## SETI energy usage

Message boards : Number crunching : SETI energy usage
Message board moderation

1 · 2 · 3 · 4 . . . 5 · Next

AuthorMessage
Steven Miller

Joined: 3 Jul 09
Posts: 2
Credit: 5,328
RAC: 0
Message 917095 - Posted: 12 Jul 2009, 7:55:18 UTC

In a recent NPR Science Friday radio program on May 22nd, I heard Dan Werthimer say that in the ten years since SETI@home was started, five million SETI@home volunteers in 226 countries donated over three million years of computing time. That's an amazing accomplishment. I remember running SETI soon after it first became available, watching the numbers crunch and hoping someone would soon see some interesting signal... even a curious false signal.

However, it got me thinking about 3 million years of computing time crunching a CPU intensive algorithm. Is there some way to measure the economic cost of this, some way that we can quantify it? In running the program at home today, I realize that one method would be to estimate the approximate energy use.

I've spent a fair amount of time measuring the energy usage of various devices in my house, including my computers during various states (off, sleep, idle, and during peak activity). Normally my computers (2 PC's and 1 Mac for my family of 4) automatically go to sleep about 15 minutes of idle. Sleep is typically 1 or 2 watts, idle is about 60-100 (w/o monitor), and peak activity is 100-200 depending on whether it's only number crunching (like SETI) or during 3D game playing which utilizes the graphics card (not counting monitor).

Based on my computers, the energy usage and cost of these 3 million years of running SETI@home, assuming that about half the computers would normally be sleep or off if they weren't running SETI@home, is:

years: 3,000,000
hours/year: 8760
total hours: 26,280,000,000 hours
Typical Watts running SETI: 120 watts (w/o monitor)
Watts for idle computers (not in sleep): 100 watts
% of computers that would be normally off/sleep: 50%
watts for off/sleep: 2

Total energy used running SETI: 3,153,600,000 KWH
Energy if not running SETI: 1,340,280,000 KWH
Net energy used due to SETI: 1,813,320,000 KWH

Price per KWH: 8 cents (0.08 dollars)
Energy cost: 146,065,000 dollars
Gigawatt days: 75.56

So, if my calculations are correct, the total cost to run SETI@home to date has been about \$146 million dollars in electricity or roughly the equivalent of a 1000 megawatt power plan running for 2.5 months.

I guess it just demonstrates the adage "There's no such thing as a free lunch"

Regards,

Steven Miller
booboobaby@yahoo.com
ID: 917095 ·
TCP JESUS

Joined: 19 Jan 03
Posts: 205
Credit: 1,248,845
RAC: 0
Message 917105 - Posted: 12 Jul 2009, 9:05:44 UTC

That's a great post.
I am TCP JESUS...The Carpenter Phenom Jesus....and HAMMERING is what I do best!
formerly known as...MC Hammer.
ID: 917105 ·
Joseph

Joined: 9 Mar 01
Posts: 42
Credit: 4,191,922
RAC: 0
Message 917107 - Posted: 12 Jul 2009, 9:09:27 UTC - in response to Message 917095.

I think 146 million is low estimate. I think it is a lot more than 146 million dollars. I for one had 11 computers which were running all time 24x7 for 2 years crunching for SETI among other things. And I am sure there are other like me or even more.
ID: 917107 ·
tullio
Volunteer moderator
Volunteer tester

Joined: 9 Apr 04
Posts: 6156
Credit: 1,457,359
RAC: 786
Message 917133 - Posted: 12 Jul 2009, 12:24:29 UTC - in response to Message 917107.

I think 146 million is low estimate. I think it is a lot more than 146 million dollars. I for one had 11 computers which were running all time 24x7 for 2 years crunching for SETI among other things. And I am sure there are other like me or even more.

Also my electricity costs me 19 eurocents/kWh.And GPUs are more power-hungry than CPUs like mine.
Tullio
ID: 917133 ·
ML1
Volunteer tester

Joined: 25 Nov 01
Posts: 9308
Credit: 6,710,907
RAC: 4,301
Message 917134 - Posted: 12 Jul 2009, 12:44:54 UTC - in response to Message 917095.

So, if my calculations are correct, the total cost to run SETI@home to date has been about \$146 million dollars in electricity...

I guess it just demonstrates the adage "There's no such thing as a free lunch

Good post there.

Note though that a (large?) proportion of that would have been used regardless of whether s@h/Boinc was there or not... Boinc is intended to take advantage of idle CPU time for systems that would be powered up and online regardless.

For example, I have no "dedicated" crunchers. CPU idle time only contributed from here.

