SFF Haswell build critique

[mfenn]

-I never really agreed with the x% perf for y% cost method. In abstract sure it makes sense, but in reality it's $20 for an extra 200mhz that may not be very noticeable but still save me time. Given that I get paid over 20/hr, the question to me is how long it'll take to recover a full hour to show some ROI. Even if I don't notice the difference, it still adds up.

I don't spend a lot of time waiting for the machine so it's not cut and dry, but that does seem like a better metric than straight price/perf. Sound reasonable?

Price/performance with a minimum level of performance is a good measure for general home and hobbyist use. For actual "work" work (i.e. getting paid), ROI is certainly a better metric. Note, this is the first time you've mentioned that it is for actual work work.

Note that you have to be able to quantify the R in ROI. Saying "it'll probably help...maybe" doesn't help. Saying, "it'll make X task that I perform Y times per week take N hours instead of M hours, and my time is worth Z dollars per hour" makes it really easy to decide if a given upgrade is worth it or not.