Do you use or do Distributed Computing? Why or Why not?

[narzy]

I really am interested to see someone's kill-a-watt numbers if they have them. It would be interesting to see just how much of an impact DC has on power usage.

 

[Idontcare]

Originally posted by: narzy
I really am interested to see someone's kill-a-watt numbers if they have them. It would be interesting to see just how much of an impact DC has on power usage.

It will be just about the same as loading up any other application that pushes the cpu (or gpu) utilization rates to 100%.

For the vast majority of cpus out there right now the power-consumption delta between idle and fully loaded is less than 150W for even the more extreme TDP chips.

So for the folks who still use incandescent light bulbs, its about the same as turning on an extra 100W light bulb in your house. For the folks who use CFL's is about the same as turning on 4-5 100W equivalent (lumen-wise) CFL bulbs. Watching TV on a standard 27" CRT is going to use about 2-3x more electricity versus setting your otherwise idle computer to run DC.

 

[taltamir]

Originally posted by: Idontcare

Originally posted by: narzy
I really am interested to see someone's kill-a-watt numbers if they have them. It would be interesting to see just how much of an impact DC has on power usage.

It will be just about the same as loading up any other application that pushes the cpu (or gpu) utilization rates to 100%.

For the vast majority of cpus out there right now the power-consumption delta between idle and fully loaded is less than 150W for even the more extreme TDP chips.

So for the folks who still use incandescent light bulbs, its about the same as turning on an extra 100W light bulb in your house. For the folks who use CFL's is about the same as turning on 4-5 100W equivalent (lumen-wise) CFL bulbs. Watching TV on a standard 27" CRT is going to use about 2-3x more electricity versus setting your otherwise idle computer to run DC.

people still use incandescent bulbs?

 

[Griswold]

I used to. For years, on and off. First I stopped doing it during summer because of the extra heat generated that became a problem with todays 100-140W TDP performance behemoths. Then I stopped altogether because most of the projects around arent interesting enough to deal with all the drawbacks (heat and noise, energy consumption).

Though, there is one project I'll spend some CPU cycles on once it finally gets off the ground and that is Orbit@Home. Its sort of out of the ordinary and who knows, it might have a huge impact (no pun) on mankind one distant day when they/we discover a large enough lump of metal to cause severe damage coming our way.

 

[myocardia]

I used to do DC for a fairly short while, but stopped after I realized that wasn't the reason I owned computers.

 

[Philippart]

according to IBM DC has the same effect on your energy bill than an additional light bulb

when talking about cancer projects I'm talking about folding@home and worldcommunitygrid

 

[taltamir]

according to IBM DC has the same effect on your energy bill than an additional light bulb

IBM is full of it.

 

[totalcommand]

Originally posted by: taltamir

Concerns over electricity costs are valid, but irrelevant to the vast majority of folks who are appreciably above the poverty line and don't live with their parents still (meaning they are beholden to someone else who pays their bills for them and so have to justify their cost footprint on the household budget).

Hundreds of dollars per year is not irrelevant no matter how rich you are. It might be acceptable, but not irrelevant...
Also, people keep on saying "cancer research". There are MANY DC projects out there, many of which don't have such well defined benefits (ex: seti@home). I assure you that folding was the LAST project I quit. If you are gonna argue for cancer cures, then state clearly that you are talking about protein folding, not DC in general.

Although I also happen to be a poor college student... (although I never buy beer, cigs, or pizza)

I Would like to point out that underclocking and undervolting a modern CPU to get the WU of an old one is... well, a waste in the relative scheme of things. DC is not a static thing, and as costs decrease, challenges tackled increase, requiring more power. Every year n, people contribute as much FLOPS as the entirety of the calculations done from inception to n-X. with X probably being 3 or 4 years. So the entirety of donations simply accelerates research by a few more years. And to be honest, it would be better to have specialized computational farms and donate 15$ a month to them, instead of running it on your own PC. Part of it is that people are not seeing the cost.

