power consumption of my x3360 (kill-a-watt)

[graysky]

Got bored yesterday, so I ran checked out the power consumption of my X3360 system for shits and giggles with my kill-a-watt.
Pic of the Kill-a-Watt

Powered Hardware Details:

Intel X3360 (C1 stepping) @ 8.5x400=3.40 GHz
266/667 MHz Strap CPU Vcc=1.12500
NB Core=1.370 CPU VTT=1.310

DFI LT P35-TR2 (no modifications)

eVGA 8800 GTSG92 512 meg
(770/1,923/2,000 MHz : Core/Shader/Memory)

Corsair Dominator DDR2-1066 (TWIN2X4096-8500C5DF)
2x 2Gb @ 5-5-5-15 (performance level 6) @ 1,000 MHz (4:5) @ 2.100V

Corsair HX620
2x HDs (seagates, one a 10th generation and the other an 11th)
1x DVDROM
4x120 mm Tricool fans (came with the p182 case) all on low
1x120 mm S-Flex SFF21F (1600 RPM) on the HS
1x40 mm silent fan on the NB

That's it... no other hardware (speakers, monitor, etc.) was in the loop.

Kill-a-Watt Readings
Idle in BIOS screens - 158 W
Idle in BIOS screens with both HDD's unplugged - 144 W

Idle (with speedstep active in XP x64) - 137 W
Load (prime95 v25.6 small FFT) - 213 W
Load (prime95 v25.6 large FFT) - 216 W
Load (prime95 v25.6 blend) - 210 W
Playing Crysis - 237-241 W
x264 encode - 197 W
Standby - 0 W

No real point to this post beyond just trivial information! It is interesting to me that the large FFTs consistently read 3-4 watts higher than the small ones did which is inline with the software's description of the large FFTs as generating the max heat/power consumption. Here is the similar analysis of my older Q6600-based system, but there are too many changes (MB, settings, video board, memory, etc.) for a "which processor is more power efficient" comparison.

What is little bit crazy is that both systems draw the same while idle, and even @ 3.4 GHz, and faster memory, the X3360-based system uses less wattage on p95.

 

[Idontcare]

Would be interesting to see the power #'s when your CPU is at stock, for idle and load.

 

[graysky]

I did something similar to that w/ my q6600 (see link in first post). 7 % reduction idle and 18 % reduction on load. Problem is that I haven't minimized my vcores for the stock 8.5x333 setting.

 

[myocardia]

Originally posted by: graysky
I did something similar to that w/ my q6600 (see link in first post). 7 % reduction idle and 18 % reduction on load.

There isn't a link in your first post, except the pic. Interesting, though. That means that say, two Q6600's @ 50% overclocks are much more power efficient under load than the three Q6600's it would require to replace their workload. I assumed that overclocked, the efficiency would be considerably lower than that, making the two 50% overclocks use roughly the same amount of power as the three stock parts.

 

[Idontcare]

Originally posted by: myocardia

Originally posted by: graysky
I did something similar to that w/ my q6600 (see link in first post). 7 % reduction idle and 18 % reduction on load.

There isn't a link in your first post, except the pic. Interesting, though. That means that say, two Q6600's @ 50% overclocks are much more power efficient under load than the three Q6600's it would require to replace their workload. I assumed that overclocked, the efficiency would be considerably lower than that, making the two 50% overclocks use roughly the same amount of power as the three stock parts.

Your assumptions would be correct if we were solely comparing CPU power-consumption for stock versus non-stock...but there is a non-zero power-consumption overhead built into this current analysis owing to the ram/chipset/hdrive/GPU/etc power consumption which scales per system count and not (considerably) per overclock percentage.

Now if we were talking about power-consumption of OC'ed versus stock CPU's on an n-socket mobo (say Skulltrail) then the non-CPU power consumption overhead would be effectively a constant and then CPU-specific power consumption conclusions could be extracted from the data.

 

[graysky]

Originally posted by: myocardia

Originally posted by: graysky
I did something similar to that w/ my q6600 (see link in first post). 7 % reduction idle and 18 % reduction on load.

There isn't a link in your first post, except the pic.

Sure there is... the word "Here" in the sentence, Here is the similar analysis of my older Q6600-based system, but there are too many changes (MB, settings, video board, memory, etc.) for a "which processor is more power efficient" comparison.

 

[zach0624]

I just have to say that those are some killer numbers, for a quad 241 under load is awesome.

 

[myocardia]

Originally posted by: Idontcare
Your assumptions would be correct if we were solely comparing CPU power-consumption for stock versus non-stock...but there is a non-zero power-consumption overhead built into this current analysis owing to the ram/chipset/hdrive/GPU/etc power consumption which scales per system count and not (considerably) per overclock percentage.

I'm not sure why I didn't think of that, since it's obvious. Thanks for pointing it out, though.

Originally posted by: graysky
Sure there is... the word "Here" in the sentence, Here is the similar analysis of my older Q6600-based system, but there are too many changes (MB, settings, video board, memory, etc.) for a "which processor is more power efficient" comparison.

Okay, it seems yesterday wasn't my day, as far as brilliance goes. I honestly looked for it, and didn't see it, though I maybe I didn't look any lower than the Kill-A-Watt readings. At least I hope that's why I didn't see it.

 

[Drsignguy]

Here is a Linky for all to look over as each state differs in charges. (just for our info). This is now putting my mind at work to see what my charges per month really are. Nice....