Rather than harping on, I’ll let this graph explain the situation.
The idle results were static and pretty close to their actual values which is good to see. When the tests were at full power, the rails began to vary by a large amount and were beginning to encroach on the 5% leeway given by ATX specification. However, this is a 520W power supply, and it’s not designed for 3 computers. That said, it is still within the spec and it managed to power the rigs flawlessly.
I was quite amazed by the results, especially with all the crap that manufacturers spout about NEEDING a horrendously expensive and powerful PSU to run their latest components. We tested the amount of power that the Mach 1 was sucking for the mains, and it maxed at 450W. Three graphics cards, four hard discs, two Physx cards and countless fans and 450W? 2000W power supplies seem a little silly. If you are wondering how much the slave PC power supplies we pulling, neither used more than 100W a piece.
The reason why manufacturers say that a high-end power supply is necessary is more to do with rail amp distribution rather than actual watt output. Much like clock speed isn’t everything with processors (except with Pentium 4’s of course), overall wattage means pretty much nothing. If you have a component pulling 12A on the 12v rail on a 2000W power
upply that only provides 100W on the 12v rail, but 900W on the 3.3v rail, then you’ll run into problems. Rail distribution means much more than overall muscle which is where some power supplies fall down compared to their lower overall wattage cousins.