Originally Posted by PCCstudent
When I do not know I admit it.I will admit that I have never before scene the term "latency" applied to a power supply.I certainly have scene it applied quite frequently to data transmission and devices that get in the way of throughput and slow things down but in what sense is latency being applied to a power supply in this paticular case? Do not interpet my question as anything more than a question.I certainly cannot remember our OCN power supply guru speak of latency in regards to a power supply spec, but perhaps it fits in somewhere. Perhaps you are speaking of some defect in the power supplies rectification ability,I just do not know.
I did do a google search and it is just not turning up latency as a power supply spec.Now we all know that a google search is not the be all and end all of technical searchs.My best bet is something to do with the rectification process.
No offence, I have an electronics background, so that is why I get it.
Basically, ATX PSUs are made of two parts:
An AC-DC converter
And a DC-DC step-down regulator
The reason we call this a switching PSU is exactly that. From the high voltage DC regulated section, the DC-DC converter is literally switching on-off-on very fast (typically 25kHz or above). Then there is a buffer circuitry (capacitors and coils).
As the system drains more power, the voltage needs to stay stable. That means that the switching PWM (pulse width modulation, basically the ratio on/off) has to change to compensate.
Now, motherboard VRMs have a lot of sensors to make sure the voltage and current going to the PC is the correct one. Theses sensors are made to react in a mater of milliseconds, as the CPU and basically anything on the PC is very sensitive to over-voltage/over-current.
If the PSU cannot adjust fast enough (say 300ms, which is REALLY bad... a good PSU should be under 25ms), the VRMs and motherboard sensors detect an overcurrent just like there was a thunderstorm or a PSU failure. They will stop the PSU to protect the sensitive parts. In this case, when everything tries to start at the same time, there is a huge demand for current. That in turn causes the voltage on all rails of the PSU to drop a little. And because the PSU takes a long time to react, the current in the parts spikes (Ohm's law... voltage drop = current spike for the same load). That causes a lot more heat in small micro parts and can burn-out anything on the board.
Now, most components won't flash-melt because they are built to dissipate a given maximum power in heat over time (TDP.... that is a know value in CPU specs). But most small parts, especially the MOSFETs in the VRMs are way more sensitive and can only dissipate under 1W. If you push 2-3W in there, the temperature will double ever time value... that is bad for the parts. Common electronic parts have a max operating temperature of 75c, with military-grade/medical-grade usually going up to 175c (145c typical).
You can get a very similar behavior with a dead PSU. When you completely drain the power (remove power cord and hold power button 10 seconds), at the first try all the system fans will "kick" but nothing more. Any subsequent attempt won't do anything. That is the VRMs/BIOS blocking the PSU to prevent damage.