Electricy costs vary as well, (and cooling costs), and it can come up to much more than 15$ a month.

A few years of research is not an insignificant thing at all. The costs of operating a lab are very high.

Plus, a 15$ donation would not be the same as a 15$ electrical WU donation. Infrastructure costs/overhead for running server farms is high as well. Of your 15$ cash donation, probably half that would actually go to computational power.

 

[Idontcare]

Originally posted by: taltamir

according to IBM DC has the same effect on your energy bill than an additional light bulb

IBM is full of it.

I posted links above to the idle and load power consumption data for modern cpus...not sure exactly what basis you are making your argument with here but the data supports IBM's claims.

Originally posted by: totalcommand
Plus, a 15$ donation would not be the same as a 15$ electrical WU donation. Infrastructure costs/overhead for running server farms is high as well. Of your 15$ cash donation, probably half that would actually go to computational power.

And that is the crux of the argument for the existence of DC - the incremental cost donated by the individual contributor can be as little as the added cost of running their rig loaded versus idle.

All the other investments involved are already covered by the person buying the computer (and location/room for its housing) presumably for other purposes entirely (gaming, email, etc).

The research teams that use the DC results can never have a cost structure per WU this low.

 

[elconejito]

I'd say I don't DC simply because i don't know enough about them. I used to SETI many, many years ago. I don't recall why I stopped, I think after a reinstall i just never put it back on. Also, I don't normally leave my main PC on 24/7. I only have it on while I'm using it, except for the occasional overnight video encode in which case, the CPU cyles are already being used.

I wouldn't do it where I live now in the summer simply because it's an old house and all the AC goes right out all the drafty holes in the wall, so adding another source of (significant) heat would just be a nono. And the power bill is really really high already. Maybe in the winter I might do it, we'll see. I do have a couple computers on 24/7, but I've specifically made them to be as low-power as possible (yes, each one is about a light-bulb worth of usage) so then cranking them up to 100% would be counter-productive for me.

 

[Philippart]

you don't have to leave your computers on 24/7 to contribute to DC...

 

[ShawnD1]

Originally posted by: narzy
Why do most people run their computers 24/7? why not set up a sleep/hibernate time?

Because every computer I've seen doing that runs like total garbage. I hear that the on/off power surging isn't a big deal, but there seems to be some correlation between people who turn the computer off every day and having the hard drives screw up. I've been using the same 120gb 5400rpm maxtor drive for 6 years of 24/7 operation and there are no signs of it degrading. I've only ever lost 1 hard drive and it was because I dropped it on a tile floor. The hard drive failures that plague the people around me don't seem to happen on any of my 24/7 computers.

I run F@H because the computer is on anyway. If the computer is already doing file sharing and defragging and virus scanning, it might as well do some calculations on top of that. My UPS says the power difference between full load and no load on an E6600 is only about 10-20W while the whole system takes over 200W, so it's really a non issue. The case fans and hard drives take more power than the CPU does.

edit
I took a few pictures to share with the rest of the class.
Full load power: 324W
No load power: 302W
power used by F@H on both cores: 22W

The power where I live is about 11 cents per kwh. That works out to be $22 per year to run F@H all day and night.

 

[waffleironhead]

Originally posted by: taltamir


people still use incandescent bulbs?

I do. In my experience with many brands of florescent bulbs they flicker and cause headaches and eye strain. So the energy savings not worth the benefit.

 

[PlasmaBomb]

Originally posted by: narzy
I really am interested to see someone's kill-a-watt numbers if they have them. It would be interesting to see just how much of an impact DC has on power usage.

Full load (2 or 4 SMP units on two Clovertown Xeon's @ 2.5 GHz) draws 238W according to my Kill-a-Watt. I don't seem to have written down the idle power

 

[narzy]

Originally posted by: ShawnD1

Originally posted by: narzy
Why do most people run their computers 24/7? why not set up a sleep/hibernate time?