Happy crunchin',
Martin

See new freedom: Mageia Linux
See & try out for yourself: Linux Voice
The Future is what We all make IT (GPLv3)
ID: 917134 ·
1mp0£173
Volunteer tester

Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
Message 917178 - Posted: 12 Jul 2009, 16:40:33 UTC

The original idea for distributed computing came from a sudden realization one day that in some university department somewhere there was a stack of computers on desks that were being used for word processing, and between keystrokes they were doing nothing.

SETI@Home classic, and later, BOINC, were designed to harvest those unused cycles.

So, I think we have two different kinds of crunchers:

Most of the people active in the forums have purpose-built crunchers (or at least, machines that were selected with crunching in mind). They're very credit-driven. The vast majority of participants have never posted (9 out of 10) and are not fanatics.

I suspect that most of the non-fanatics are not doing anything extra for SETI.

I have two machines actively crunching.

One is a small server that exists mainly to track web statistics. It isn't very busy -- or very big. It runs 24/7 because it is "on" for that task. It's a Via C7 so it's power efficient (and not the fastest cruncher).

The other is my workstation. I turned it off when I went to sleep around ten last night, and turned it on this morning when I woke up. It's old, it's an Athlon XP 3000+ and will be retired soon -- replaced with a machine chosen for low-power (and low heat).

In both cases, the power draw would be a little lower if I didn't crunch, but not by much.

I think the majority are "casual crunchers" harvesting a waste process, and you can't count that against the SETI@Home power bill.
ID: 917178 ·
HAL9000
Volunteer tester

Joined: 11 Sep 99
Posts: 6403
Credit: 170,874,466
RAC: 54,029
Message 917297 - Posted: 13 Jul 2009, 5:35:39 UTC

Your post got me to do something I have been meaning to check for a while now.
Grabbed my meter and checked the power usage of most all my home machines.

The values for current and va are: system off/system idle/system w/BOINC

HAL9000II-Current: 0.0/1.0/1.3-Volt Amps: 0/120/156
Intel Core 2 Duo CPU E8400 @ 3.00GHz [x86 Family 6 Model 23 Stepping 6](2 processors)
HAL9000-Current: 0.2/1.0/1.6-Volt Amps: 24/120/192
Intel Pentium 4 CPU 3.00GHz [x86 Family 15 Model 2 Stepping 9](2 processors)
SAL9000-Current: 0.0/0.2/0.3-Volt Amps: 0/24/36
Intel Pentium M processor 1.50GHz [x86 Family 6 Model 13 Stepping 8](1 processors)
Intel Pentium 4 CPU 2.40GHz [x86 Family 15 Model 2 Stepping 7](1 processors)

AE35 uses 2 2.5" notebook drives to run. So the numbers would be higher if I was using standard 3.5" drives.

HAL9000 is loaded up with 3 HD's and 2 DVD drives. I'll be converting it to 4 2.5" SATA drives and no DVD drives soon.

HAL9000II has 3 HD's likes HAL9000, but they are the 1TB green western digital ones. While my meter was hooked up I did some 3D gaming tests. With the ATI 4870 in there the peak current draw was only 1.4A. I did that test w/ and w/o BOINC running. This system just doesn't seem to use more then 168va.

Using BoincView her are the wu/day numbers it is generating atm.
HAL9000II-39.3
HAL9000-18.9
AE35-9.25
SAL9000-8.6

So I took the machines power usage for the day and divided by the wu's a day. To get some sort of computing efficiency for each of them.

This data displays va per day with the resulting wu's per va per day.

HAL9000II-3744/0.0104967948
HAL9000-4608/0.0041015625
AE35-2880/0.0032118055
SAL9000-864/0.0099537037

This seems to show that my old notebook processes wu's nearly as efficient as my new desktop for the given power used. Which I find really interesting.

I just did the math for it, and HAL9000 + AE35 would cost \$464 for a year to run them. Together they use exactly twice the power of HAL9000II. So by making an exact twin of HAL9000II I could save \$232 a year!

There is also the cost of added home cooling costs that is very hard to track.

I think this might give me a good excuse to upgrade to a Core 2 Quad and get a cheap board to run my Core 2 Duo on. Which could be done for under \$200 now I bet.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the BP6/VP6 User Group today!
ID: 917297 ·
-= Vyper =-
Volunteer tester

Joined: 5 Sep 99
Posts: 1358
Credit: 489,601,979
RAC: 169,856
Message 917300 - Posted: 13 Jul 2009, 5:51:10 UTC

Cool thread, but the energy price hasn't a fixed price in every country though, in our country it varies from 0,110 about 0,148 dollars per kWH up today..

