A very nice and detailed article. There's some nice background information (e.g. the temperature dependence of the power consumption) and the main advantage is that they measure the GPU power consumption directly, not just the entire system.

They are testing with FurMark and 3D Mark, though. Both should generate a higher load than GPU-Grid.

MrS
____________
Scanning for our furry friends since Jan 2002

If one doesn't have a power-meter there may be a way to estimate GPU-Grid power consumption:

1. Set your fan to a fixed speed. Not a very low one, we're going to need 3D mode.

2. Note your idle temperature, check that the clock speeds are the same as mentioned in the article.

3. Do some 3D Mark runs where you can find the real power consumption of your card in the article. Several loops may be neccessary until the temperature stabilizes.

4. Log the temperature (e.g. GPU-Z with "refresh while in background").

5. Now look for the idle and 3D Mark power consumption of your card in the article.

6. Note your temperature while running GPU-Grid, give it some time to stabilize.

7. Interpolate linearly.

Idle
Your Temp: 40°C
Measured power consumption: 40W

3D Mark load
Your temp: 60°C
Measured power consumption: 100W

GPU-Grid
Your temp: 50°C
-> approximate power consumption = (100W-40W)/(60°C-40°C) * (50°C-40°C) + 40W = 70W

And if you run seti@home on your GPUs you should see less power draw than under GPU-Grid.

MrS
____________
Scanning for our furry friends since Jan 2002