PCIe 3.0 (and for that matter, even 2.0) was mainly introduced to provide adequate bandwidth for dual graphics card setups. The performance difference between PCIe 1.x and 2.0 depends on a variety of factors such as your GPU, chipset, the specific application/game (for example, in some games, PCIe 2.0 may make a bigger difference on an NVIDIA card than a Radeon, and vice versa), etc.
Just for information, PCIe 2.0, taking into account bandwidth overheads, delivers about 4GT/s in and out. The fastest PCIe 2.0 implementation known to date is the nForce 780 series which runs part of the bus at 4.5 GT/s (of course, practical performance difference is NIL because the additional nForce 200 chip adds latency too. However, multi card performance does show a marginal improvement on these nForce boards).
There is some evidence that the current generation cards do saturate the PCIe 1.1 bus. A good read on the matter can be found here:
The impact of PCI Express on GPUs (page 1: Introduction) - BeHardware
It shows that on a GTX 570 or an HD 6970, the performance difference between PCIe 2.0 x16 and x8 (which is the same as PCIe 1.x) is about 8%.
Thus, the practical conclusion is that if you have anything less than a 560 Ti/HD 6950, PCIe 2.0 will not make a difference for a single card configuration. However, it makes quite the difference for dual GPU setups and higher end graphics cards.
