Nice! Still pretty satisfied with where the X650 sits.

How is Pout() being determined for each PSU, though? Sorry (I've probably missed this), but is there a per-rail I*V sum being done for these figures? With what meter(s)?
Edit: and do the readings take phase into account, or not? I.e. is it really power or does it include some reactive load?

Thanks.

Phase? DC doesn't have phase, does it?

I've been making exactly the same graphs on a dutch forum!

One final step to help people pick a power supply is to divide the power lost through the purchase price of a psu so you get power lost per dollar you spend on a psu.

On these graphs, looking up the most efficient psu's is a disappointing task. They're either very expensive or not even atx size. With a power per prize graph, it's just a matter of picking the lowest line in the graph at your desired power range.

This is impressive, thanks CA_Steve !!
I noticed that the power lost goes down at lower loads for some PSUs (X400 for example, 19W at 150W and 16W at 200W). Is this still withing the margin of error or is there a logical explanation ? Just wondering...

You can read about SPCR's PSU test method here. The load tester uses resistors. So, any reactive load would be small and stray inductance/capacitance from wiring/components.

If only I could find somewhere to buy a lazer platinum in the UK I'd have looked at one already. Are Kingwin not selling in Europe (or at least not the UK)? As it is I guess I'm sticking with my ageing HX520. Given how long i've had it, I'm impressed it's still regarded relatively well.

Yeah, there were a few that did this little shuffle. Here's the possible reasons:
- margin of error in testing
- it's baked into the PSU design
- data entry error in the original review
- data entry error by me

Hopefully, I've screened out my data entry errors. Plotting the results is a great way to see if something's wacky.

Super Flower mfgrs the Kingwin branded PSU. The comparable model is the Golden King Modular.

Silly me! I was stuck in AC mode. I guess there could be inrush/ripple/switching characteristics in a real system, but they would have to be negligible..

The test setup looks awesome. Feeling lab envy.

There is a specific procedure used for 80 Plus testing to proportionally load the various voltage rails of the PSU to simulate PC load conditions. SPCR follows this as do many other sites. Some opt to roll their own - part of the many reasons why you can see variance of results across sites for the same model. That said, my guess is resistive load testing is going to give better results than what we might see in a real system.

I'm hoping to have the time to swap out the 520W Bronze PSU in my signature for an X-560 Gold in the next day or two and plan to take some before/after measurements with specific loads (idle/Hulu-flash/Netflix-silverlight/gaming). I'm curious to see how well the delta power tracks vs a test setup.

Very nice article!

Figure 6 does not link to a larger version of the graph like the others.

And the recommended PSU list hasn't been updated since February, many PSU reviews since then. It would be nice to see them added to the list.

Nice!

Also: bad link at the last pic:
http://www.silentpcreview.com/files/ima ... oad600.png

should be

http://www.silentpcreview.com/files/ima ... c-load.png

I approve this is roughly how power supply design works as well; the actual efficiency number is not much more than an afterthought because component dimensioning and balancing performance has to be done with physical quantities: voltage, current, power, energy. Within the world of topology design and power supply research two oft-used tools to graphically gauge losses are the proportional diagram (representing components of power consumption as proportionally larger bars/areas/volumes in a 1/2/3-D plot) and Sankey diagrams (depicting energy flow and losses as arrows). Here's two examples from my prize-winning (I love saying that) sub-6W computer. First a proportional diagram:



And a Sankey diagram:



What you've done in this article is take a useful subset of this methodology and apply it to the sum of PSU losses, which as I explained is actually pretty obvious to PSU designers, but seems to be 'new' knowledge in the review world.

CA_Steve is doing WORK. Nice job!

Some of the results are not exactly surprising, but it's good to see a bigger picture than one traditionally gets. I've always diligently done my math or looked up reviews with the appropriate figures when picking PSUs for builds I've been involved with, and it's always a multiple-step operation of calculating the loads, calculating PSU efficiency at a given load and then comparing the PSUs at given loads. When there are ready comparisons of PSUs at a given wattage, things are much simpler, as you don't have to do the calculations yourself - and possibly mess up (although with Excel sheets it's pretty hard to do so).

The article convinced me that power loss is a good metric, especially in the case of rather small margins and high efficiency units.

I'm glad to see the discussion and good information in this thread too, it's about time this aspect of silence basics gets discussed in detail too.

Very nice!

If I may make a request; if you use this metric in future reviews, could you pick just a few (3-4) other PSUs for comparison to make the graphs easier to understand?