I can see your point too, and yes is is a vary valid one as well; the industry also needs some momentum/sales to get money for R&D to in turn create even more efficient machines.

I have no numbers on what the effects would be if more people would buy used hardware but my general feeling is that too little is being reused/repaired today in comparison to how much is produced/bought new. Still, this was not my main point after all - hardware reuse is good yes, but my main point is that it is probably OK to use old(er) hardware for quite some time when counting energy and Joules, since even a 30% less efficient machine can contribute to WCG and the research here - because when compared to how much more energy such a PC uses in relation to how much energy can be saved by reducing energy consumption from other sources (heat/AC, petrol, turn off lights, usage of water, baths, shower frequency etc) it is actually not that bad. Of course, if you are planning on buying 10 PCs, then that might be sort of a bigger deal perhaps :)

I concur that using efficient architectures is a good thing to aim at.
My oldest PC is based on an Intel Penryn (45nm) i7 870. 4 cores/8 threads. It might seem old, but it returns about 20-30 MCM WUs per day (each WU currently takes 5-6 hours to compute) and I think that is pretty good.

Actually, going for core/thread count may be more important than pure architectural efficiency in most cases. Let's say I can afford to buy a 4 core/8 thread CPU and platform of an older architecture, but only a 4 core/4 thread one if I will buy it new. Since BOINC and WCG research greatly benefits from parallelism I would place my bets in the older 8 thread machine.

Anyway, this subject is very tricky and all-in-all I believe that WCG research will benefit simply by getting as many machines /that fit the minimum requirements for each project/ added to the pool. ;)
At least at this moment in time...

Yes, for us power crunchers, modern CPUs usually are worthwhile when counting electricity costs in and not only net energy used. I agree.

But if adding $ into the mix, and assuming that money is not infinite :) then one may ask:

"If I have $1000 dollars to spend on WCG crunching, should I buy a new system with capabilities to run 8 threads most efficiently as available today, or should I buy two systems for that money, crunching a total of 16 threads but a bit less efficiently but with the same effectiveness (getting WUs crunched validly)?"

Newest hardware may be most efficient, but also the most expensive usually as well. So with 2 systems @ 16 threads you may get less efficiency (and a higher electricity bill) but you also still crunch may more WUs per unit of calendar time than that single new one PC - even though the purchase cost is the same for both system alternatives.

Basically I don't think that (normal or power user) people should refrain from joining the grid or adding an older PC into the mix just because it is not "top notch". In the end every WU needs to be crunched and for every WU an extra added "slow" system crunches it will mean it will not need to be crunched by someone else.

Also, I find that electricity (and other forms of energy) is wasted on much worse things today that the fact that someone adds a 486:er to the grid is almost a blessing in terms of energy being put to good use! :D

Adding slow systems may or may not be beneficial. I'm not sure how boinc projects work so...

1. If there are a fixed number of work units, lets say 1 million, it is much better to only use efficient computers. If we rule out old systems, maybe it will take twice as much time to complete all WUs, but it will use 1/6 as much electricity.

2. If a project runs for a fixed time and the most WUs have to be crunched during that time, adding lots of slow systems is beneficial.

I guess the truth lies between these two theories?

In the end, everyone will have to decide for themselves which rigs are worth running and which aren't.

At month 62 you are 5 years out. By that time there will be a new latest and greatest.
Cheers