Because every computer I've seen doing that runs like total garbage. I hear that the on/off power surging isn't a big deal, but there seems to be some correlation between people who turn the computer off every day and having the hard drives screw up. I've been using the same 120gb 5400rpm maxtor drive for 6 years of 24/7 operation and there are no signs of it degrading. I've only ever lost 1 hard drive and it was because I dropped it on a tile floor. The hard drive failures that plague the people around me don't seem to happen on any of my 24/7 computers.

I run F@H because the computer is on anyway. If the computer is already doing file sharing and defragging and virus scanning, it might as well do some calculations on top of that. My UPS says the power difference between full load and no load on an E6600 is only about 10-20W while the whole system takes over 200W, so it's really a non issue. The case fans and hard drives take more power than the CPU does.

edit
I took a few pictures to share with the rest of the class.
Full load power: 324W
No load power: 302W
power used by F@H on both cores: 22W

The power where I live is about 11 cents per kwh. That works out to be $22 per year to run F@H all day and night.

Thanks so much for the pictures, I'll be very interested to see how others are also affected by the power usage. This thread has gotten me on to somewhat of a curious adventure .
I really do appreciate all of the feedback thus far to my question. It has gone much more in depth than I had imagined. All very valid points of view and all very much appreciated.

 

[taltamir]

Originally posted by: waffleironhead

Originally posted by: taltamir


people still use incandescent bulbs?

I do. In my experience with many brands of florescent bulbs they flicker and cause headaches and eye strain. So the energy savings not worth the benefit.

flourescent? I hope you mean CFL... the CFL bulbs are very different than old style column flourescent bulbs.

anyways, the figures of 22 watt difference seem bogus to me... ESPECIALLY due to 302W idle, it should be closer to to 100 watt idle and 350 watt load.
I just finished my finals (finally), so I have time to take some measurements... tommorow, right now i gotta sleep, i am dead on my feet

 

[waffleironhead]

Originally posted by: taltamir

Originally posted by: waffleironhead

Originally posted by: taltamir


people still use incandescent bulbs?

I do. In my experience with many brands of florescent bulbs they flicker and cause headaches and eye strain. So the energy savings not worth the benefit.

flourescent? I hope you mean CFL... the CFL bulbs are very different than old style column flourescent bulbs.

anyways, the figures of 22 watt difference seem bogus to me... ESPECIALLY due to 302W idle, it should be closer to to 100 watt idle and 350 watt load.
I just finished my finals (finally), so I have time to take some measurements... tommorow, right now i gotta sleep, i am dead on my feet

Yep CFL=flourescent. They all flicker. Not everyone sees it, but for those of us who do, it can be very bad. I'm just hoping LED becomes more mainstream and cost/lumen drops.

 

[taltamir]

i honestly see more flickering on incandescant bulbs... every time a major power drain adds up like a drier or a laser printer the lights all flicker. With my CFL I don't really notice those problems...

My biggest problem with CFL actually is the mercury.. did you know that if it breaks you are supposed to RUN OUT OF THE ROOM and only come back to clean the debris 20 minutes later?
http://www.npr.org/templates/s...ry.php?storyId=7431198

funny that every "environmentalist" from here to kingdom come tells you to REPLACE exist bulbs with CFL (another stupid thing, the production cost of the bulb is WAY greater than the saved amount... you first let your current bulbs naturally burn out... THEN you replace them with CFL... otherwise you create MORE pollution and spend more money)

 

[taltamir]

Off: 2 watt
Idle: 92 watt (stable)
CPU OCCT: 164 watt - stable
CPU Linpack on OCCT: 186 watt (exact, stable)
GPU OCCT: 248 watt - stable
GPU + Linpack (PSU test): 280 to 300 watt fluctuating.
Second idle measure after tests: 118 watt (while showing 0% utilization), waited 5 minutes, no change.

