I'm not entirely in agreement with you on that, Chris, though I do agree that having the latest-and-greatest is neither a necessary nor sufficient condition for success in pursuing a CS degree.
My main reason for disagreeing with you is that much (I'd even be willing to say most) of the learning will take place outside of assignments through independently exploring and playing with what you've read and tried to apply in coursework. Having immediate access to tools that can speed up that process can not only help one gain new understanding or realize a misunderstanding more quickly, but it can actually encourage a student to begin such explorations in the first place.
At the risk of generalizing my own experiences too much, when I had no computer at home, having one of those "hmm, I wonder if..." moments either meant a long walk to campus or waiting until the next day to try it out. Once I had a computer at home, I spent much more time asking those kind of questions and trying to answer them. I learned much more than I otherwise would have as a result.
Similarly, if I'd had a 4-core machine when taking the parallels classes, I could have spent more time investigating some ideas I was curious about than waiting in the queue.
And keep in mind, I'm not just talking about the required testing for assignments here. Your advice that the OP look into the specifics of his program are right on, and he should definitely do that before making any kind of decision. I'm talking about testing ideas of his own, on his own time. Again, playing with what he's learned. Going above what is just in coursework, which no matter the rigor of the program, can never teach you everything. The best students I knew were ones that spent an hour on the assignment, and spent 8 on trying their own thing.
It's like that in every field, really, but in CS, some of the questions you might ask are about the very tool you're using. I mean, in math I never wondered, "hmm, would this still happen if I switched from 0.5 to 0.7 lead?" But it's entirely expected that you might wonder if you get the same payoff going from two to four cores that you did going from one to two. And when the answer is "no" (and it will be, more often than you'd want), having the ability to find out why on your own is very helpful.
So in short, I agree that the demands of his specific program may not require anything more than a G3 (which is also all that I had), but he should also keep in mind that the written demands aren't all he should be considering.