CPU power usage

Message boards : BOINC client : CPU power usage
Message board moderation

To post messages, you must log in.

AuthorMessage
stuartd

Send message
Joined: 5 Jun 06
Posts: 2
United States
Message 4652 - Posted: 5 Jun 2006, 15:38:48 UTC

I just got a watt meter ("Kill-a-Watt" model P3, about $20 online), and I was wondering how much power my computer actually used. I discovered that running BOINC (as opposed to just idle) increases the PC power consumption by ~50 Watts. This is a 2.8GHz "hyper threaded" Pentium, Windows XP, lots of RAM.

This amounts to ~10 cents a day (assuming about 20 hours a day of run time -- I turn the darned thing off at night), or $36.50 a year, at PG&E's consumer rate of about 10 cents a KWh.

So, donating "unused" computer power is actually donating real electrical power -- at least on this computer. Has anyone else done measurements? What sort of values are you all seeing?

Happy trails,
Stuart
ID: 4652 · Report as offensive
Aurora Borealis
Avatar

Send message
Joined: 8 Jan 06
Posts: 448
Canada
Message 4653 - Posted: 5 Jun 2006, 16:43:58 UTC

Energy considerations. There was a thread on the Seti number cruncher board in the middle of last month discussing this but I don't have the time at the moment to locate it.

Boinc V 7.4.36
Win7 i5 3.33G 4GB NVidia 470
ID: 4653 · Report as offensive
stuartd

Send message
Joined: 5 Jun 06
Posts: 2
United States
Message 4658 - Posted: 5 Jun 2006, 21:34:58 UTC - in response to Message 4653.  

Thanks! I see my measurements are right in line with the "Energy considerations" page -- I hadn't seen that before. Do other processors have similar costs? (Well, I guess with everyone going to Intel, maybe it's moot.)

Other interesting numbers -- I find my fridge costs about 45 cents a day, and when my computer is powered down but everything else still "on" (monitor sleeping, powered speakers still powered but quiet, cell phone and PDA chargers plugged in but not charging, etc.) the plug strip is pulling about 7 watts -- not bad.

Energy considerations. There was a thread on the Seti number cruncher board in the middle of last month discussing this but I don't have the time at the moment to locate it.


ID: 4658 · Report as offensive
ylp88

Send message
Joined: 7 Jun 06
Posts: 1
Australia
Message 4664 - Posted: 7 Jun 2006, 5:16:45 UTC

I did some measurements using a DIY built power meter and compared it to readings made using my multimeter and a borrowed oscilloscope.

My results and similar disussion is here.

ylp88
ID: 4664 · Report as offensive
Augustine
Avatar

Send message
Joined: 10 Mar 06
Posts: 73
Message 4673 - Posted: 8 Jun 2006, 7:01:44 UTC
Last modified: 8 Jun 2006, 7:03:40 UTC

FWIW, I measured the additional power used by a BOINC project using the UPS monitor on an old system a while ago too.

Using the same UPS on a new, dual-core Opteron 175 Sun x2100 server that replaced that system, two instances of the Malaria Control application increases the total system power consumption by 25W from 82W.

ID: 4673 · Report as offensive
Clarence

Send message
Joined: 21 May 06
Posts: 3
United States
Message 5048 - Posted: 18 Jul 2006, 0:52:15 UTC - in response to Message 4652.  

I just got a watt meter ("Kill-a-Watt" model P3, about $20 online), and I was wondering how much power my computer actually used. I discovered that running BOINC (as opposed to just idle) increases the PC power consumption by ~50 Watts. This is a 2.8GHz "hyper threaded" Pentium, Windows XP, lots of RAM.

This amounts to ~10 cents a day (assuming about 20 hours a day of run time -- I turn the darned thing off at night), or $36.50 a year, at PG&E's consumer rate of about 10 cents a KWh.

