Sign in with
Sign up | Sign in
Your question
Solved

Overclocking and power consumption

Last response: in Overclocking
Share
February 25, 2012 2:00:51 PM

I was just using a psu calculator and when I used the overclock function my wattage doubled . If your planning to oc to say 5g what is the voltage ( never oc'd before ) and does this sound right or did i do something wrong ? :whistle: 
a c 190 K Overclocking
February 25, 2012 3:04:36 PM

That would depend on your hardware, and yes, overclocking requires more power, just as higher revs on an engine makes the car go faster,
and as a result creates more heat to get rid of, another overclockers issue :) 
Its hard to say voltages for 5GHz off the question because you haven't posted your specs, even then its only guideline figures, my 2500k might reach 5GHz at 1.325v but your may need 1.4v, a lot of clocking is finding your particular hardwares sweet spot, and a deal of that involves trial, error and patience :) 
Moto
m
0
l
Related resources
a c 190 K Overclocking
February 25, 2012 6:27:02 PM

No, it was a generalisation showing that what one chip requires for a successful clock maybe insufficient for another chip, I should have pointed that out :) 
1.325v is in my brain because thats what my 975BE takes for 4GHz
Moto
m
0
l
February 25, 2012 7:17:33 PM

K . I watched a couple of youtube vids and they all stated 1.3 to 1.4 as you stated but back to the required wattage .. my system ( proposed ) went from 745 to 1447 when I added oc .... Im thinking I must of mussed up ....the system is i5 2500k at hpefully 5g oc . asus p8z68 mobo nvidia 570 x 2 sli , rx480x2 based waterloop with cpu and gpus . and 120 mm led x 8 , 120 mm x6 , 140 mm led x 3 .....
m
0
l
February 25, 2012 8:12:19 PM

Increasing frequency, even without an increase in voltage, does increase power consumption.

http://en.wikipedia.org/wiki/CPU_power_dissipation

Quote:
For a given device, operating at a higher clock rate always requires more power.
...
The power consumed by a CPU, is approximately proportional to CPU frequency, and to the square of the CPU voltage:[1]

P = C*V^2*f

(where C is capacitance, f is frequency and V is voltage).


That formula is true for dynamic power but there is also static power which will not change with frequency but will go up with a voltage increase. It is basically the same formula but without the "f", so P = C*V^2

Modern CPUs can consume as much static power as dynamic power, so assuming that is the case then by doubling the frequency it would consume 50% more power. For example if static and dynamic power at stock were each 50W, then doubling the frequency you would consume 50W static and 100W dynamic for 150W.

Then you have to figure that is only for frequency increases. If you increase the voltage your power increases proportional to the square of the voltage, so doubling your voltage (not that you would) would cause you to draw 4x as much power.
m
0
l
a b K Overclocking
February 26, 2012 3:07:30 AM

You must have messed up typing in the calculator. With 2x570, a 850w psu would let you oc as high as the i5 can go.
m
0
l

Best solution

a c 197 K Overclocking
February 26, 2012 4:01:01 AM

Sigh.

Forget dynamic and static power. For estimating CPU power consumption, they are meaningless.

Real simple power estimate (based on actual measurements):
Power comsumption of highly overclock CPU = TDP *1.25.

Example:
VID of my Q6600 = 1.2625 volts. I actually discovered that the chip will reliably run at stock speed on 1.20 volts.

OC'd to 3.6 GHz, core voltage = 1.425 volts. So 1.425^2/1.2625^2 = 1.27.
This is not quite valid because a CPU is not a resistive load, but as I used to say, "Close enough for government work."

90 watts TDP *1.27 = 114.7 watts.
TDP is just an approximation of CPU power consumption. It is based on every CPU pin toggling, and that normally will not happen. So TDP is really a little higher than actual CPU power consumption.

Actual measurements:
Using a lab calibrated clamp-on ammeter, I measured 8.5 amps (102 watts) going into the CPU power plug. OC'd, I measured 10 amps (120 watts). These figures include the losses in the motherboard power regulator, so CPU power consumption will be a little less.

My Kill-A-Watt power meter indicates 19 watts more power consumption when the Q6600 is overclocked.

Overclocking my Q9550 yields similar results.

For more of my thoughts about system power consumption:
http://www.tomshardware.co.uk/forum/forum2.php?config=t...
Share
February 26, 2012 10:43:00 AM

Out come the techies . Ty all for the replies . Theres alot of info here and I appreciate the time you put in to them . I may just learn something yet . I did re-visit the calculator and I must have entered something wrong because I came out with 830 something as being needed . K114 you were very close . So for some headroom I'm thinking 1000 w ... What are your thotts ?
m
0
l
February 26, 2012 12:28:00 PM

jsc said:
Sigh.

Forget dynamic and static power. For estimating CPU power consumption, they are meaningless.

...

OC'd to 3.6 GHz, core voltage = 1.425 volts. So 1.425^2/1.2625^2 = 1.27.
This is not quite valid because a CPU is not a resistive load, but as I used to say, "Close enough for government work."


My post wasn't meant as a means to estimate his power consumption but just as an example that increasing frequency and voltage will significantly increase power consumption. I like your calculation though O.C. V^2/Stock V^2 = wattage multiplier. Simple and easy. :) 

k1114 pretty much nailed it though you need at least an 850W which would give you a little bit of headroom but a 1000W is not a bad idea either. Just depends on your budget I guess. In guru3d's 570 SLI review that's exactly what they come up with too. They measure almost 600W consumption with the GPUs being stressed (but not the CPU).

http://www.guru3d.com/article/geforce-gtx-570-sli-revie...
m
0
l
February 26, 2012 1:07:06 PM

Cool ty again all . Im gonna go over this thread many times I think to try to wrap my head around all the ( techy) input ... Thinking I should just go take an A+ cert course .. I think 1000w would make me comfortable about having room for any future upgrades anthough http://www.newegg.com/Product/Product.aspx?Item=N82E168... looks nice too :)  if I digress into a fan controller question are the powers that be gonna slap me ?
m
0
l
a c 190 K Overclocking
February 26, 2012 1:12:09 PM

Probably not, but you can always mark your best answer on this one and start another up, just to be safe :) 
Moto
m
0
l
February 26, 2012 2:32:26 PM

Holy you could run your whole house off DC power with that thing! lol j/k
m
0
l
February 26, 2012 3:03:51 PM

Motopsychojdn said:
Probably not, but you can always mark your best answer on this one and start another up, just to be safe :) 
Moto

Ok your probobly right ..I'll start another thread ..watch for it .. ty all
m
0
l
February 26, 2012 3:05:42 PM

Best answer selected by rufus_22.
m
0
l
!