I attended a no-name college and went to a top-tier university for my phd (think mit/stanford/cmu/berkeley). I also taught at a different top-tier university. Everything below is specific to CS.

I think that top-tier universities are different in the top 20% and bottom 20% of students. The top 20% because you won't find those types of people at no-name. The bottom 20% because they're at least not completely incompetent (unlike the bottom 20% at no-name).

Here's how I break it down:

Top 20% of students at top universities: They really are that good. Also, they are the ones who benefit most from the intense curriculum at top universities. So they start off really good and actually do get a force multiplier effect from the comparatively very rigorous course-work. Most of these students end up on rapid trajectories in industry or with NSF fellowships.

Average students at top universities: Most of these students would be in the top 10% of the class at a no-name college. Still quite good, but nothing special.

Bottom 20% of students at top tier universities: Really nothing special. Comparable to the average student at no-name. Contrast with the bottom 20% at no-name, many of whom couldn't fizzbuzz on their first attempt in our senior interview prep course.

So, top tier places are really characterized by their top 20% and bottom 20%. The top 20% really are quite amazing to work with. The bottom 20% are at least not incompetent.

One last note:

> Or maybe the best of these students go on to work at companies that pay 250k/yr and I just don't meet them.

Most of my students at top-tier had offers in the $150K - $250K range. So if your sample of top-tier university CS students is sampled exclusively from non-entry-level engineers who did not receive offers in the $250K range, you're almost certainly sampling from the bottom 20% of top-tier graduates.