So the median needs to be increased a bit :)

Kind regards Vyper

P.S All in all you could say that some people like me which has bought hardware to run it dedicated for s@h have atleast donated perhaps 10M\$ for seti at home :) D.S

_________________________________________________________________________
Founder of GPU Users Group
ID: 917300 ·
HAL9000
Volunteer tester

Joined: 11 Sep 99
Posts: 6403
Credit: 170,874,466
RAC: 54,029
Message 917301 - Posted: 13 Jul 2009, 6:10:01 UTC - in response to Message 917300.

Cool thread, but the energy price hasn't a fixed price in every country though, in our country it varies from 0,110 about 0,148 dollars per kWH up today..

So the median needs to be increased a bit :)

Kind regards Vyper

P.S All in all you could say that some people like me which has bought hardware to run it dedicated for s@h have at least donated perhaps 10M\$ for seti at home :) D.S

Mine is \$0.17 per KWh this time of the year. I just take my bill and divide the KWh and cost. That is much easier then looking at generating cost, transmission cost, line maint cost. The cost goes down in the winter as I am 100% electric and something about state laws making them do that.

\$0.08 is good for a base line. The new water heater I had put in uses \$0.086 as it's basis for the energy star power rating. I think that number is a US national average given out by the DOE.

On a side note. I just ordered a mobo and memory to setup a machine for my Core 2 Duo as mentioned on my last post. Only cost me \$183.94! Now it's just a matter of the \$300 for the Q9650. :/
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the BP6/VP6 User Group today!
ID: 917301 ·
Joseph

Joined: 9 Mar 01
Posts: 42
Credit: 4,191,922
RAC: 0
Message 917334 - Posted: 13 Jul 2009, 11:33:05 UTC

so, can we say that keeping our computers busy curnching for SETI cost us on average ,say, 50\$/month per person?
ID: 917334 ·
tullio
Volunteer moderator
Volunteer tester

Joined: 9 Apr 04
Posts: 6156
Credit: 1,457,359
RAC: 786
Message 917336 - Posted: 13 Jul 2009, 11:44:35 UTC

That amounts to my total electricity bill, 30 euros/month, including lights, washing machine,refrigerator and freezer,TV, radio, stereo, No AC and no heating. But I am running 5 BOINC projects, not only SETI, with no GPU, only a CPU running 1.8 GHz 24/7. An occasional lawnmover cut also spends some.
Tullio
ID: 917336 ·
Reuben Gathright

Joined: 8 Mar 01
Posts: 213
Credit: 14,594,579
RAC: 0
Message 917396 - Posted: 13 Jul 2009, 17:55:19 UTC

Nice post HAL9000, I use a P3 International P4400 meter to measure the power draw of my systems. Over a 24 hour period, I found an HP Proliant GL360D5, Dual Xeon server pulls 20Kw!
Overclock with the MSI G31M3-L and Intel E8600 3.33Ghz
Intel D865GLC Socket 478 Motherboard
~How To Overclock The Eee ASUS 1005HA Netbook To 1.9Ghz~
ID: 917396 ·
Dave Sedgwick

Joined: 14 May 99
Posts: 27
Credit: 1,642,558
RAC: 38
Message 917416 - Posted: 13 Jul 2009, 18:33:09 UTC

I'm not sure you can really put a price on finding the discovery that will render the major faiths moot, and finally make people realize that we do really have almost everything in common - thus treating others a whole lot better!
if by chance an honest man like yourself were to make enemies, they would become my enemies. and then they will fear you.
ID: 917416 ·
j2satx
Volunteer tester

Joined: 2 Oct 02
Posts: 404
Credit: 196,758
RAC: 0
Message 917442 - Posted: 13 Jul 2009, 20:34:46 UTC - in response to Message 917178.

The original idea for distributed computing came from a sudden realization one day that in some university department somewhere there was a stack of computers on desks that were being used for word processing, and between keystrokes they were doing nothing.

SETI@Home classic, and later, BOINC, were designed to harvest those unused cycles.

So, I think we have two different kinds of crunchers:

Most of the people active in the forums have purpose-built crunchers (or at least, machines that were selected with crunching in mind). They're very credit-driven. The vast majority of participants have never posted (9 out of 10) and are not fanatics.

I suspect that most of the non-fanatics are not doing anything extra for SETI.

I have two machines actively crunching.

One is a small server that exists mainly to track web statistics. It isn't very busy -- or very big. It runs 24/7 because it is "on" for that task. It's a Via C7 so it's power efficient (and not the fastest cruncher).

The other is my workstation. I turned it off when I went to sleep around ten last night, and turned it on this morning when I woke up. It's old, it's an Athlon XP 3000+ and will be retired soon -- replaced with a machine chosen for low-power (and low heat).