Disabled all power saving: gigabyte DES (EIST) and intel C1E, and intel TM2.

idle: 105 to 120 watt fluctuating
CPU OCCT: 170 watt - stable
CPU Linpack on OCCT: 186 watt (exact, stable)
GPU OCCT: 251 watt - stable
GPU + Linpack (PSU test): 280 to 305 watt fluctuating.
Second idle measure after tests: 124 watt (while showing 0% utilization)
CPU OCCT second time: 172 watt
third idle: 120 (very stable) - waited five minutes, no change in value.

OCCPT v3.1.0 used in test on 2009 august 4th. Kill-A-Watt meter used at wall outlet of PC BOX... monitor, speakers, router, etc not included.
System: Q9400 CPU running stock, 2x2GB of DDR2-800 ram, gigabyte EP35-DS3R mobo, intel X25-M 80GB G2 + WD 640GB HDD.

i suspect the higher idle wattage after the tests is due to a bug in the mobo. probably to do with its system of controlling the amount of active CPU phases.

 

[taltamir]

also... i gotta ask. How many of you use the GPU folding client?

 

[Philippart]

most people who are into DC now run a gpu client too! there are various projects that support it

 

[Denithor]

taltamir - GPU clients are all that I run on my boxes. I have EVGA cards across the board with a lifetime warranty covering each. I figure why stress the rest of the system when the results are so much lower?

I have these folding:
8800GS
8800GTS
GTX 260 x 2

Yields about 20k points per day on average.

I need to get one of those Kill-A-Watt devices and measure my power consumption while folding.

Can you actually fire up a copy of the F@H GPU client on your GTX 260 long enough to get a reading? I'm curious how much difference it will make versus idle.

 

[elconejito]

I've got a kill-a-watt and some time to burn. I'd install a DC client for testing purposes to see what idle and load are while using the client. Which one should i choose?

 

[IEC]

I used to run CPU F@H, but that was when I didn't pay for electricity... now I run the GPU2 client on a GTX260 and a 9600GSO for 10K PPD (not for the TeAm, sorry).

My electricity bill was $45 last month. It definitely adds to the bill, but I don't mind it.

 

[Idontcare]

Originally posted by: taltamir
My biggest problem with CFL actually is the mercury.. did you know that if it breaks you are supposed to RUN OUT OF THE ROOM and only come back to clean the debris 20 minutes later?
http://www.npr.org/templates/s...ry.php?storyId=7431198

I with you here, as a chemist I really don't like the idea of my children being exposed to mercury that is in a form which can readily be taken in thru the lung tissues.

I've broken a CFL before, by accident, it really doesn't take too much carelessness and that spiral can break.

I just bought a couple of those LED replacements for canister lighting, that'll be my first foray into replacing CFL's. They were as Sam's club.

Originally posted by: taltamir
funny that every "environmentalist" from here to kingdom come tells you to REPLACE exist bulbs with CFL (another stupid thing, the production cost of the bulb is WAY greater than the saved amount... you first let your current bulbs naturally burn out... THEN you replace them with CFL... otherwise you create MORE pollution and spend more money)

Its true, CFL's aren't very green once the total carbon footprint, including production costs, are factored in. The same with solar cell electricity. What is even more criminal in my mind about the CFL's is that the mercury is being treated like an acceptable landfill material when you dispose of the CFL's in your garbage.

Really this mercury ought to be recaptured and disposed of like any other toxic waste, a process that would elevate the carbon footprint of the life cycle of the CFL even higher still.

The environmental propaganda that comes with CFL's is a total sham, but that doesn't mean the consumer doesn't stand a chance to actually save some money in the process. Just don't delude yourself into thinking you are saving money and the environment at the same time.

(data point: converting my entire house from incandescent to CFL's resulted in a $100/month electricity bill reduction...it cost me ~$700 to buy all those CFL's but I recovered the investment in less than 8 months)

Originally posted by: elconejito
I've got a kill-a-watt and some time to burn. I'd install a DC client for testing purposes to see what idle and load are while using the client. Which one should i choose?

Same here, (1) have kill-a-watt, and (2) am willing to generate data. Need input on which DC to download and try out.