no measurable reduction in power consumption/cpu-freq

[graysky]

I have a Kill-a-watt that measures a power draw in Watts. I was curious to see what sort of real energy reduction the acpi in my kernel was affording to me so I measured the idle system under the two available multipliers both at the stock level for my system, and the overclocked level.

The bottom line is that there was no difference!

Hardware Details:
CPU: Intel X3360 (Xeon version of the 45nm quad-core Q9550)
Memory: 2x2 Gb DDR2
Motherboard: DFI LT LP P35-T2R
HDD: 2x Seagate 7200.11 (640 Gb)
Fans: 5x120mm 1500 RPM fans
Videocard: PCI-E 16X 8400GS

The stock setting was a FSB of 333 MHz under both the 6.0x and 8.5x multiplier w/ the RAM running @ 1,066 MHz. The low idle was 2.00 GHz and the full idle was 2.83 GHz. The overclocked setting was a FSB of 400 MHz under both the 6.0x and 8.5x multiplier w/ the RAM running @ 1,000 MHz. The low idle was 2.40 GHz and the full idle was 3.40 GHz.

Under stock conditions, the power consumption idle (just logged into Gnome with nothing running) was 92 Watts @ 2.00 GHz and also 92 Watts @ 2.83 GHz. Under overclocked conditions, the power consumption idle (just logged into Gnome with nothing running was 98 Watts @ 2.40 GHz and also 98 Watts @ 3.40 GHz.

What gives? I was led to believe that adjusting the CPU frequency saved power on idle. According to these readings, there is no difference between the two multipliers.

 

[Idontcare]

Graysky that does sound very suspicious. I too have a kill-a-watt and it reports sizable power consumption differences when I move around my multiplier and Vcc for idle and load.

Do you have a plug-in-lamp to reference checking your kill-a-watt to see if it has gone bonkers? (plug in a lamp with 100W bulb, measure power consumption, then swap bulb for a 60W, then a 40W or 25W)

 

[graysky]

Yeah, it shows differences in different wattage bulbs... when you adjust your multiplier, do you leave the vcore constant?

 

[Idontcare]

Interesting...Yeah I have the data and it was all for the same Vcc, all I changed was the multiplier. Let me dig up my notebook and pull the numbers so you can see.

The data was generated as part of the tests that went into this graph (which does focus on Vcc, but I have same Vcc at different GHz too as you can see there are two data points already in that graph).

 

[fffblackmage]

Off the top of my head, I don't happen to remember the formula exactly, but power dissipation is calculated by something like voltage square times freq times something else (P = (V^2)*freq.*something).

I just know for sure that power dissipation has the greatest change with the voltage, and freq has a much smaller change, though it should be noticeable.

 

[nyker96]

it could be smaller change if you don't increase vcore much but just 6w diff from 2,4 to 3,4 does sound suspicious. I'd put it around 15-20W diff. since you didn't push vcore the power increase on the cpu is linear to speed. i think you have a 40% increase so the power is 40% more but that's just on the cpu but you are measure system wattage.