What effect does lowering voltages have on wattages.

[Admetos]

I'm running my i5 750 at stock speeds, but was curious of how low voltages could go while staying stable.

I'm using a Gigabyte P55M-UD2, QPI and Chipset are both down as low a they go in BIOS. CPU is set 6 steps down from stock, and showing 1.008v while performing Intel Burn Test. Memory is down to 1.4 volt also. Currently everything is stable, but not tried going more.

I do this because at stock speeds the i5 750 does everything I want, and other then CPU cooler, running near silent with a low rpm 140mm case fan. The fan on the x-650 does not spin. In a years or so time i'll look into overclocking when I want to extend system life.

Incidentally I use to use a OCZ 600w, however that only allowed 2 stops down on CPU voltage while staying stable, as mentioned i'm 6 stops down on the X-650.

Can anyone speculate what effect running low voltages has on CPU wattage use? Anyone else here running like this, on an overclockers site guess not really..

 

[ThomasTheTankEngine]

Watts = voltage x current

So lower the voltage, you'll lower the power consumption.

 

[lord filbuster]

I reckon you would save maybe 10w at full tilt, 20w tops. I did the same with my q9450, and i measured about a 15w drop at the mains socket. There was relatively little effect at idle, 1-2w. Still, not bad for no effort, plus it will run a bit cooler.

 

[Tealc]

I find that downclocking my Blck by 15% with Gigabyte's Dynamic Energy Saver decreases my idle power usage by a few more watts, this is then added to the saving in dropping in vcore. In addition I also underclock my graphics card in 2D mode.

Not sure how I can drop it further. Dropped already from 240w to 160w at idle so I'm happy.

 

[liquidex]

I measured a drop of ~15W at load for a change of 0.05V on my e8400.

 

[Admetos]

Interesting responses, I'm down at least 0.2v at the moment. Some point i'll try even more.

As said it's down to the Seasonic X650, my previous OCZ 600w would only drop maybe a 1/3 of the voltage in comparison before system crashed on load test.

 

[Rroff]

Due to the inefficencies of most power supplies and a few other things its pretty rare to see any meaningful gains from undervolting to reduce power useage - unless your folding 24x7 or something like that and then you probably don't care about the power useage anyhow.

 

[darkwildchild79]

dont know never tried but thinking about it i might try and see what happens

 

[Admetos]

Due to the inefficencies of most power supplies and a few other things its pretty rare to see any meaningful gains from undervolting to reduce power useage - unless your folding 24x7 or something like that and then you probably don't care about the power useage anyhow.

I kind of agree. I was really only doing this to how low I could go on all voltages really, also don't like excess heat / noise if it's not required, and I guess it all adds up. The biggest waste of energy in my system is the GTX260 thats always stuck in 3d mode due to dual monitors.

Just took the CPU down another notch - 7 steps down from stock, still stable on the burn test.

 

[gurusan]

Assume power consumption is the same as TDP, so 95W, then assume 1.15V as stock voltage or something.

That means it would pull around 83A at full pelt. Then if you set voltage to .9V then it would pull around 75 watts. Some would call this significant, others wouldn't care at all.

 

[Rroff]

Thats the thing... how many people are running their PCs at full pelt for enough time for the amount saved be significant enough to be noticeable.

For your average user your talking a saving of around 30pence for a whole month if that if your really lucky, from some quick scratching your talking more in the region of 10-15pence unless you own a monster gaming PC.

 

[Gzero]

Put it in perspective: 3kw for your kettle, 300 watt for your average gaming PC load.

Kettle works for less than 5 minutes, your average gaming session ~hour.

Mind you, when my pc is really doing nothing I do try setting it to sleep... only to turn on again to see the replies.

I've underclocked my 4870 from 500mhz to 400mhz on the core, plus dropped the voltage from 1.26 to 1.16, small but helps keep the temps down and should easily shave a few pounds off the bill at the end of the quarter.

I didn't see much change on fiddling the voltages on DDR2, neither any effect on the idle wattage when lowering the FSB on my Q9550.

 

[Stanners]

You have a 3kW kettle, that is some nice hardware

 

[Big.Wayne]

Hello Rroff

This is a subject I am interested in so would welcome a polite discussion if you have time?

unless your folding 24x7 or something like that and then you probably don't care about the power useage anyhow.

It's a statement based on the assumption everyone thinks and cares about the same things as your good self . . .

Thats the thing... how many people are running their PCs at full pelt for enough time for the amount saved be significant enough to be noticeable.

I have no idea what the rest of the population are doing with their PCs and how long they have them running for . . . do you?

I'm not sure what your personal definition of "significant" is but anything I can do to save money and reduce my carbon footprint without incurring too much inconvenience is a winner in my books . . . then factor in the hundreds, thousands and millions of other computer uses like myself? . . . once you think outside of yourself I would say the difference could be very significant? . . . maybe not financially to each individual but on a global scale it could be something like switching off a nuclear power station? . . . with little or no inconvenience to any of us other than undervolting a PC?

For your average user your talking a saving of around 30pence for a whole month if that if your really lucky, from some quick scratching your talking more in the region of 10-15pence unless you own a monster gaming PC.

