[latch]

Aside from an SSD, a 2006 computer (with the introduction of the first intel Core2Duo) with 8 gigs of ram is really the last time things got noticeably faster for the average programmer.

Sure, game programmers and the like 'need' to upgrade more often, but for almost everyone else, it's just been a waste of money in relation to what you get.

Laptops are a different story, though I think that's quickly coming about (or it already has with i5/i7s)

I have noticed a tendency for some of the bulkier tools to get slower over time, making it seem like you need a better computer, when in reality, we just need better tool makers (VS 2010 was a notable step back that made you feel like it was time to upgrade).

[barrkel]

I frequently run things like:

    find -iname '*.foo' | xargs -P 8 -n 1 process-foo

The more cores, the better. I feel I'm really restricted from targeting the CPUs of tomorrow by lack of parallelism; if you're writing a piece of software you expect to possibly last 20+ years, you really shouldn't over-assume today's limited CPU count, in particular.

Work per clock cycle is improving at a not too shabby rate either. When I upgraded from a Q6600 to an i7 920, build times reduced by about 60%; Handbrake transcode times improved by even more.