> The power output rating is calculated by increasing load current until the voltage drops unacceptably, then calculating the power output at that point.
It really isn't. Sure, the engineers who design the power supply do that test, but that is not at all what determines the advertised rating of the final product. At 100% of the advertised load, most computer power supplies are still delivering nominal voltage or slightly above, not 5% below as allowed by the ATX spec.
> Efficiency is very flat in switching power supplies once load gets above about 10% of rating. [...] A 750W power supply is just as efficient running at 250W as it is at 750W.
Yes, on either side of the peak efficiency, you'll have points of equal efficiency. And in the middle, you'll have a few percentage points higher efficiency. But more importantly, at the ~30W a typical desktop will actually be drawing most of the time, a power supply with a smaller rating will be substantially more efficient.
Buying 750W and larger power supplies just doesn't make sense for single-CPU, single-GPU systems. Power supplies with lower ratings already have plenty of headroom both built-in to their rating and in the difference between 500W and what a real desktop actually uses on real workloads. To the extent that having excess capacity helps longevity, a 550W or 650W model is already well past the point of diminishing returns and going up to 750W is pure vanity. If you want reliability, shop for PSUs that use high-quality fans and capacitors, don't just stupidly add an extra 30% on top of what's already for more PSU than you really need.
> It really isn't. Sure, the engineers who design the power supply do that test, but that is not at all what determines the advertised rating of the final product. At 100% of the advertised load, most computer power supplies are still delivering nominal voltage or slightly above, not 5% below as allowed by the ATX spec.
I assumed power supply manufacturers would want to slap the peak power number on their supplies for marketing purposes. If they are actually being conservative, then you are correct.
> Yes, on either side of the peak efficiency, you'll have points of equal efficiency. And in the middle, you'll have a few percentage points higher efficiency.
My point is that efficiency is a plateau, not a "peak". Once in the plateau the fluctuations of efficiency from one load point to another are not significant. Below a certain minimum load and past the peak power "knee" is a different story, but that plateau is very large.
> But more importantly, at the ~30W a typical desktop will actually be drawing most of the time, a power supply with a smaller rating will be substantially more efficient.
I never disputed this. I disputed the nonsense that power supplies have a meaningful "peak" efficiency, and that it is a function of it's rating. 30W is probably not enough of a base load for efficient operation of larger computer power supplies, but once that point is hit it no longer matters what the actual load is.
I'm also not suggesting it is a good idea to waste money on a larger supply than you really need.
It really isn't. Sure, the engineers who design the power supply do that test, but that is not at all what determines the advertised rating of the final product. At 100% of the advertised load, most computer power supplies are still delivering nominal voltage or slightly above, not 5% below as allowed by the ATX spec.
> Efficiency is very flat in switching power supplies once load gets above about 10% of rating. [...] A 750W power supply is just as efficient running at 250W as it is at 750W.
Yes, on either side of the peak efficiency, you'll have points of equal efficiency. And in the middle, you'll have a few percentage points higher efficiency. But more importantly, at the ~30W a typical desktop will actually be drawing most of the time, a power supply with a smaller rating will be substantially more efficient.
Buying 750W and larger power supplies just doesn't make sense for single-CPU, single-GPU systems. Power supplies with lower ratings already have plenty of headroom both built-in to their rating and in the difference between 500W and what a real desktop actually uses on real workloads. To the extent that having excess capacity helps longevity, a 550W or 650W model is already well past the point of diminishing returns and going up to 750W is pure vanity. If you want reliability, shop for PSUs that use high-quality fans and capacitors, don't just stupidly add an extra 30% on top of what's already for more PSU than you really need.