Power Tests : Voltage and C2Ds

MattC

Distinguished
Oct 1, 2004
132
0
18,680
Hello! I hope you all find this useful.

I tested my new system with the Kill-A-Watt meter under a variety of conditions. I am posting the more interesting of the results here.

First, specifications:
EVGA 8800 GTS 320
Gigabyte DS3 (rev 3.3)
C2D E4320
G.Skill Micron D Memory
OCZ GameXStream Powersupply, 600w
No discreet sound card
Asus DVD
Samsung HDD (an older 7krpm 160gb drive)

The power draw is not including the monitor, which is a Samsung SyncMaster 940b and draws 36 watts during use.

(comment about frequency and power use removed - not sure what I was thinking when I wrote that, as it contradicted my entire post )

With the CPU Voltage at 1.1 and the system idling (1600mhz, 6x266), the power draw is 121 Watts (not bad, but I'm guessing AMD can crush that number)

At load, the system runs at 2.314 ghz (330.7x7), same voltage settings. Power draw is 152 Watts.

Increasing the Voltage to the stock voltage settings (1.35) at the same clock settings yields and idle use of 131W and a load use 180W

Using this information we can speculate as to the increased heat dissipation from the cpu that results from increasing the CPU Vcore from my undervolted 1.1 volts to the stock 1.35 volts. Assuming an 80% efficiency, an extra ~8 or so watts at idle and 24 watts at load are going through the cpu and probably the mb. If you want an idea of how much heat that is, try feeling some different-wattage bulbs. Your typical compact flourescent bulb using 9-13 watts, and a nice bright one uses around 20. leave one on for a while to see how much extra heat that cpu heatsink will need to dissipate.


A little side note / review:
The gigabyte ds3 rev 3.3 (f10 bios) has built in overclocking on demand, and it works rather well. While this cpu can't scale as much as the AMD CPUs, going from 1.6 to 2.3 is quite nice. I may work on finding a way to further reduce voltage when idle (perhaps RMclock?), since 1.1 volts is enough for full load (and therefor overkill for idle speeds).


p.s. Yes, I know others have done tests like this already. I think having a little redundancy is nice, and if you don't like it you don't have to read it.

Edit 1: Load = TAT full load. I was not stressing the GPU in any of these tests, though I fully intend to test that, along with a variety of other things. I actually just got my kill-a-watt meter, so I have just begun playing with it.
 
First off, Frequency does not seem to have any significant effect on power draw. I tested this very briefly, so it may be be entirely correct. I will update when I have tested this again.
With a constant voltage, CPU power is fairly linear in relationship to frequency.

With the CPU Voltage at 1.1 and the system idling (1600mhz, 6x266), the power draw is 121 Watts (not bad, but I'm guessing AMD can crush that number).
This figure says more about the 88xx series GPU’s being power hogs than anything else.
I measured my Core 2 Duo system at 2.67GHz 1.28V under full load and it consumed 126W at the wall. I used Prime95 at the power hungry setting.
 
I kept some records from before, for comparison -

Specs:
C2D x6800
Asus P5W-DH Deluxe
One stick of 1GB RAM at 2.2v, 3 FSB:4 DRAM
Radeon x1900xtx - stock
Antec TP3-650 PSU
No discrete sound
1 idle DVD drive
1 HDD 7200rpm 320GB

1.225vcore, 273*11 = 3.0GHz idle 137w load 151w
1.325vcore, 291*11 = 3.2GHz idle 153w load 173w game 210-235w
1.4875vcore, 329*11 = 3.6GHz idle 179w load 208w

Load stressed integer units. An FPU load like Orthos-small would draw 4-7w more. Game was a single-player UT2004 script with the integer load running in the background at a lower priority. CPU-z does not show any downclocking when idle (power-saving deactivated from BIOS).

Either our boards adjust Vcore differently, or our power supplies are not equally efficient. The Antec model was marketed as high-efficiency (>80%), though this generally breaks down at low wattage.
 
For comparison:

Pentium D 805 @ 1.15 vcore
Asrock 775i65G
2x256MB PC 3200 ram
GF3 Ti200
200GB Seagate 7200.7
80GB Seagate IV
Winfast Capture Card
Aureal sound card
12X DVDRW
12x DVDrom
3.5 floppy
250 watt Dell PSU

Idle: 100 watts. single prime: 125 watts. dual prime: 150 watts

Same as above, but with IGP instead of GF3 (10 watts), no capture card (5 watts), AC97 instead of sound card (5 watts), an no 200G hdd (7 watts).

Idle: 73 watts. dual prime: 123 watts.



Dell inspiron 1200 Pentium M laptop, 600MHz 0.700 vcore, GMA 900, 14" TFT, 30GB 4200rpm hdd, 256MB pc2700 DDR ram, 24xCDRW/DVD combo, PCMCIA wifi card.

Idle (lowest backlight setting): 12 watts. Load: 15 watts. No display: 9 watts. Standby: 1 watt
 
Heh, I have the kill-o-watt meter.. Just too lazy to do the readings, and remember the readings I had on my P4 setup.

I was going to take pics of the meters and use 3dmark 2003 or 05 to do load tests with it and prime95. TAT wouldn't work on my current system, but on my dads E4300 it does.

But.. like I said.. I'm just too lazy. 😳