PSU question

You measure the wattage with a watt meter plugged into the outlet.

The efficiency of a PSU can have an effect on how much it draws. Poor low end efficiency will draw more power than a PSU will good low end efficiency.

300 watts or 600 watts. If the system only needs 200 watts to operate, the PSU with the higher efficiency will draw less power to run the system.

This all depends on the hardware setup, and how much power it draws.

I will explain more in another post below. I haven't seen the thread you are referring to, but it isn't possible to give a single efficiency figure, as I'll explain below.

The question about measure the power consumption, plug a power meter in between the wall and your PSU.

It is very important to understand that the efficiency of a power supply is not flat, but a dome shaped curve. Efficiency varies depending on the percentage of max power the PSU is working at.

The efficiency figures often quoted by manufacturers are the top of the curve - its maximum effiency. This often occurs at approxiamately 50-55% of the max rated load of the PSU. For a 500W PSU this would be around 250W load, and the effiency at this load would be around perhaps 85% for a fairly good one.

At 480W load (top end), and at 20W load (bottom end) the 500W PSU will perform at worse than 85% effiency. You will notice, if you look into the different 80 Plus gradings (ordinary, Bronze, Silver and Gold) that they specify different efficiency hurdles at different percentages of max load, so:

edit: This table hasn't come out neatly, because the forum is auto-deleting multiple spaces. I don't know the syntax to stop that.

20% 50% 100% of Max PSU rating

Ordinary 80 Plus 80% 80% 80%
Bronze 82% 85% 82%
Silver 85% 88% 85%
Gold 87% 90% 87%

You'll notice:
1) The drop-off in efficiency that 80 Plus expect at the top end is quite subtle
2) They don't specify any required efficiency level when the PSU is operating at less than 20% of max load. That is because the drop-off at the bottom end is usually quite sharp. 80 Plus accept that the performance of most PSUs is poor at much less than 20% of their capacity.

Point 2) above is not well understood, and people can make a mistake here. Wanting scope for extra GPUs, overclocking etc they go and buy a very high capacity PSU 'to give lots of headroom'. Worried about effiency, they find one with an 80 Plus Silver rating, perhaps a 1000W power supply.

The power consumption of a computer varies depending on whether it is nearly idle browsing the web, or worked hard to play the latest game.

At idle the computer might well only be drawing about 130W or less (even a midrange system if its upto date). Don't believe me? Check out the ATI Radeon 5750 system on this graph from a Tom's Hardware review:
http://www.tomshardware.co.uk/rade [...] 04-15.html

Now at just over 100W drawn from a 1000W PSU, that's a 10% load - not a level at which any PSU works optimally. Even for our 80 Plus Silver rated PSU, the effiency level may have dropped off to 70% or less at this point.

Now at maximum load (look back at the graphs on the review I posted) the midrange system might be up to about 200W or a bit less. If its silver rated then we've got up to 20% load so we should now be hitting 85% efficiency. BUT THAT IS MAX COMPUTER DRAW. How much of the computers life is really going to be spent at that kind of workload? Normally not much. So therefore in this example the computer is mostly going to be drawing at less than 20% of what the PSU can provide - and this is any PSUs worst performing area.

The best thing to do is to choose a power rating for your PSU so that it operates in its sweet spot around 50%. Look at the power consumption graphs within TomsHardware reviews of graphics cards like yours, with CPUs like yours, if possible. Make adjustments down where you know your CPU is less power hungry.

Take your guesstimated power draw at idle (say 150W) and your guesstimated power drawer when your computer's working hard (say 220W).

Now, if you get a 400W power supply, your computer will be drawing at 37.5% of PSU capability at idle and 55% of PSU capacity when working hard. The PSU would then be operating in its effiency sweetspot. So in this example you would restrict your search to PSUs of 400-500W rating, THEN WITHIN THAT you look for one with the best 80 Plus rating.

An appropriately powered PSU with a bronze rating will likely outperform a silver rated PSU with an excessive power capability, because in that case the silver is operating well below its sweetspot.

Where a PSU doesn't have an 80 Plus rating, but just quotes its max effiency (remember usually this happens at 50% of rated load), you may find that the efficiency curve is more 'domed', with a worse drop-off at each end. In this case it is even more important to get the right power.

Of course, err slightly on the high side with power rating, and do make allowances for up-speccing your machine, but only if you are serious about doing that in the near future.

That's my exhaustive guide to maxing your PSUs effiency with your particular computer. Its a bit long, I may edit it down later.

No. 1 is not exactly difficult. It just requires test equipment (a clamp-on ammeter) that is not commonly available.

For instance, I know that my Q6600 pulls 8 amps at stock speed and 9.5 amps when OC'd to 3.6 GHz through the CPU power plug. I took those measurements by pushing back the sleeveing and clamping the ammeter around the yellow wires.

Repeat for all the different cables and colors and total the results. For convenience, you need to do this on a breadboarded system. And it's tedious, but not particularly difficult.

Hey. I was curious one day.

Yes. I thought it was obvious that after you have the current demands on the various outputs that it was then a simple exercise to total the watts - not the current, then divide that by the input power reading to calculate the efficiency at that power point.

I just assumed that someone who is discussing efficiency at this level knows the difference between input and output and the difference between volts, amps, and watts.

Maybe I was wrong.