The number of citations has nothing to do with the total benefit from an article. For example people (and by people I mean scientists) always cite the referee article, not the arxiv one. Or they add the arxiv link for convenience, but if possible they prefer to cite the more prestigious one. Why? Because this adds prestige to their own articles. It's a whole mafia. How, then you could possibly compare the over-all value of the two ways of publishing, when they have a completely different purpose in scientific society?

Not to mention the fact that scientific journals are very selective to what they publish (and quite requiring) and thus, the articles published there are much more valued than the free ones, simply because they had passed the selection phase.

But their "value" aside, when I make a research and I hit a subscription-only article, for me, it's like hitting a wall. Because my university doesn't have the money to secure me a subscription for all the journals I need. And who's suffering from that? Me and my research. Is that the way it should be?

I think it's high time to change the idea of paid-articles. If it's printed edition-ok, I will pay for it. But the online editions can and should earn from ads. Not from readers or authors.