Do you mean 7...dxc4 8 Bxc4 b6 instead of 8...Bb6?
The reason 7...dxc4 is not played is it allows white's bishop to come to c4 in one move from f1. Usually Black waits for the bishop to develop to e2 or d3 and then captures on c4 to disturb the bishop. It is so early in the game that it is difficult to show a concrete refutation. Surely you can accept that 7...b6 is a better move and that this is why GMs play it.
The d5 pawn doesnt block the bishop--the bishop protects the d5 pawn.
Even if White pushes on e4 and d5, The Black light squared bishop and f6 Knight eye those squares.The Black l.s.bishop also eyes the castled King(if he castles Kingside, which most players do in the the QG).
Even if white succeeds in advancing e4 by placing his Bishop on d3, it will be a target of attack for Black. ???
In the QGD lines for black I doubt there is a better choice than the Tartakower variation. It has been endorsed by at least 3 world champions : Spassky, Kasparov, Karpov and often playing both sides of it ! Other great players may have also used the line, I havent checked.
Another way to reply 1.d4 d5 2.c4 is by declining the gambit with 2...e6.
The Tartakower Variation runs
1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Bg5 Be7 5.Nf3 h6 6.Bh4 O-O 7.e3 b6 8.cxd5 Nxd5 9.Nxd5 exd5 10.Bxe7 Qxe7 .
The d5 pawn blocks the BLACK Queen's bishop. I think that 7...dxc4 would be better: 8.Bxc4 Bb6. The Bishop now has long diagonal. I wonder why this move is not in any Opening Book.
Please tell me what you think of the variation.
I found 4 games with 7...dxc4, and white won all after Bxc4.
We should not always go according to what Grandmasters say. If it was for an opponent rated the same as you, the one who opts for this line as Black may find a very interesting variation and can win the game.
We should always be trying out the variations by ourselves rather than only listening to Grandmasters.
BLACK with his bishop on the open diagonal can dominate the game if he knows how to play well! 
If you want to play this plan, it's probably better to just play a QGA. In (similar variations of ) the QGA, white has to play e2-e3 quickly so his dark squared bishop stays at home.
Instead, you've chosen to play the QGD, the main strategy of which is to keep the center and defend it, hoping to slowly neutralize white's activity.
To suddenly give up the center and to switch to something QGA-like while white already has an active dark squared bishop is inconsistent and can't really be as good as either the QGA or the normal QGD plans.
Of course, none of this is likely to matter at amateur level, say below 2300.
In the Tartakower lines of the QGD white has several good moves at move 8 and according to my database 8 cxd5 isnt the best one.
ouachita, why is that the case? I have always wondered about this. Isn't it bad to limit the variety of openings available to a developing player?
As I've said on numerous other forums, a player should master the e4-e5 and d4-d5 openings, and only then explore the more modern openings and variations.
Sure, but we're discussing d4-d5 here, right?
As I've said on numerous other forums, a player should master the e4-e5 and d4-d5 openings, and only then explore the more modern openings and variations.
Sure, but we're discussing d4-d5 here, right?
Dont confuse him ! 
It's okay to try your own ideas. By playing them, you can assess if they're good, average or bad. 
It's okay to try your own ideas. By playing them, you can assess if they're good, average or bad.
I've already tried the line many a times and it works quite well.
My problem is why does that line does not appear in any opening book. There must surely be a refutation to it. What is it?
My problem is why does that line does not appear in any opening book. There must surely be a refutation to it. What is it?
Plenty reasons have been mentioned why it's an inferior move. Not all slightly worse moves have a concete refutation!
7. ... dxc4 is not a blunder, but it does appear to be rather bad from the stats (68% white in my database)
On reason for this is that black has committed the bishop to e7, but taken on c4 before white has played Bd3. Waiting until afterwards is a tempo better. If black follows up with c5, relatively speaking white may gain a tempo by dxc5. These are both reasons for this not being a great time to play a delayed queen's gambit accepted. But my guess (and that's all it is) is that black can still reach equality with care, just with a little more difficulty than in popular lines.
All important points have already been said, just to summarize:
(1) 7...dxc4 is not a mistake. It is not much played because there are better lines. Books only mention important, much played lines.
(2) by deferring dxc4, and playing it after white plays his bishop, black actually gains a move. Thus white keeps his bishop development for the last, and black tries to play dxc4 after white's bishop-move (Bd3 or Be2).
(3) 7...dxc4 8.Bxc4 (c5 9.0-0) is present in databases (under D53: Queen's Gambit Declined, Modern, 4.Bg5 Be7 variation), not in books - only very few play it.
It's okay to try your own ideas. By playing them, you can assess if they're good, average or bad.
I've already tried the line many a times and it works quite well.
My problem is why does that line does not appear in any opening book. There must surely be a refutation to it. What is it?
Have a look at games with 7.e3 b6 8.Be2 dxc4 and 7.e3 Nbd7 8.Bd3 dxc4 and compare them with the move-order you've chosen, then make up your mind
Another way to reply 1.d4 d5 2.c4 is by declining the gambit with 2...e6.
The Tartakower Variation runs
1.d4 d5 2.c4 e6 3.Nc3 Nf6 4.Bg5 Be7 5.Nf3 h6 6.Bh4 O-O 7.e3 b6 8.cxd5 Nxd5 9.Nxd5 exd5 10.Bxe7 Qxe7 .
The d5 pawn blocks the BLACK Queen's bishop. I think that 7...dxc4 would be better: 8.Bxc4 Bb6. The Bishop now has long diagonal. I wonder why this move is not in any Opening Book.
Please tell me what you think of the variation.