Crunching energy efficiency: PCs vs Smartphones

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 419,532,512
RAC: 39,168

I gave up. And installed

I gave up. And installed BOINC on my 3 years old smartphone. Suddenly I found that it has a 600 MHz processor (ARMv6) and 400 Mb of free memory and is capable to run BOINC and even Einstein on it, but only in internal memory. When installing to the SD it tells me that it have not enough permission for something to run. So I moved it to internal memory and it lies right now in front of me, charging and crucnching already 10% of WU for about 11 hours. Ok. I'm glad.
I remembered that I have a wattmeter too and tried to measure the power draw.
While it is charging from a USB 3.0 port on my home computer, the computer 17" LCD monitor and access point consumes 280 Watts running BOINC both on CPU and GPU. The battery is charged already to 80%. So, I disconnected the phone and found that the power consumption hasn't changed at all (???)
Then I plugged the USB cable from smartphone to the standalone charger that is capable to produce 1.0A on a single 5V USB port (original charger for Palm Tungsten E2 - for those who remember those days). The power meter shows me that smartphone consumes 15 Watts (!!!) while it is charging and running BOINC and... 17 Watts while charhing without BOINC and whith screen turned off.
May be it's a charger? Noway!
I plugged in another (very chin..., ok, noname) charger. It consumes almost 22 Watts when charhing and more than a Watt while idle. Moreover - it very slightly whistles! Well, I have to forget to sleep and charge with it :)
Ok. Than I found another one hightech charger. It is a small Samsung travel adapter model ETA0U10EBE made in China too and capable to produce 0.7 Amps at micro-USB connector. Right now it consumes almost 15 Watts too as a original Palm charger.

All the above shows us that:
1) BOINC itself doesn't affect on a power draw.
2) LCD screen doesn't affect on a power draw. My screen uses only 3% of a battery charge according to an internal power measurment of the smartphone.
3) Cellular module is 10x more powerful than all other things (at least in my smart). It allows me to crunch even while running on battery.
4) Quality charger can save up to 30%-50% of power.
5) We can optimize our job by using standard USB port of home PC for charging thus opting out of standalone charger lowering overall power draw by 15-22 Watts per hour. It happenes I guess from the redundancy of a computer power supply.
Unfortunately I cannot afford this to myself because I sleep in another room and use smartphone as an alarm clock, so I don't want to run in the morning to another room to stop it ringing :)

P.S. My position is still the same. Android device is more efficient with BOINC ONLY when it is bought for some another reason and suddenly appeared useful with BOINC (or you found a way to exchange the old computer with the device without additional charge). Any other way even old PC is more efficient because you don't have to pay some hundred bucks for the device that is already yours, while Android bought especially for crunching have to work for more than 10 years to overcome with its electricity economy it's own cost.

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

Wow, that must be a really

Wow, that must be a really shitty charger you're using.

Although there's always losses associated with transforming 110V/220V AC into low Voltage DC power, 22W is pretty ridiculous.

My unused smartphone charger doesn't even get warm when the battery is full and it's really tiny.

There have been calculations by a german IT magazine though, indicating that running several smartphones from a single, efficient USB multi-charger can save some money running BOINC. The more inefficient your individual chargers are, the bigger the savings.
They also recommended testing if the device can operate without the battery installed (some do, many don't), as it's not needed for older devices used exclusively for BOINC, merely serves as a UPS power backup then.

A computer's USB port (or comparable) should work as well, although their power supplying abilities are always limited - actually charging a BOINC-busy phone with less than full battery will take much longer. But for crunching only this is hardly relevant.

Note though that most typical power meters run into serious (!) inaccuracy when going into single digit Watt regions of measurements, anything of 5W or less will be a hughe challenge for almost all consumer-type measuring devices. So take whatever reading you get with a large grain of salt, the reading just might be way off.
You need specialized equipment to take meaningful readings in such low-Watt regions.

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 419,533,012
RAC: 39,217

RE: A computer's USB port

Quote:
A computer's USB port (or comparable) should work as well, although their power supplying abilities are always limited - actually charging a BOINC-busy phone with less than full battery will take much longer.


Only if it is a USB 2.0 or USB 1.1 port. USB 3.0 produces almost 900 mAmps, more than most of chargers these days.

Jeroen
Jeroen
Joined: 25 Nov 05
Posts: 379
Credit: 740,030,628
RAC: 3

I spent some time today

I spent some time today seeing how low I could get the power consumption with my Intel 3930K. I have the CPU at stock 3.8 GHz frequency and stable CPU load core voltage at 1.167v measured with a volt meter.

Using the same formulas as used earlier in this thread, here is what I have so far.

