Comparing system idle across large generations is somewhat tricky, but we can pull some ballpark comparisons. A ~2013 workstation Xeon has a system idle of ~75

Comparing system idle across large generations is somewhat tricky, but we can pull some ballpark comparisons. A ~2013 workstation Xeon has a system idle of ~75W (https://www.anandtech.com/show/7852/intel-xeon-e52697-v2-and-xeon-e52687w-v2-review-12-and-8-cores/2 https://www.anandtech.com/show/7852/intel-xeon-e52697-v2-and...). This should be similar to the HP system from the start of the linked article. Modern system idles sit in the ~50W range (https://www.techpowerup.com/review/intel-core-i9-12900k-alder-lake-12th-gen/20.html https://www.techpowerup.com/review/intel-core-i9-12900k-alde...). Yeah, I know the components aren't matched... but that's what you get hunting for data. 25W isn't nothing, but it's not gigantic either. It's hard to get precise figures about the embodied energy of consumer electronics, but we have a range of 2000-6000MJ (https://energyeducation.ca/encyclopedia/Embodied_energy https://energyeducation.ca/encyclopedia/Embodied_energy). At 2000MJ, the payoff time for 25W of idle saving is quite large - on the order of 900 days. That's a lot of time to play with to try to design a hardware refresh approach to minimize total energy expenditure over your lifespan. And obviously from a $$$ standpoint, even taking some of the higher energy costs (~50cents/kWH), the cost delta over an entire year is pretty modest - $110 over a year. Once again, not insignificant, but given the discount factors involved (you can possibly get a $400 dollar used machine instead of a $1000 new), still quite possibly worth while. Obviously all of these are modified by what your actual system loading is, what your hardware refresh cycles look like, and local cost of electricity.