Would it be possible to show me the maths behind those "quick scratching" figures please Rroff ?

1. How much wattage does the PC consume when idle?
2. How much wattage does the PC consume when loaded?
3. What is the daily blended average between idle/load states?
4. How many hours a day is the P.C switched on?
5. How many days a week is the the P.C switched on?
6. How many weeks a year is the the P.C switched on?
7. How much is the user paying per kWh?
8. How many other computer users globally?

 

[Admetos]

Well i'm down 9 stops down on the voltage now. CPUZ is reporting 0.992 volt under burn test load - everything still fine. If system is still running well in a week, i'll try some more.

 

[mmj_uk]

Reducing the voltage too much will cause instabilities, you need to be stress testing the CPU to ensure any reduction in volts isn't sacrificing stability.

It strikes me as strange that anyone concerned with power usage would even buy a quad core CPU though...

 

[Duff-Man]

Would it be possible to show me the maths behind those "quick scratching" figures

1. How much wattage does the PC consume when idle?
2. How much wattage does the PC consume when loaded?
3. What is the daily blended average between idle/load states?
4. How many hours a day is the P.C switched on?
5. How many days a week is the the P.C switched on?
6. How many weeks a year is the the P.C switched on?
7. How much is the user paying per kWh?
8. How many other computer users globally?

You can work it out easily enough using primary school arithmetic...

Say, for arguments sake, 350 days per year computer usage, averaging 2 hours load, and 8 hours of idle. You can adjust these figures to suit your typical usage, or your idea of average-joe's computer usage (which will clearly be less than that). We assume a fairly high performance gaming PC, for arguments sake, which will draw around 200W under idle, and 500W under load. Add an extra 50W for the screen, which won't change dramatically between load or idle.

So, per day we have:

0.25Kw for eight hours (2.0KWh)
550W for two hrs (1.1KWh)
So around 3.1 KWh per day.

Electricity comes in at around 12p per KWh (depends on your supplier), so around £0.37 per day. Or £130.20 for your 350-day year.



-- To guess at the global computer usage you need to make a number of assumptions and approximations. The best you can do without intensive research would be to get an order of magnitude estimate (i.e. an answer correct to roughly a factor of ten). So we shall do this.

science portal estimates that there are over 1 Billion computers in the world, but the majority of these will be low-power laptops, and/or old low-power machines. This is also balanced by a very small (proportionally) number of high-power servers. For arguments sake, we shall say 75W at idle, and 200W at load, with 150 days use per year, 4hrs total usage per day, with a 3:1 split for idle:load.

This gives us, very roughly:

(0.225+0.2)*150 = 63.75 KWh per computer per year.

Multiplied by 1 billion computers, this gives a total of 63750 Gigawatt hrs.

There are around 8760hrs per year, so to maintain this average energy output you would need a power plant of capacity 7.28 Gigawatts.

A typical modern nuclear power plant produces on the order of 0.5 to 2.0 gigawatts (according to wikipedia, so we can assume around seven modern nuclear power plants are required to power all the world's computers.


As for how much of a difference under-volting the CPU would make... Well, the vast majority of computers in use around the world are NOT high-performance machines. I would estimate (very roughly) that around 5% of power output comes from high-powered machines such as servers and enthusiast PCs. So, a (say) 10% reduction in power output from each of those machines represents around a 0.5% reduction in total power output from PCs. This is around than 3.6% the output of an average (1GW) nuclear power plant. And this is being optimistic, assuming that server PCs (which make up the vast majority of high-power PCs) can be reduced in power without sacrificing number-crunching output. In reality, reducing their power output would lead to a reduction in performance, which would in tern lead to more PCs being bought.

Anyway, that might give you an idea of scales involved.

 

[gurusan]

I was surprised when I found my E8400 used around 10-15W idle when clocked to 4.5GHz....it used pretty much exactly the same power at the stock 3GHz.

 

[Big.Wayne]

Hi Duff-Man,

You can work it out easily enough using primary school arithmetic...

Are you implying that anyone present can't do basic arithmetic?

I was asking Rroff how he arrived at his figure behind his statement? . . . figures that involve a number of factors outside of basic arithmetic like "assumptions and approximations? . . .

for arguments sake

There is no arguement involved . . . it's up to each individual to do the maths for themselves . . . and be aware the resulting figures are bespoke to them and their own hardware/usage/kWh cost ££ and not the whole world . . .

To guess at the global computer usage you need to make a number of assumptions and approximations.

Indeed! . . . good effort on your behalf though! . . . . if your "assumptions and approximations" are in anyway factual then the idea of closing down a nuclear power station from undervolting a billion computers doesn't seem realistic . . . however reducing the global power usage by 36 million kWh sounds awesome!

 

[Big.Wayne]

I was surprised when I found my E8400 used around 10-15W idle when clocked to 4.5GHz....it used pretty much exactly the same power at the stock 3GHz.

Did both chips use the same vCore while idle? . . . I guess not but not sure how that works then?

How about the difference in wattage between a stock 3.0GHz vCore under full load an the same chip at 4.5GHz under load?