In both cases, the power draw would be a little lower if I didn't crunch, but not by much.

I think the majority are "casual crunchers" harvesting a waste process, and you can't count that against the SETI@Home power bill.

Why not.....BOINC does not use "idle cycles", so the cost difference between an idle cycle and BOINC processing cycle should go to BOINC.
ID: 917442 ·
1mp0£173
Volunteer tester

Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
Message 917444 - Posted: 13 Jul 2009, 20:42:31 UTC - in response to Message 917442.

I think the majority are "casual crunchers" harvesting a waste process, and you can't count that against the SETI@Home power bill.

Why not.....BOINC does not use "idle cycles", so the cost difference between an idle cycle and BOINC processing cycle should go to BOINC.

These machines don't have a significant energy cost difference between idle and running BOINC. It might be a few percent at most.

Compare that to the cost of a dedicated SETI cruncher, or the cost of leaving a machine on overnight when it could have been turned off.

If a machine doesn't have a reason to be on, around here it gets turned off. Unused cycles go to BOINC only if the machine would be on for some other reason.
ID: 917444 ·
HAL9000
Volunteer tester

Joined: 11 Sep 99
Posts: 6403
Credit: 170,874,466
RAC: 54,029
Message 917466 - Posted: 13 Jul 2009, 22:01:22 UTC - in response to Message 917444.

I think the majority are "casual crunchers" harvesting a waste process, and you can't count that against the SETI@Home power bill.

Why not.....BOINC does not use "idle cycles", so the cost difference between an idle cycle and BOINC processing cycle should go to BOINC.

These machines don't have a significant energy cost difference between idle and running BOINC. It might be a few percent at most.

Compare that to the cost of a dedicated SETI cruncher, or the cost of leaving a machine on overnight when it could have been turned off.

If a machine doesn't have a reason to be on, around here it gets turned off. Unused cycles go to BOINC only if the machine would be on for some other reason.

Based on my measurements of my machines. Most of them use on average 60% more power then when idle. I'd like to collect more data across more machines. However, It does seem reasonable that an average machine would use 50% more power under full cpu load.

I bet if you compare the extra power used by processing vs the wasted power of phantom devices in your home I would guess that the phantom devices are sucking up more power.

I just look at the power used for processing is part of the "donation" to the project. I would like it to be as efficient as possible though. To that end I'm cooking up an idea to run my router, switches and cable modem from solar. They are all 12v devices using 10-30watts each. So I'm going to grab some of those 25w solar car charges off of ebay and see how much power I can average from my office windows.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the BP6/VP6 User Group today!
ID: 917466 ·
1mp0£173
Volunteer tester

Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
Message 917474 - Posted: 13 Jul 2009, 22:18:28 UTC - in response to Message 917466.

I bet if you compare the extra power used by processing vs the wasted power of phantom devices in your home I would guess that the phantom devices are sucking up more power.

Not here. All of the phantom devices have been tracked down and either reduced or eliminated.
ID: 917474 ·
Jim_S

Joined: 23 Feb 00
Posts: 4665
Credit: 44,699,850
RAC: 31,030
Message 917494 - Posted: 14 Jul 2009, 1:21:36 UTC

I run 5 dedicated older computers all year long and our electric bills average \$70.00 to \$120.00 a month.
I consider this a small price to pay to have friends around the World. Also I have learned much about cultural differences.

I Desire Peace and Justice, Jim Scott (Mod-Ret.)
ID: 917494 ·
Virtual Boss*
Volunteer tester

Joined: 4 May 08
Posts: 417
Credit: 6,415,441
RAC: 367
Message 917505 - Posted: 14 Jul 2009, 2:36:53 UTC

My old 500MHz P3 uses 39W idle, 42W crunching. Not much difference there.
ID: 917505 ·
HAL9000
Volunteer tester

Joined: 11 Sep 99
Posts: 6403
Credit: 170,874,466
RAC: 54,029
Message 917523 - Posted: 14 Jul 2009, 5:09:03 UTC - in response to Message 917505.

My old 500MHz P3 uses 39W idle, 42W crunching. Not much difference there.

Back in the way back times I was using 40 or so celeron 400mhz machines. Even running 2 motherboards off of 1 power supply with special Y cables I had made. I don't recall if I ever measured the power fomr them but they were all running off of a 15A breaker. Good ole P3's were good on power. :)
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the BP6/VP6 User Group today!
ID: 917523 ·
1 · 2 · 3 · 4 . . . 5 · Next

Message boards : Number crunching : SETI energy usage