied depending on the temperature of the heatsink with the diode packs.
Unfortunately, the fans don’t have an integrated velocity sensor while our optical tachometer isn’t very correct with objects of so small a size, so I had to limit myself to measuring the voltage of the fans. As you see, it does depend on the PSU load (on its temperature, to be exact) and is rather high, over 7V, even at the min load.
As for my subjective impressions, the VGA Power is noisy. Its sound can be heard even at small loads. At loads from 150 to 200W the VGA Power can well become the noisiest component in your computer. The position and the type of the fans only make the things worse. They are at the front panel of the system case, so the latter doesn’t suppress their noise at all. Small and working at a high speed, the fans produce a characteristic high-frequency hiss that rises above the rest of the system noises.
On the other hand, if you use graphics cards with standard cooling solutions, the fans of the VGA Power won’t be distinctly heard as their noise is similar in loudness and spectrum. But if you use quiet coolers from Zalman, Arctic Cooling or other manufacturers, you will hardly like the VGA Power.
Otherwise the PSU leaves a nice impression. The output voltage is almost ideally stable, varying from 12.07 to 11.96V at loads of 50 to 300W. This is not surprising if you remember that the problems with stability of the output voltages of ordinary ATX power supplies are all due to the fact that such PSUs have to be seeking a balance between their multiple output voltages. Here, there is only one output voltage.
I had apprehensions that the VGA Power might be too hot and dangerous for the nearby optical drive. The diagram above shows the temperature of the hottest spot on the cover of the VGA Power case (if was not installed in the computer, but was lying on a desk; the air temperature in the room was about 22°C). If your system case permits, it is better to leave one bay free above the VGA Power because even though its cover is not used for cooling (besides an air gap there is a thick insulating pad between the cover and the heatsinks), its temperature is over 40°C at max load. So if the VGA Power is under high load, and the neighboring DVD-drive is working with a disc (like when you are playing a game that is regularly reads its data from the disc), the drive may feel not quite well. Modern high-speed optical drives heat up a lot just in cramped cases and heating them up from below won’t make things any better…
The output voltage ripple at full load was about 90 millivolts (73 millivolts of a high-frequency, 285kHz pulsation and the rest is a low-frequency, 100Hz pulsation) which easily fits in the acceptable range (120 millivolts).
The efficiency of this power supply is a little below 85% at high loads. I had expected a better result considering the simpler design in comparison with standard PSUs (particularly the lack of a group regulation choke which is rather hot at work) and the high pulse-width modulation frequency. The power factor is quite standard for a PSU with active power factor correction and reached 0.99.
The main question is whether the user actually needs a VGA Power. The parameters of the PSU are within the acceptable limits as I have checked in the previous section, so it will work correctly in a computer. But what load can it take over from the main PSU? How many cards can it power up? What else can you do with it?
In most hardware reviews the total consumption of a graphics card is mentioned because when the whole system is powered by a single PSU, it is the total consumption that matters. It doesn’t matter which connector the current flows through because this doesn’t affect the amount of dissipated heat or noise.
But there are two power supplies in our case. The graphics card has two connectors, too. The PCI Express slot it is plugged into not only gives it control signals, but also power, up to 75 watts. By the way, graphics cards with a maximum consumption of 50W and less are not even equipped with an additional power connector lately because the PCI Express slot alone gives them more power than they need.
I asked my colleagues from the Video lab to additionally measure the power consumption of today’s top-end graphics cards, dividing it into the consumption through the PCI Express slot and through the additional power connector. Here are the results:
So you can see that graphics cards do consume up to 50 watts of power from the PCI Express slot (and you may also note that the weakest of the tested cards – GeForce 7900 GT – doesn’t in fact need external power). Everything above that is consumed from the additional power connector. The consumption from the +12V rail is only shown in the diagram because it bears most of the load; the share of the +3.3V rail is negligible.
The VGA Power can only be connected to the graphics card’s additional power connector whereas the mainboard and the PCI Express slot are powered by the main PSU. So you shouldn’t expect the VGA Power to fully unload the main PSU as if there were no graphics cards at all – if you’ve built a SLI or CrossFire subsystem with two top-end graphics cards, it will consume as much as 90-100 from the main PSU.
The maximum load on the VGA Power that can be achieved in a modern system is a little less than 150W (if you’ve got two Radeon X1900 XTX) or half its maximum load capacity.
It means that:
Here I’d want to return to the beginning of the article where I said that on modern mainboards the CPU power connector is electrically separate from the mainboard’s main connector, and another power supply can be connected to it without any problems (of course, you should check this out on each particular mainboard with a multimeter – if there’s a near-zero resistance between the 12-volt contact of the ATX12V connector and the 12-volt contact of the main connector, they are linked together). But if the VGA Power is redundant even for a SLI system, why can’t it also be used to power the CPU? What you need is a new cable that can be made in 15 minutes if you’ve got just the basic skills.
The VGA Power from FSP is an interesting, but rather specific device. It is mainly targeted at PC enthusiasts and overclockers (who can never have quite enough of PSU wattage), at potential owners of the upcoming quad SLI, etc. On one hand, the VGA Power meets competition from ordinary, high-wattage PSUs. FSP Group itself turns out 700W models that should suffice with a big reserve for any modern computer and are going to be quieter with their 120mm fans than the VGA Power. But on the other hand, the VGA Power can guarantee excellent voltage stability with a deflection of ±0.5% from the norm across the entire load range. I guess some overclockers are going to appreciate this characteristic and the fact that the VGA Power is a ready-made solution.
The VGA Power may also be interesting for such a small group of users as owners of barebone systems and other exotic system cases in which the power supply cannot be easily replaced with a higher-wattage one due to its non-standard form-factor. But there are not so many such users, especially if you count in only those of them who want to install such a processor and graphics card that the native PSU cannot cope with. And quite a lot of such systems just don’t have a free 5.25” bay. And even if they have, it wouldn’t be good to install the VGA Power close to the DVD-drive because the case of this power supply is very hot at work, as I mentioned earlier.
Ordinary users will hardly be interested in the VGA Power. For them, replacing the main PSU is going to be a better solution than purchasing an additional one, especially since a new 400W PSU (for example, an ATX-400PNF from FSP) costs less than $40 and the VGA Power will hardly cost as much less as to justify the purchase. The ATX-400PNF can not only power up almost any computer configuration (at least, the owner of a configuration it cannot cope with surely has enough money to buy an even better PSU), but is also going to be much quieter than a VGA Power plus an old, low-wattage main PSU.