For me, a long-time admissions person (and now-and-then director of grad studies) in mathematics, in a large state school in the U.S.: I do not care very much about GPA. Reason: there are too many reasons, unrelated to mathematics, why someone might get good or bad grades. Even "chronic good grades" can be a somewhat negative indicator, potentially, if it means that the student is obedient and conformist above all else.
Undergrad programs (not so much "universities", and not so much "higher-ranked") with richer programs obviously tend to have more resources to offer students. Having inaccessible big shots does not guarantee a rich program, for example. Nor does everyone take advantage of the potential in their environment. And so on.
And, specifically, "identical in every other regard" would have to mean identical letters from the same faculty members. Well, this never happens, for many reasons, especially to students in different universities.
So, although I understand the intent of the apparent premise, on one hand if I play along with it then the answer is something like "they're still essentially identical", but/and, more realistically, this does not happen. That is, the premise is essentially never met in practice.
To amplify: being in different universities usually means that one has access to different faculty, and has different people in one's cohort. In my appraisal of grad school applicants, this distinction is the origin of my comparisons of applicants... which is manifest specifically in letters of recommendation (and applicants' statements of purpose, which reflect the students' perceptions of their experiences).
EDIT: Also, now that I review things, the notion of "easier curriculum" is quite a bit of a phantom, too. It can only sensibly refer to "minimum requirements", which are rarely an attractive profile to have, in the best of cases. It's not that mathematics (e.g.) itself is easier in some locales than others. Students can avoid engagement or not, especially with the internet, in nearly any physical location. The question of the local minimal requirements is hardly relevant to anything (apart from the unfortunately common misinterpretation of "minimum" as "desirable"...)
One can be passive and disinterested at an elite place, or proactive and engaged at a less-elite place... and in both cases it's still impossible to predict "GPA" since the latter is badly correlated with actual mathematical activity, etc.