Ten minutes of investigation yields…
To try to be really careful and controlling about this experiment, I closed Chrome, then re-opened it, and closed the window (zero tabs). Then I openend tabs one at a time, recording measurements from Activity Monitor (after they settled down) in Fathom. MacOS 10.6.8, Chrome 19.0.1084.54. MacBook Pro, not new.
Results do not match the Mythbusters “testing” post referred to by Shawn here. Of course, the original test opened some “top 40” sites list for their 40 tabs. But this gives some interesting insight into what may be going on.
The number of threads was praactically constant at 26 or 27.
A Best-Case Scenario 
I assume “real” means actual RAM and “virtual” means virtual memory. These are cool results. Now I want to know where the knee is across different browsers or if in fact they have similar curves. Again I’m more interested in performance which is what the memory usage is hinting at. While a multiprocessor wouldn’t change this memory picture, it could drastically change performance but that would depend on how the browser was written – threading. Also does this differ on different platforms? In Windows there is a wonderful tool called the performance monitor that lets you sample and graph all kinds of interesting measurements of your running computer. It could serve as the basis for a statistics lesson by itself. You can set up long running captures or just watch real time. Easy to use and comes with all the Windows O/S’s.
Stepping back, maybe you need to find investigations that excite your audience, like this one does me. What statistical measurements matter to high school students? Percentage of students with with certain designer labels. Likelihood of showing up dressed the same as someone else …
Yeah, I’m taken with these too, partly because it’s data you can get quickly by just sitting there and doing a little planning, and partly because it’s about computer performance, which is intensely relevant to my everyday life.
For kids in the class, this might not be exciting. The Census stuff usually seems to be engaging, maybe because of the first reason: Fathom makes it so easy to get the data right away; you can ask and answer a question (and generate new questions) very quickly. I see students also engage with downloaded data when it has a connection to social justice (i.e., it pushes an outrage button) or to people like them or their parents; or if it involves cool or gross phenomena. Death and disease seem to be good no matter what.