[mythtv-users] Efficient Power Supply Units/Saving money

[Michael T. Dean]

So, last year a PSU of mine blew up and newegg had a sale on an Antec 
Earthwatt EA430 PSU.  I decided to try it out because I had heard a lot 
of good things about the 80 PLUS ( http://www.80plus.org/ ) program and 
the EA430 was 80 PLUS certified.  I was very happy with the PSU, 
but--because a friend had borrowed my Kill-A-Watt--testing the power 
usage of the system with an old/cheap/non-80-PLUS PSU and with the new 
PSU was a lot more work than I made time to do.  (Though I promised to 
do the test a couple of times.)

However, I now have 3 other PSU's that are starting to fail, so I 
watched for another sale and ordered 3 more (the new ones are actually 
the newer EA430D).  This time, I actually measured the system--my master 
backend--power consumption, and I'm now sold on 80 PLUS PSU's.

Some system configuration information:
   Mother board: PC Chips M848A v3.0 (SiS746FX NB & SiS963L SB)
      Processor: Athlon XP 2400+ (Seems to be about 65W or 70W TDP)
            RAM: 1GB
          HDD's: 2x1.5TB HDD, 1x750GB HDD
          Video: 1xNVIDIA GeForce FX 5200 (discrete)
          Audio: Realtek ALC655 (integrated)
  Capture Cards: 2xpcHDTV HD-3000's
          Other: 1xRosewill SATA/IDE card

The system is a dedicated backend.  It does not run X, it does not have 
an attached keyboard/mouse/monitor.  The processor does not support 
frequency scaling.  I don't spin down the HDD's.  Note that load was not 
identical for the tests (as I had different numbers of recordings 
occurring/different amounts of commflagging), but by measuring for "long 
enough" durations, I should have a reasonably accurate portrayal of 
average usage.

  Old PSU (w/ BOINC/SETI at home):
    20090130 13:30 - 20090131 15:55: 26h25m/4.40kWh, 3.997kWh/d = 
119.924kWh/mo
                   - 20090201 14:10: 48h40m/8.15kWh, 4.019kWh/d = 
120.575kWh/mo
                   - 20090202 13:30: 72h00m/12.06kWh, 4.02kWh/d = 
120.600kWh/mo
    20090206 02:40 - 20090208 08:25: 53h45m/9.03kWh, 4.030kWh/d = 
120.960kWh/mo
                     20090210 16:45: 110h5m/18.62kWh, 4.06kWh/d = 
121.784kWh/mo

Average draw: 169.14W

  Old PSU (w/o BOINC/SETI at home):
    20090202 13:45 - 20090204 18:00: 52h15m/7.89kWh, 3.624kWh/d = 
108.723kWh/mo
                     20090206 02:40: 84h55m/12.87kWh, 3.64kWh/d = 
109.123kWh/mo

Average draw: 151.56W

  New PSU (w/o BOINC/SETI at home):
    20090210 17:15 - 20090211 14:53: 21h38m/2.82kWh, 3.129kWh/d = 
93.855kWh/mo
                     20090213 12:53: 67h38m/8.81kWh, 3.126kWh/d = 
93.788kWh/mo

Average draw: 130.26W

  New PSU (w/ BOINC/SETI at home):
    20090213 13:30 - 20090215 15:53: 50h23m/7.47kWh, 3.558kWh/d = 
106.750kWh/mo

Average draw: 148.26W

So, looking at the data, after simply changing the PSU, the average 
power consumption dropped from about 169W to 148W (just over 20W 
savings) while running BOINC/SETI at home and from about 151W to 130W (just 
over 20W savings) when not running BOINC/SETI at home.  When the system 
runs 24/7, that equates to about 500Wh/day or, 15kWh/mo.  Assuming 
$0.10/kWh, that's a savings of $1.50/mo.

The data also shows me that running SETI at home is costing me about 18W.  
Therefore, I could choose to stop SETI and save some additional money, 
or I could choose to interpret the data as saying that swapping PSU's 
allows me to "run SETI for free" (i.e. the amount of power SETI takes is 
about equivalent to the amount the PSU saves).

If anyone else is looking at getting a new PSU, newegg has a nice deal 
on the Corsair CMPSU-400CX 400W ( 
http://www.newegg.com/Product/Product.aspx?Item=N82E16817139008 ), but 
note that it has a $20 mail-in rebate.  They also have a handy search 
criterion under system Power Supplies that allows you to find only 80 
PLUS (including higher level) PSU's.

Make sure that you verify your current PSU is less efficient than the 
new one if you decide to "upgrade" for the cost savings alone.  Note 
that you can find details (including test results from certification 
showing actual measured efficiency) at 
http://www.80plus.org/manu/psu/psu_join.aspx .

And, the /best/ way to save power/money is to shut down your systems 
when they're not in use.  You can do this for myth systems with 
mythshutdown (and mythwelcome on frontend systems).  Chris Pinkham's 
upcoming Wake-on-LAN patches, which allow the master backend to wake 
slave backends when they're required (which will likely be available in 
0.22), will allow for the greatest power savings with the least effort.

I also highly recommend a Kill-A-Watt or some other "at the wall" usage 
meter to find out exactly what your systems are costing you.  Again, 
assuming 24/7 usage and $0.10/kWh, my master backend is /now/ (with the 
more efficient PSU) costing about $10.68/mo (with SETI) or $9.38/mo 
(without).

Mike


*** The PSU's were:
Old PSU:
RaidMax RX-630A ATX12V V2.01 Switching Power Supply
http://www.raidmax.com/specs/volcano.htm
Efficiency: "Up to 80%"
+3.3V @ 0.5A /30.0A
  +5V @ 2A   / 38A
+12V1 @ 1A   / 20A
+12V2 @ 1A   / 18A
 -12V @ 0A   / 0.8A
+5VSB @ 0A   / 2.5A

New PSU:
Antec Earthwatts EA430D 430W ATX12V v2.2 80 PLUS Certified Active PFC PSU
http://www.newegg.com/Product/Product.aspx?Item=N82E16817371006
Efficiency: "> 80%"
+3.3V @ 0.5A / 20A
  +5V @ 0.5A / 20A
+12V1 @ 1A   / 17A
+12V2 @ 1A   / 16A
 -12V @ 0A   / 0.8A

Yeah, I know the values differ from newegg's, but these are 
correct--they're straight off the labels.

And, yeah, I realize that the RX-630A was way, way overpowered (whereas 
the EA-430D is just way overpowered) for the system, so the RX-630A 
running at low load (which is where non-80-PLUS PSU's tend to be very 
inefficient).  Note, though, that the 80-PLUS PSU's must be >80% 
efficient from 20% to 100% load (and the BRONZE/SILVER/GOLD have higher 
requirements).