To my understanding (which, as usual, could very well be wrong, particularly with newer stuff!) PCIE 2.0 cannot be reduces to, say, PCIE 1.0. Instead What can happen is an expansion card might get fewer PCI-E Lanes then is optimal.
Now, that said, I've also never heard of on-board graphics chipsets affecting performance of installed dedicated cards in this way. Typically when you install a dedicated graphics card, if the motherboard as a separate, on-board chipset, it disables altogether specifically to avoid the problem of the on-board using up PCI-E lanes and causing the dedicated card to throttle to say 8x lanes.
This reduction can happen when you populate specific slots. Some motherboards have two x16 slots, but if both are populated, then they both run at 8x. I'd guess that internally they are just "one slot" that gets divvied up between two physical slots. I experienced this first hand without even realizing until, probably a year later, I started up GPU-Z to look at some detail of the card and noticed it listed it was running at 8x speed. The cause was that I had installed an expansion card into the other x16 slot over a year before.
I partly agree with Computer_Commando, as I think spending money on these sorts of upgrades is sort of throwing money down a hole; money which realistically you could instead put towards a much more modern system, which can be one of AMDs many more modern CPUs. Even using the lowest end components on the market today (eg value-oriented components) would likely outperform the Athlon II x4, which you can then repurpose for other tasks.
