Possible mistake in case study

[samuBe]

Hi,

I was reading the case study in the sci repo (sci/case-studies/eshoppen.md at main · Green-Software-Foundation/sci · GitHub). During this, I noticed a few irregularities of math that doesn’t really add up, mainly with the energy part.

The formula:
E = Server utilization * Number of hours * Number of cores * TDP * TDP co-efficient
= (0.18 * 1 hour * 2 cores * 205 TDP * 0.32 TDP co-efficient)/1000
= 0.023
Contains two mistakes (according to me):
The usage is considered twice: 0.18 and 0.32, more explanation below.
The TDP is for the entire chip not for a single core, so I believe it should be 2/16. From Intel: Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload.

For the first point, let’s instead of using a TDP co-efficient, use the power curve (https://www.researchgate.net/figure/A-typical-CPU-Utilization-Server-Power-scatter-plot_fig4_352107996) directly. I think if the CPU uses 20% of the entire resource it will use 60W. If 2 cores of the 8 are used, 15W. A carbon intensity of 200gCO2e/kWh results in 4 gCO2e/hour of this resource.

Of course, I could be wrong and would be happy to be.

Thank you so much for your attention and participation.