PSU #'s don't look right. - Page 5

mdocod, your observations are well thought out but you've made a mistake. The +5V and +3.3V loads are not subtracted from the labeled +12V rating. It is is assumed that the +12V has the additional capacity to take the load from the +5V and +3.3V as well as its native load. And conversion efficiency doesn't come into at all in any way. The +12V load in a DC-DC PSU is essentially labeled wattage divided by 12 and rounded. That is all. It's sometimes labeled lower, but for unfathomable reasons, perhaps just because the product manager in charge did not understand how DC-DC PSUs work. There are people in the industry who aren't very good at their jobs, you know.

Hello Phaedrus,

In the case of many units I believe you are absolutely correct. For example, in the FSP example I gave. The only way the numbers on the label agree with the laws of conservation of energy is if we "assume" that there is unlisted overhead on the 12V rail to power the lower voltage rails. It's very likely that the label was just brought over from a previous generation PSU that had a separate transformer and the whole old school works in the secondaries. I don't believe that the assumption [that there is unlisted but intended overhead] is necessary for all PSUs that use modern DC-DC topology for the secondaries. The Seasonic and Thermaltake examples both work out as labeled to a conversion efficiency from the 12 to 5/3.3 volt rails that is right in the ballpark of what I would expect for varying quality levels of decent buck regulation in this range. These units are both examples of a rating that needs no assumption of additional power on the 12V rail to meet the TOTAL rating given. In fact, to assume that more power on the 12V rail is required than what is listed, to meet the demands of the low voltage rails would have to come with the assertion of less efficiency than I have calculated above. Being familiar with DC-DC buck regulation ~90-95% is not unreasonable at these voltage ranges at all. The elegance with which the numbers resolve themselves with many of these labels suggests that they have in fact included the power for the 5/3.3V rails in the 12V rail rating accordingly.

In any case, with or without the assumption, I believe we can agree on the reality that in PSUs with this topology, it is very possible to have a scenario where the 12V rail has more power drawn from it than the combined total output rating, in order to achieve the output rating.

Best Regards,
Eric

you guys are insane, you know that??

thx for blowing my mind btw

Possible; but improbable.


Anyway, almost all high-end PSUs have ridiculous excess +12V capacity. A PSU with a +12V rating of 750W might have 1000W of +12V capacity, hardware-wise. This is for better efficiency, voltage regulation, transient response, durability, and peak load handling. It can handle an extra 50W from the +5V and +3.3V.

Hi Phaedrus,

I'm in full agreement about the available overhead otherwise. I was hoping to imply a condition that the power supply is loaded to it's rated capacity for the point I was making there. In any case, I am 100% in agreement with you about the typical excesses that *are* available, but may have a respectful disagreement on whether or not it's necessary to tap into those excesses to achieve label ratings in all cases.

Best Regards,
Eric