Efficiency vs load analyzed for 20 PSUs

PSUs: The source of DC power for all components in the PC & often a big noise source.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee, Devonavar

Post Reply
CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Efficiency vs load analyzed for 20 PSUs

Post by CA_Steve » Mon Oct 22, 2012 10:10 am

The discussion in the Super Flower PSU thread on 80+ methods showing results as percentages rather than heat loss and the tradeoff of cheap vs. efficient got me thinking. I spent some time data mining the SPCR PSU articles, throwing the test results back into Excel for some chart-fu.

First some caveats:
- Ignore differences of a few watts. It’s probably within the variance of the test equipment accuracy, the test setup, test repeatability, the phase of the moon, and how many craft beers Mike C and the other intrepid reviewers have consumed during the testing and editing process. :D
- These results span from Aug 2006 to Oct 2012. I’m sure there have been changes in the test setup during this time.
- Any discussion of power use is the DC power supplied by the PSU, not power at the wall socket.
- Curve fitting. Take it with a grain of salt.
- I’ve probably made a data entry error or two.

First up, the 14 most recently reviewed PSUs: 360W to 1050W, Bronze to Platinum, in one big eyesore of a chart. Data is plotted as power lost vs. power out. It’s a blob at 300W and below, but look at the higher loads and the variation in power loss.
Image

Next, the same data zoomed in to look at typical idle, low CPU utilization and low power configurations. A new ATX desktop build can idle in the 20-25W range. These PSUs vary in efficiency up to 10W. Even going up to 60W load, the variance stays within 11-12W. A couple of things pop out:
- You can see the efficiency hit you take for using a higher power PSU than you need for low CPU utilization tasks.
- You can sorta see how low power efficiency has improved over time. Extreme example: Silverstone ST50NF with 16W power loss at 20W Power output vs. Kingwin LZP-550 with 6W loss @ 20W out.
Image

Here’s a chart for high CPU utilization and typical (single graphics card) gaming builds. Some figures of merit: An i7-3770K system running Handbrake will be in the 100W range, a PC with a midrange GPU (HD 7850/7870, GTX 660/660 Ti) playing at 1080p can run from 100W to 250W. For this range, the variation in efficiency widens up. At 100W, the spread is 13W. At 200W, the spread is 20W. At 300W, it’s 28W.
Image

To get a better look at efficiency improvements over time, the next series of charts compares all of the Seasonic PSUs reviewed (since 2006). Bronze vs. Gold stands out.
Image

Zooming in to idle and lower power use, Bronze parts (S12, M12, Corsair HX) have idle power loss of 10 to 22W. Big gap between the 430W and the 520/620W. The 350-650W Golds stay within a ~5W spread at idle as well as over the 20-60W power range.
Image

For high CPU utilization and gaming, Bronze vs. Gold is pretty clear cut. The average Bronze rated PSU uses 10W more at 100W output, 15W more @ 200W, and 25W @ 300W.
Image

I may do a bit more chart-fu later on. Have fun discussing!

bastiaan
Posts: 98
Joined: Fri Oct 19, 2012 4:23 am

Re: Efficiency vs load analyzed for 20 PSUs

Post by bastiaan » Mon Oct 22, 2012 11:01 am

These charts are awesome, especially the low power ones.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: Efficiency vs load analyzed for 20 PSUs

Post by MikeC » Mon Oct 22, 2012 11:28 am

Much thanks Steve! Those tables are worth running in a main SPCR article. We'll talk later.

bastiaan
Posts: 98
Joined: Fri Oct 19, 2012 4:23 am

Re: Efficiency vs load analyzed for 20 PSUs

Post by bastiaan » Tue Oct 23, 2012 2:16 am

It's very interesting that the Kingwins (both in the 500W range) do so well at low loads, especially the LZP-550; the graphs seem to indicate that it's the most efficient power supply for loads below 300W, even beating the PSUs rated for 300W from other brands. On the other hand it might be unfair to draw conclusions like that from the few measurement points indicated.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Efficiency vs load analyzed for 20 PSUs

Post by CA_Steve » Tue Oct 23, 2012 7:23 am

One of the many reasons why I've been recommending it for gaming builds.

Esben
Posts: 83
Joined: Sun Nov 11, 2007 11:35 am
Location: Denmark

Re: Efficiency vs load analyzed for 20 PSUs

Post by Esben » Mon Nov 05, 2012 2:04 pm

Thanks a lot for taking the time to make this comparison between the power supplies. I like very much your methodology with lost watts vs watts output, as it gives a much clearer comparison of the power supplies of different ratings. It is difficult to compare efficiencies of two power supplies, one which is 88.xx % efficient at 10 % load of XX watts, to another which is 89.yy % efficient at 10 % load of YY watts. Hopefully the editors of SPCR will consider using a similar methodology.

Both Kingwin power supplies and the TFX Seasonic SS-350TGM are not available here. They look like good contenders for maybe even surpassing the PicoPSU + DC-brick. How would that combo rate in idle power waste?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Efficiency vs load analyzed for 20 PSUs

Post by CA_Steve » Mon Nov 05, 2012 2:55 pm

Ralf recently compared the Seasonic G-360, Seasonic SS350 and a pico-PSU and brick combo here.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: Efficiency vs load analyzed for 20 PSUs

Post by MikeC » Wed Nov 07, 2012 8:04 am

Esben wrote:Thanks a lot for taking the time to make this comparison between the power supplies. I like very much your methodology with lost watts vs watts output, as it gives a much clearer comparison of the power supplies of different ratings. It is difficult to compare efficiencies of two power supplies, one which is 88.xx % efficient at 10 % load of XX watts, to another which is 89.yy % efficient at 10 % load of YY watts. Hopefully the editors of SPCR will consider using a similar methodology.
If you mean to make comparisons of efficiency, perhaps its worth adding to the reviews, but they are already long and complex, and an occasional "here's a summary" comparison article is probably better for this data. But if you mean test at specific output loads instead of at % of rated load....

Since the very first PSU review back in 2002, I've used wattage load points for testing. At first, they were 90W, 150W, 300W, etc... but over the years, the points have been expanded for finer detail and extended all the way down to 20W. iirc, the current test points are 20W, 40W, 65W, 90W, 150W, 200W, 250W, 300W, 400W.... and what happens after that point depends partly on PSU rating. I've never tested for power as % of rated load.

The amount of power lost as heat was introduced into the PSU tested data table of the Kingwin Lazer Platinum 550W Power Supply review back in May 2011... and it's been standard in all PSU reviews since then. I've also encouraged looking at power lost as heat -- as a way of thinking clearly about efficiency -- in many PSU reviews and forum discussions for years.

Steve did a great job of collating the test data into comparison charts, but that data already existed in the reviews.

drees
Posts: 157
Joined: Thu Jul 13, 2006 10:59 pm

Re: Efficiency vs load analyzed for 20 PSUs

Post by drees » Tue Nov 13, 2012 10:56 am

Wow, these charts are awesome and really help visualize the difference in efficiency between PSUs - in particular the low-power portion which is of particular interest to me. Nicely done!

The SS-350TGM really stands out in comparison to the rest at low power thanks to these charts.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Efficiency vs load analyzed for 20 PSUs

Post by CA_Steve » Tue Nov 13, 2012 7:41 pm

Thanks - glad you liked 'em. More to come in a bit.

Post Reply