So, donating "unused" computer power is actually donating real electrical power -- at least on this computer. Has anyone else done measurements? What sort of values are you all seeing?

Happy trails,
Stuart


Interesting discussion on power use. Am in process of upgrading my home network of 4 fulltime older pc's, all running DamnSmallLinux and BOINCing full time. Would be interested in seeing comparison of watt per gigaflop or other ratio of effective crunching vs. actual power use by cpu, motherboard, and or pc. Has anyone done or seen something like that?

ID: 5048 · Report as offensive
mnb-fin

Send message
Joined: 25 May 06
Posts: 6
Finland
Message 5069 - Posted: 20 Jul 2006, 11:06:26 UTC

My PC:
None of the components overclocked.
Processor is AMD Athlon XP 2100+ Palomino 1.73 GHz (thermal design point 72 watts)

110W Idle
187W Running BOINC at 100% CPU utilization

(measured average of time period of 16 hours.)


ID: 5069 · Report as offensive
Clarence

Send message
Joined: 21 May 06
Posts: 3
United States
Message 15066 - Posted: 21 Jan 2008, 17:36:54 UTC
Last modified: 21 Jan 2008, 17:48:22 UTC

PC______OS______W/BOINC_Float___Integer__F/w____I/W____ RankF/W__Rank I/W
GX1____ 32_____ 50_____ 313_____444_____6.26____8.88____6_______6
L4100___62_____ 62_____ 672_____1430____10.84___23.06___4_______3
600_____47______47______710_____1031____15.11___21.94___1_______4
466 ____40______85______656_____945_____7.72____11.12___5_______5
gxa ____57______64______163_____232_____2.55____3.63____7_______7
duron __81_____ 81______1223____3161____15.10___39.02___2_______1
sempron_110____ 110_____1508____3906____13.71 35.51___3_______2

Using the Kill a Watt, I obtained the information above. There are conflicting ranks depending on whether floating point or integer. Which matter more for BOINC? All OS's are Linux. GX1, 600, 466, gxa all on DamnSmallLinux toram; all but Gx1 and GXa CPU upgraded; duron, sempron, 4100 either ubuntu hardy alpha 3, or gOS beta 2.

Clarence
ID: 15066 · Report as offensive
MikeMarsUK

Send message
Joined: 16 Apr 06
Posts: 386
United Kingdom
Message 15068 - Posted: 21 Jan 2008, 18:35:45 UTC


Hi Clarence,

The best thing to do is to ignore the (fairly meaningless and unreliable) Boinc benchmarks and look at the performance of each system on a particular project. Each project will be different.

ID: 15068 · Report as offensive
Profile Ananas

Send message
Joined: 27 Jun 06
Posts: 305
Germany
Message 15239 - Posted: 5 Feb 2008, 23:08:10 UTC

Electricity costs in germany currently are somewhere between 25 and 30 cts/kWh (depending on provider and location) but some providers allow the installation of a dual counter with cheaper electricity for 8 hours at night, reducing the costs by about 5 cts/kWh.

For people with a big farm the installation of such a dual counter might pay off.
ID: 15239 · Report as offensive
rroonnaalldd

Send message
Joined: 7 Jan 08
Posts: 31
Germany
Message 15244 - Posted: 6 Feb 2008, 9:29:14 UTC - in response to Message 5069.  
Last modified: 6 Feb 2008, 9:37:20 UTC

My PC:
None of the components overclocked.
Processor is AMD Athlon XP 2100+ Palomino 1.73 GHz (thermal design point 72 watts)

110W Idle
187W Running BOINC at 100% CPU utilization

(measured average of time period of 16 hours.)



That's pretty high.
My Dell E520 with E6600, 4GB of DDR2-667, 320GB WD, Gf7300LE needs:
idle: 92W
full: 116W and QMC-WU's on both Cores.

edit
Messured with APC PowerChute Monitor.
ID: 15244 · Report as offensive

Message boards : BOINC client : CPU power usage

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.