How much extra power does your overclock consume?

[ShawnD1]

I've been screwing with my overclock settings and looking at power demand reported by my UPS. My processor is an E6600 which has a TDP of 65W. When I change the clock from the stock 2.4ghz to a nicely overclocked 2.9ghz, power consumption increases by 50W. That means I get roughly 20% overclock and it consumes 77% more power.

For those of you who have extreme overclocks like a 3.8ghz quad core, how much extra power does overclocking consume? Have you ever measured it?

 

[fffblackmage]

On stock cpu voltage and clock frequencies (1.86GHz, 1.35V), my computer uses around 140W on idle.

Undervolting my cpu to around 1.125V saves me around 10W. For my 4830, using 300/450 MHz for 2D clock frequency saves me about another 10W, compared to 3D clock frequencies (800/1000 MHz).

Using 1.4V for cpu voltage and overclocking to around 3 GHz increases power consumption by about 20W over stock (on idle).

I think my system used around 240W or so under full load (orthos+furmark) when the cpu is at 1.86GHz, 1.125V undervolt. Total power consumption increases to just under 300W with my maximum overclock on the cpu (~3 GHz, 1.4V). In both cases, my 4830 is at 800/1000 MHz, voltage is fixed.

These numbers are approximations. I did actual measurements using my Kill-A-Watt meter, but I don't have the real numbers saved.

The increase in power consumption when overclocking shouldn't come as a surprise. It's why heat dissipation is such a huge issue.

 

[Acanthus]

I'm sure mine is substantial.

 

[Rifter]

I dont want to know, but i had to upgrade my 550W PSU because of instability, i went to a 850 for some breathing room.

 

[ShawnD1]

I think my system used around 240W or so under full load (orthos+furmark) when the cpu is at 1.86GHz, 1.125V undervolt. Total power consumption increases to just under 300W with my maximum overclock on the cpu (~3 GHz, 1.4V). In both cases, my 4830 is at 800/1000 MHz, voltage is fixed.

That's pretty damn good. Check this out:

E6600 at 2.9ghz, 1.30V, video card is an 8800GTX at stock settings. I should get a frying pan shaped heatsink so I can fry some eggs 😀

 

[aigomorla]

lol u guys dont want to know how much nadeshiko eats.

let me put it to you this way.

IF she was a car, she'd be a Indy car. LOL..

Before the gulftown i was over 500W on full load.

Stupid X2's on Xfire eat a lot of watts.

 

[faxon]

hate to burst your bubble, but the kill-a-watt isnt always accurate when measuring power draw on a PFC power supply.
http://www.jonnyguru.com/forums/showthread.php?t=3019&highlight=kill-a-watt&page=4

 

[MrK6]

I was curious about this so I tested my i5 750 at three different settings. For idle, I measured the lowest consistent power draw on the desktop, for load I used prime 95 64-bit with 4 threads on large FFTs and took the highest consistent power draw. "Real" voltages were measured using CPU-Z. Power draw was measured from the wall using my Kill-A-Watt. Details and results:

Current overclock - 4.0GHz (190x21) @ 1.35V BIOS, 1.344V by CPU-Z, no EIST, no C1, no C4,C6,C7, rest of set-up in sig
Desktop Idle - 123W @ 1.344V, Load - 267W @ 1.344V

Stock - 2.66GHz (133x20), Turbo on (2.8GHz load), "Normal" voltage set in BIOS, EIST on, C1 on, C4,C6,C7 on, RAM @ 1333MHz
Desktop Idle - 76W @ 0.848V, Load - 180W @ 1.248V

Undervolt - This is a quick and dirty undervolt I did while I was waiting for my Noctua cooler to come in when I first built the system (was on the Intel stock heatsink, which is a loud POS) - 2.66GHz, Turbo off, EIST on, C1 on, C4,C6,C7 on, RAM @ 1333MHz
Desktop Idle - 79W @ 0.976V (lolwut), Load - 138W @ 0.992V

Pretty interesting I'd say. Obviously, turning off power saving features kills the idle power efficiency of the chip, but it greatly stabilizes the overclock at lower voltages, which is a trade-off I'm more than happy to make. This chip is still much more efficient than my Q6600 was, and come this summer if the heat is a problem, I can always tinker around with the power saving settings at a lower overclock. Anyway, hope the data was insightful.

 

[ShawnD1]

I was curious about this so I tested my i5 750 at three different settings. For idle, I measured the lowest consistent power draw on the desktop, for load I used prime 95 64-bit with 4 threads on large FFTs and took the highest consistent power draw.

Interesting. I was playing with stress testing tools and watching the power draw on the UPS. Power draw in one particular set of tests:
Prime95 small FFT - 318W
Prime95 large FFT - 311W
Intel Burn Test - 265W

Some "burn" test eh? This computer has F@H running in the background all the time and power draw actually decreases when running the Intel Burn Test. I wonder if Intel supplying horrible stock coolers has anything to do with their stress test software putting nowhere near as much load on the CPU as a regular program like F@H does.

 

[Markfw]

What I would love to see, is the usage of my I7 920@3.7 @100% W/HT, and both video cards @100% (my 24/7 usage). Just got another power bill, and with cheap hydro power, it was $325, and I have 3 systems (6 video cards) turned off.