Najdorf 6.Be2 e5 7.Nf3 Be7 8.Bg5: why Nf3?

I have read, can't remember where, that in the following case (see diagram below), 8.Bg5 is the point of having played 7.Nf3 instead of 7.Nb3. First the diagram, then the question:

So what's the point?

I have read that 8.Bg5 after 7.Nb3 (the usual move) is a subtle difference. See diagram below just to be sure. So what's the "subtle" difference?

So what's the subtle difference?

I came up with the following so far: Bg5 appears to attempt to undermine the (Boleslavsky) hole d5. But what has 7.Nf3 to do with the "Bg5 method"?

Be3 would be the normal square for that bishop to move to in the Opocensky (6. Be2). 

This will allow f4 to be played. If the knight is on f3, f4 can't be played anyway, so the bishop may as well go to g5. 

Also with Bg5 you would typically be castling queenside, so maybe you don't want to have another knight target on the queenside if you're castling that side. 

This line with Nf3 is a very positional one with the main goal of fighting for the weak d5-square. After ...e5, there is no great square for the knight available to participate in this strategic battle. That's why you often see maneuvers like Nf3-d2-f1-e3 later in the middlegame. However, on f3 it is at least not blocking any of White's pieces. White can still go Bc4 and if necessary Bb3 if attacked.

in the second diagram, does ...Nxe4 work after Bg5? 

edit: nevermind, it doesn't.

As I understand it, the main theme is that both parties always fight for d5-square anyway. But when white plays the Nf3-variation, so to speak in this case, Bg5 seems to be the usual follow-up. Such is not the case when white choses to plant the knight on b3 instead like they usually do. In this case Bg5 is not the main course anymore, though still entirely possible off course.

So what is this subtle difference?

I told you the difference, did you not read what I said? 

Be3 is the normal move when you are playing Nb3 and then you can move f4 without black's queen or bishop attacking that diagonal. There is also the fact that you will be usually castling queenside with Bg5, so you want less of a target with the knight at that side. White pretty much has to move a4 or at least a3 in the Opocensky variation.   

There's your clear and fairly obvious answer, just because you keep saying it's "subtle" doesn't make it so. 

Since you don't remember where you read this, maybe you are mixing it up with 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 Nc6 -- the Classical Sicilian. Here 6.Be2 e5 7.Nf3 Be7 (7...h6) 8.Bg5 is (or was, in the days of yore) considered a little better for white, but 7.Nb3 Be7 8.Bg5 is considered equal. In the Najdorf I don't recall anybody making this distinction between 7.Nf3 and 7.Nb3. In the Classical Sicilian after 6.Be2 e5 7.Nf3 h6, black's recommended antidote is a correctly timed retrograde ...Nc6-b8!, with the idea ...Nb8-d7-f6, challenging the d5 square. Since in the Najdorf the black knight was not "misplaced" on c6, black saves two tempi on the ...Nb8-d7-f6 plan.

Really it's all equal, you can find strong players on either side of the board who are not too impressed by the supposed theoretical subtleties.

Quit speculating. I looked it up. It's a quote in a PGN with Anish Giri's course on the Najdorf. An amazing course if only I would have more time. But my question here was meant as a follow on this quote:

" This is similar in nature to 6.Be2  e5  7.Nf3  Be7  8.Bg5  with barely relevant and very subtle differences. This system is a lot older than it's 7.Nf3 colleague, but not any more promising for White.  Most likely we will get the same knight + dark-squared bishop vs knight + light-squared bishop position, where the knight on b3 is as far away from the d5 square as it would be on f3. "

Thanks to your answers I can place it now in a  much better context. Thank you all.