X-bit labs

ATX Power Supply Units Roundup: Part II (page 2)

Category: Other

by Oleg Artamonov

[ 04/25/2005 | 03:45 PM ]

It’s clear that an adapter from the 20-pin plug of the power supply to the mainboard’s 24-pin connector cannot solve the problem of burned-out pins: just different pins are going to overheat. The only proper solution is the use of a power supply with a native 24-pin connector. It’s of course possible to make an adapter that would attach to two PSU connectors – to the main ATX connector and to a Molex connector where the adapter could take two of the missing lines, that is +5V and +12V (the +3.3V line wouldn’t be available even here, though). This would settle the matter, though partly, but I haven’t yet met such adapters. All solutions I know don’t use any additional connectors.

Besides graphics cards, the power consumption of central processors is also growing ever higher. And CPUs load the +12V power rail, too. It seems they can just increase the wattage of the power supply together with the maximum allowable current on the +12V rail, and that’s all. It’s even simpler than increasing the current on the +5V rail (and allowable currents on this rail exceed 40 amperes on some PSU models), but the developers confront the safety regulation EN 60950 which says that the maximum power on user-accessible contacts (the outputs of a power supply belong here, of course) cannot be higher than 240 volt-amperes. Thus, it’s impossible to provide a current above 20 amperes on the +12V output without violating this regulation.

How did the engineers solve this problem? They wind the secondary +12V winding of the transformer depending on the required load current; let’s suppose it is 30 amperes. The rectifier’s diodes and the throttles are installed depending on this total current, too. But after the rectifier the +12V rail is split into two wires, each of which has independent protection against current overload with the threshold set to, say, 15 amperes. This slight redesign of the power supply (the protection circuitry is quite simple) allows achieving a higher output capacity on the +12V rail (for example, one wire feeds up to 15 amperes to the CPU, while the other powers up the graphics card, hard drives, etc that can consume up to 15 amperes in total, too; the total consumption of all devices on the +12V rail can reach 30 amperes in this case), meanwhile complying with the safety regulations because the protection works as soon as there are more than 12 amperes on any of these two PSU outputs.

Summarizing this briefly, I can put it like that: new power supply models have one internal high-current +12V source that looks like two sources to the user. The load characteristic of the PSU depends on the capabilities of that internal source rather than on the artificial limitations on the outputs the user has access to. You will see below, in the description of the power supply from CoolerMaster, why I put so much emphasis on this fact.

The last thing to be discussed in this introduction is the efficiency factor of power supplies. Among other parameters we do measure the efficiency in our tests, so our readers ask what exactly efficiency value should be considered good. Besides the obvious rule “higher efficiency is better”, there are requirements set down by the industry standards. According to the latest specification, ATX12V 2.0, the power supply is required to have an efficiency of no less than 70% at full and half load and no less than 60% at a load that equals 20% of the PSU’s capacity (these requirements had been somewhat less strict in the previous version of the standard). The recommended efficiency values are 75% at full load, 80% at half load, and 68% at 20-percent load. All manufactured power supplies meet the requirements, but not all of them meet the recommendations. Not all units can boast an efficiency of 80% (but some models can exceed this value considerably).

I would also like to note that I perform my tests at 220 volts on the PSU’s input and this voltage is maintained quite accurately. If the input voltage goes down or if you switch to an 110V power grid, the PSU’s efficiency factor will decrease due to the higher loss in its high-voltage circuitry. But the difference is going to be small, a few percent at most, because the main loss falls on the PSU’s low-voltage circuitry (more exactly, on the rectifier’s diode assemblages) that don’t depend on the PSU’s input voltage at all.