(86,400 / 2931 Avg) * 6 * 62.5 = 11054 credits / day
11054 / 170 watts = 65 Total daily credits / watt
11054 / 170 watts / 6 cores = ~ 11 daily credits / watt / core

The hardware is about as barebones as I can get. Configuration includes headless, three fans, water pump with H100, SSD, memory, and board. I might be able to lower a few voltages in BIOS while not compromising stability to save a few watts. I am looking forward to seeing what the new Ivy Bridge E processors can do for power consumption reduction.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4,276
Credit: 245,572,442
RAC: 11,047

We did some investigation

We did some investigation ourselves recently (a few weeks ago), measuring power consumption of computers and chargers powering Android devices that were running BOINC. The basic outcome was that the efficiency (flops per watts) of modern computers and mobile devices is pretty much comparable. If you care about energy efficiency, there is nothing that can beat the efficiency of GPUs.

BM

BM

mountkidd
mountkidd
Joined: 14 Jun 12
Posts: 175
Credit: 11,157,814,265
RAC: 4,158,191

RE: If you care about

Quote:

If you care about energy efficiency, there is nothing that can beat the efficiency of GPUs.

BM


To put some numbers on Bernd's quote... My GTX 670's generate 280 credits/watt and the HD 7970 generates 500 credits/watt.

Gord

Manuel7d9
Manuel7d9
Joined: 7 Aug 13
Posts: 10
Credit: 8,228
RAC: 0

@Mike: Yes, I RTFM! :-) It's

@Mike: Yes, I RTFM! :-) It's a simple watt meter, I can just set the price of energy and reset totalizers. That's it.

@FalconFly: I know that there might be inaccuracy at low watt, but here my thought. While charging the wattmeter shows 4W (in accordance to what's written on the charger, 3,5W, approximate to 4). When the battery is at 98% the consuption drops and at full charge shows alternance from no consumption to charging. Both are expected behavior. So I think it's quite accurate. I'm not pretending to get a precise number, but I think that if it tells me that my smartphone consumes 1W while crunching I'm pretty sure that's not consuming 3W.

@Bernd: Thanks for your input. If you tested and found comparable efficiency, that means that I'm definitly messing up with calculation here. :-) I thought that my Ace2 beats PCs with i7!

Anyway for what interested me, with the watt meter I found that in my case, my new Ace2 is more efficient than my old PC. That's enough for me.
Btw I'm attracted by GPUs! I might buy a Radeon HD 6450 (just 30 euro on Amazon) to increase PC crunching efficiency and let it works in the weekends.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,550
Credit: 288,971,740
RAC: 89,859

RE: Interesting to know. I

Quote:
Interesting to know. I also have a Note II; so far I've resisted putting Boinc on it. My previous phone had a lot of trouble with overheating and this one sometimes gets hot too.


While thinking of all this, you've reminded me of issues with lithium ion batteries especially toward the end of their life. You have to start counting down their lifetime when they were manufactured, and not when first used ( they degrade on shelf ) and most have a useful life of about 3 years max ( could be sooner though ). They can get rather hot with degradation of an internal separating membrane plus the concentration of lithium in metallic form rises ( explosions & fire are known ). A possible sign of this evolving may be high temperatures and higher ( than expected ) charging currents.

So check ( & replace ) your batteries !

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

joe areeda
joe areeda
Joined: 13 Dec 10
Posts: 285
Credit: 320,378,898
RAC: 0

I have a back of an envelope

I have a back of an envelope calculation.

I built an I7 3770 system with a GTX 670. It ran only E@H for about 3 weeks during burn in.

Power measured with the battery backup was 225W with GPU running 2 tasks and 110W with cpu tasks only. I had a kill-a-watt (or some other brand) hooked up for part of the time and confirmed those numbers were in the ball park.

RAC for that machine hovers around 60K. So it looks like about 60K/(225 * 24)= 11 credits/wH. Which costs me about $0.17/KwH in Los Angeles. Or about 65K credits/$US. Now I refuse to divide my total credit by that, or at least to tell my wife. Fortunately not all of that is on my home electricity bill.

I haven't done the calculation with my Android tablet (old toshiba) but it peaked at about 40 RAC which would mean equal power consumption would be 0.15 w

I think it makes a little more sense to calculate credits per watt-hour as that can be converted to cost. To convert my numbers to Watts multiply by 24.

Joe

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,550
Credit: 288,971,740
RAC: 89,859

My Galaxy Note II is at

My Galaxy Note II is at around 185 RAC after a month of 24/7 activity with four cores assigned. However that doesn't seem to have stabilised yet - it was 170 three days ago - so I'll leave it another 2 weeks to see where it evens out.

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.