Real Power Consumption

[Phaedrus2129]

I just got a new Kill-a-Watt (old one got lost like a year ago) and I'm measuring the AC pull from the wall for my main system.

That is,

Q9550 @3.4GHz 1.05V
Asus P5Q Pro
4GB OCZ DDR2 1066 CL5
2x Sapphire Radeon HD4870 1GB @775/1000
Creative X-fi XtremeGamer
Two mechanical hard drives and one SSD
DVD+/-RW drive
Five 120mm fans, one 92mm fan, all with LEDs
Antec HCP-1200 engineering sample



Methodology:
The Kill-a-Watt power meter is plugged into a power strip, and my computer's power cord is plugged into the Kill-a-Watt meter. I perform the computer task I am testing and watch the Kill-a-Watt's display. I record as the continuous range the lowest and highest readings sustained for at least one second. I will also record any notable <1s peak readings. All readings are rounded to the nearest 5W. I then calculate an approximate DC wattage based on a graph of my PSU's 120VAC efficiency.

PSU efficiency:
I've acquired an efficiency curve graph for this power supply. It's a different marque, so efficiency may vary, but this will give an approximate idea as to DC power consumption.


Misc:
Standby: 0.8WAC (probably an erroneous reading--Kill-a-Watt isn't too accurate that low)
Startup: 60-250WAC (280W peak) - 40-220WDC (245W peak)
Idle on desktop: 245-250WAC - 215-220WDC
Heavy Multitasking: 245-255WAC - 215-220WDC
Music from CD: 255-260WAC - 220-230WDC
720p Youtube: 255-260WAC - 225-230WDC

Stress Tests:
3DMark06 GPU test 1: 325-360WAC - 290-320WDC *
3DMark06 GPU test 2: 325-360WAC - 290-320WDC *
3DMark06 CPU tests: 285-300WAC - 250-270WDC *
3DMark06 GPU test 3: 355-400WAC - 315-360WDC *
3DMark06 GPU test 4: 315-325WAC - 280-290WDC *
Intel Burn Test (High): 340-345WAC - 300-305WDC
Prime95 Large FFTs: 315-345WAC - 280-305WDC
Furmark Multi-GPU 1.8.2: 405-415WAC - 375-390WDC **
Furmark+IBT: 430-500WAC - 400-455WDC **

*3DMark06 was tested at the default settings; that is, 1024x768 resolution, low AA&AF.
** Furmark was tested at 1920x1080, Xtreme Burn mode, displacement mapping, PostFX, MSAA x8

Games:
(All games are tested at 1920x1080, or if not then at their highest available resolution, with all available graphics settings set to maximum)
Crysis (Typical level): 350-380WAC - 315-345WDC
Crysis (Final level): 390-420WAC - 360-395WDC
Just Cause 2: 330-385WAC - 295-350WDC
Fallout: New Vegas: 335-355WAC - 300-315WDC
Minecraft Beta: 280-320WAC - 245-285WDC
Super Meat Boy: 250-260WAC - 220-230WDC


Highest load seen: 455WDC / 500WAC
(Under Furmark + Intel Burn Test)


Upcoming:
Maybe some more games?

Final note:
Back when I had my original Kill-a-Watt and only one 4870 I recorded as the maximum in Furmark+IBT just 325WDC, and I recall gaming usually being in the 200-275W range. I don't have those precise figures though. This does indicate that a second 4870 added approximately 130W maximum; since the second card doesn't consume as much power as the first, and a single 4870 is recorded as taking about 150-170W on its own, this jives with common knowledge.

 

[E4g1e]

Actually, the Kill-A-Watt is also inaccurate: The readings can be off by as much as 30 percent, especially if your PSU has active PFC. In fact, KAW is no more accurate than the load meters built into many UPSes.

The only way to accurately determine power draw from the AC wall outlet is to get a load testing device that costs hundreds of dollars.

 

[Phaedrus2129]

Kill-a-Watt *can* be that inaccurate, but isn't always. And the APFC glitch makes the wattage appear larger, so if my numbers were off by 30% they'd be 30% high, but not low. Which would make the resulting wattage lower.

What I'm seeing matches pretty well with what I'd expect from this hardware.



Anyway, if you'd like to buy me a Brand power meter, I'd be much obliged.

 

[[Spectre]]

Kill-a-Watt *can* be that inaccurate, but isn't always. And the APFC glitch makes the wattage appear larger, so if my numbers were off by 30% they'd be 30% high, but not low. Which would make the resulting wattage lower.

Um, no when it is wrong it reads low.

 

[Phaedrus2129]

Um, no when it is wrong it reads low.

I thought it was the other way around. :/

 

[E4g1e]

I thought it was the other way around. :/

If you account for APFC in most modern PSUs, your rig would actually be drawing closer to 600W AC from the wall (assuming the 15 percent typical inaccuracy of the Kill-A-Watt with APFC-equipped PSUs).