By the same token, if we don't support SOME older systems then we lose TON of possible contributions and risk p$$$ing off people who have more to offer than a single cutting edge machine. Most of the grid, I would suggest, is made up of machines over 2 years old. My particular situation allows me to run a limited number of older machines but prevents me from buying new hardware.
But we do have to draw a line somewhere, or code the science apps so that they recognise and use the best available instructions on whatever architecture they're on, which I strongly expect to be a substantial undertaking to say the least and could cause unpredictable work results.
I suppose the project could drop support for anything older than say 5 years without it being too suicidal ... but without pulling up a profile of what processors are currently running that could still be risky.

Note the minimum system requirements. Anything that will run XP is supported.

http://www.worldcommunitygrid.org/help/viewTopic.do?shortName=minimumreq#251

Windows XP min requirements: Pentium 233-megahertz (MHz) processor
(That would be a Pentium 2 circa 1997)

http://support.microsoft.com/kb/314865



I expect a lot of electricity is wasted due to inefficient code.

Thanks. You have satisfied my curiosity. HCC is, I believe, the least demanding project currently being run. I was debating trying to run HCC on a an old P5 just to see how it would do, but if it took a week on Pentium Pro it would probably take 8-9 days on the P5. I think that would be the minimum system you could use as I think the 486 maxed out at 64 Meg of memory, so it would probably not meet the memory requirements for the system even though the processor would do the job, chugging along for a fortnight at least.

Cheers