testing power consumption ...

[Copenhagen69]

I am just curious how people test power usage for single parts in their system. Like Idle usage and Load usage.

Are there any good guides are articles on how this is done?

Looking specifically for GPU and CPU stuff.

I think this

could tell me load stuff (maybe?)... but nothing specific ...

also what could the P3 Kill A Watt monitor be useful for? Could I get a base idle wattage and then load something like the GPU and see how much the wattage goes up? or what?


thoughts/ideas/links would be greatly appreciated!

 

[coolVariable]

Kind of sad that it's impossible to tell the power consumption with some kind of application. That would be neat.

 

[PottedMeat]

That would involve putting current sensors on each PCI/PCI-e port power lines - $$, plus sensors on external lines like PCI-e Aux connectors. Then you need to do A/D conversion on each sensor input or bring them together to a multiplexer which would be a mess on an already densely traced motherboard. The interfacing part wouldn't be too bad though; they could be attached to the I2C or SMBUS I suppose.

Could I get a base idle wattage and then load something like the GPU and see how much the wattage goes up? or what?

That's pretty much what everyone does.

 

[JackMDS]

Yeah there is a good tool, it called your Eyes. :sneaky:

Plug the computer into the Kill a watt and press the button that shows consumption.

Start the computer and wait until the boot is done wait a minute or two

Eyes on the Kill-A-Watt, and the reading will tell you idle.

Let it sit until it goes to sleep, and the reading will tell you Sleep consumption (if any).

Choose a Multi Bench mark and run it look at the Kill-a-Watt, and record the consumption according to the stage of the Bench Application.

This bench mark is rather short and goes trough all the stages.

Look here Download the portable version of CrystalMark 2004R3.

http://crystalmark.info/download/index-e.html

If you use for long time applications that working through with No gaps (like ripping audio, encoding Video, Downloading huge files).

Use your favorite applications to see what they take as compare to the previous generic reading.



😎

 

[coolVariable]

That would involve putting current sensors on each PCI/PCI-e port power lines - $$, plus sensors on external lines like PCI-e Aux connectors. Then you need to do A/D conversion on each sensor input or bring them together to a multiplexer which would be a mess on an already densely traced motherboard. The interfacing part wouldn't be too bad though; they could be attached to the I2C or SMBUS I suppose.




That's pretty much what everyone does.

Ugh?
1 sensor in the PSU. Done.

 

[kalrith]

I like using prime + furmark to get maximum load for my system. You could also use just prime for cpu usage and just furmark for gpu usage, but obviously furmark's going to use the cpu as well.

 

[kalrith]

Ugh?
1 sensor in the PSU. Done.

I think he was stating what would have to be done to test each component individually, much like AT recently did in their Core i7 980X review:

Raja has been busy with his DC clamp meter measuring actual power consumption of CPUs themselves rather than measuring power consumption at the wall.

That reminded me of another important thing to point out. The Kill-A-Watt measures AC, whereas your components use DC. For example, if the Kill-A-Watt says your system is idling at 65W, it's probably using something like 55W, and the extra wattage is lost to heat through the inefficiency of the psu.

 

[M0RPH]

OP, if you want to test the power usage of your video card alone, you'll need a computer with integrated graphics so you can run the computer with and without the video card installed and compare the power usage.

 

[kalrith]

OP, if you want to test the power usage of your video card alone, you'll need a computer with integrated graphics so you can run the computer with and without the video card installed and compare the power usage.

Either this or if he has SLI/Crossfire and tests the difference between 1 and 2 video cards. Some review sites test the consumption this way, so it would be easier for him to just read some reviews.

 

[Copenhagen69]

thanks for all the input guys i will try all the suggestions out and see what comes of it