CRush1682 :
2. A PSU is "happiest" running at approximately 80% of it's total capacity. What does that actually mean? That's where it's most efficient, and will actually COST YOU LESS! That's right, I actually tried this test myself. I have a P4 2.66Ghz, Supermicro Socket478 Server board (7230) with 4 sticks of DDR400 RAM and 6 HDD's (4SATA, 2IDE). Running the stock 550watt Antec TruePower PSU that came in the Titan case actually drew more wattage from my UPS then a 350watt Antec Smart Power! Why? Because it was being waaay under-utilized. The 350watt also ran cooler because it was more efficient. (efficiency is basically a measure of how much energy is lost in a transition from one form to another. In this case electricity is lost as heat as it is converted).
Comparing two different models of PSU isn't a valid test. Yes when all else is equal the smaller capacity unit might be more efficient but it depends a lot on what wasn't upgraded versus the higher wattage model. With very similar models there is a modular nature such that many parts are the same just some key component changes.
You should not try to run a PSU at 80% of rating even if the PSU is honestly rated. The reason is longevity, PSU aren't built towards extreme budgets and a PSU designed capable of handling more current (all else equal) tends to have better heatsinking and/or airflow, lower impedance capacitors to yield longer life and lower ripple, and overall more robust design in any area close to it's limits.
If your system is using 80% of, let's say a 350W PSU since that was the example previously, and that PSU were 82% efficient, then we have (0.8 * 350)/.82 = 341W total power consumption.
Next let's suppose another modern 80-plus PSU but the 500W model, this time it's barely making 80-plus due to running at only (.8 * 350) / 500 = .56, 56% capacity.
So then for the 500W PSU it's (.56 * 500) / .80 = 350W total power consumption, only 9W more than the 350W PSU and only 9/350 = 0.025 or 2.5% more power.
That minor difference in power has to be weighed against replacement cost when the 350W fails earlier, because it will fail earlier if all else is equal. If we wanted to be green about it, that also means more manufacturing pollution to build a 2nd PSU to replace the first, and landfill throwing away the failed PSU. Most (consumer grade, PC) psu running at 80% load will not last the life of the system.
In an idea world, and maybe at Nasa when cost is no limit, things are different than in the consumer products seen in the PC industry. Even a PSU like you'd find in a $20K copy machine is vastly superior to what's in a PC when it comes to build quality, and certainly not as good as you'd put on a space probe. The manufacturers are rating for peak power at unreaslitically low ambient temps except for a small handful, if your PSU says it can handle 350W at 20C ambient temp, you're looking at having a colder than normal room in order for the PSU intake to be at 20C inside a case with other parts producing (as per the example, 350/0.8=) 280W.
CRush1682 :
4. If anyone can come up with info to the contrary, I would be more than happy to listen...I mean hey...I only got my info from a guy who designs power supplies for outer-freaking-space...he's probably just a high school drop-out who doesn't know anything. :kaola
What he may not know is about the actual differences in PSU available for purchase. Given this info and test data as he runs on his designs, his conclusion would not be as quick and simple as your summary. There are always tradeoffs and one is how much power is saved as we aren't running from battery power and solar cells, for instance.
The requirements for a power supply in outer space are different as is the goal to build supporting power for millions to billions of dollars worth of equipment. He also has no need to design for an overclocked processor and video card, he never has anything as demanding from an extremely cost contrained build budget. NASA/etc don't run anything that only peaks at 300W or so power consumption but can have the requirement for as fast a power state transistion as an overclocked Quad Core Duo and a gaming video card. NASA was using 80486 processors for a lot of their critical stuff the last time I heard, and with good reason that a very robust PSU can be built using small solid caps in the PSU instead of having electrolytics drying out in space and subject to more temperature sensitivity. Beyond processors and equivalent logic, nothing else has a requirement for that fast a transition in current or else it will crash.
It's really apples and oranges, he may be great at what he does but that doesn't mean he could work magic given $15, some string and some glue (kidding about string and glue but really there are always corners to be cut arriving at PSU with desirable prices, if it can't hit a desirable price point the volume sold goes through the floor and the price has to rise very quicky to make it a profitable venture for manufacturers. That's why a 750W PSU costs so much more than a 300W, as the parts differences certainly aren't accounting for a typical tripling in retail selling price.