Kind of funny sometimes to see what happens. By mistake my comment became market as "negative" (some temporary error on the site made the field for comments only showing the last letter of everything I wrote among other things), as hence most probably think I'm in a bad mode here trashing the article.
No big deal, but the round-up was good. My only complaint is, and it's a miner, that since most hastily read through articles the last diagram will make a standing impression because of these rebates that aren't exactly what the test did prove.
Otherwise one of the important parts about RAM is how unpredictable it is. As the article shows it has to be tested. In a sense I think the diagram at page 6 isn't totally fair, because of you pay premium for getting something rated at PC-6400, equal to a just as compatible PC-8000, the percent overclock is kind of irrelevant. A factor difficult to predict is how stable these modules are after 24/7 for a sustainable time.
Anyway, I'm sorry if my last post came across as if I trashed Crashman's work.