[ChuckMcM]

This has been one of the most amazingly non-intuitive things for me over the years. That operating a computer that is older can be so much more inefficient than operating one that is newer, that it isn't worth it. Literally the best thing you can do is crush the chips and mine the result for gold, silver, and the few rare earths that are left over. That at least puts a finite limit on the energy consumed over it, and the value returned from it.

[DanBC]

Not sending it to places like the e-waste dump in Agbogbloshie, Accra, Ghana is a benefit. I'm not sure how to weight each thing - the increased CO2 emmissions are bad; not poisoning poor people who do things like burn the PVC insulation off cables to get the copper is a good thing.

[ChuckMcM]

Certainly agree with that sentiment! We are getting to the point where it makes sense to have 'disassembly' robots, this paper (http://www.sciencedirect.com/science/article/pii/S0967066108...) looked at that (it wasn't the focus per se but an example). I can clearly see that 'anti-factories' where raw materials are recovered and unrecoverable materials made inert will be a thing.

[e12e]

Still, the market isn't going to solve this on it's own. Other peoples children are going to be cheaper than robots for the foreseeable future. In fact, poisoning the local population could be considered a minor issue compared to contaminating the water tables. Sick people die, but a contaminated water table keeps killing people for generations.

I think strong local regulations, as well as an emphasis of recycling locally is the best short-term solution. It does expose the actual cost of recycling (safely) -- but also cuts energy to shipping, and can be used as a stimuli for a more re-cycle friendly value chain (eg: make sellers of electronics take the bill for recycling, and demand local, safe recycling) -- which in turn puts pressure on suppliers to deliver items that are cheaper to recycle safely.

[ChuckMcM]

We may have sufficient regulations that the market can push it over the top. After all if you can create a facility that converts ewaste into resalable raw materials you can both sell the service to municipalities who are required by law to dispose of their ewaste safely, and the raw materials to manufacturers as a recycled product. That combination might get you into an internal rate of return to make it worth while.

[e12e]

The real challenge is the globalization (and accompanying fragmentation of regulation) of electronics production. One might recycle heavy metals locally, but you'd have to get it back into the production pipe-line -- which generally means shipping it to China. And being able to compete on price with heavy metals from various more-or less horribly run mines around the world. One obvious alternative is to produce electronic components locally (again). But realistically, doing that cheaply enough (and well enough) is going to be a challenge.

In this sense I think Tesla is a very interesting company (even if I don't think much of the cars themselves, from an eco-perspective -- cars is a horrible means of transportation, even electric ones).

[switchbak]

If we had halfway-usable power management back in the day, it would have a lot more sense (for occasional use) as it would be idle/sleeping most of the day.

But we most certainly didn't, especially in Linux - and most people ended up having to leave them on. Heck, even the super pimped-out Linux machine I made in 2012 still has occasional issues with suspend/resume (I could probably go on a week long patch fest to try to fix it, but no thanks - I bought a macbook instead :)

[brudgers]

Though it clearly isn't a server farm, there would appear to be reasonable use cases in the consumer and small business markets.

Lowest first cost is a significant form of capital efficiency [as opposed to energy]. There are tradeoffs either way, and for a single moderate use computer reaching the break-even point via reduced energy footprint could easily be several years. The raw cost of running the most power hungry P4 [115w] is $151 per 24/7 year @ full load @ $0.15/kwH.

In addition, having the machine pre-loaded with Linux also offers time efficiency versus re-provisioning a new Windows machine with Ubuntu or shopping for components, etc.

None of which to say it's a good deal per FLOP or MIP or anything else. But it's not necessarily bad either.