Well, the article that I found states that PCI-e bandwith has little actual impact on performance of even the latest video card. While stepping down from PCI-e 2.0 at 16x, they only found a 2% slowdown at 8x, a 5% reduction at 4x and a mere 25% drop while the bus was running at 1/16th of it's maximum speed. I'm interested in results, not theory.
The only thing I assume is that PCI-e 2.0 8x will serve the video card just as well as a PCI-e 1.0 at 16x. Which is probably true considering it's written on Wikipedia, the link you posted.
Which brings us to the conclusion: if he is to spend money wisely, he is not going to buy a new motherboard just so he can have a PCI-e 2.0. It'll do nothing for him. He won't care if the highway in his computer is crowded, that isn't his problem.
Lanes has nothing to do with Bandwidth.
x16 Lanes @ 250mbs
vs
x8 Lanes @ 500mbs
Does not mean 500mbs/2 = 250.
"PCIe 1.x is often quoted to support a data rate of 250 MB/s in each direction, per lane."
vs
PCIe 2.x being capable of 500mb/s in each direction, per lane.
So PCIe 1.x means x16 * 250 = 4000MB/s (or 4GB/s)
PCIe 2.x means x16 * 500 = 8000MB/s (or 8GB/s)
So according to the article:
8000MB/s - 2% (@x8) = 7840 MB/s (which is not correct, just that the card in question is not utilizing all of the available lanes in the tests they ran)
Which is still far greater than 4000MB/s
The lanes limit the amount of interrupt requests made to/from the device, not hinder the amount of bits per lane it is capable of.
The article you linked to was all 2.0 configurations, there were no variances of port type vs card type.
The article was to show the differences of performance at different resolutions in different "modes" x1 to x16 lanes on a 2.x setup.
Just because you have 16 lanes available, does not mean they are all in use all the time, only those needed (requested/IRQed) are on.
So if lane 1 - 4 are requested for data they will send 250MB/s per lane, they do not open another lane to provide extra bandwidth for the support, they are reserved for other requests.
Would be like asking a printer to print a document, then sending another job to the same printer, and expecting it to do it at the same time or send the job to printer 2 if printer 1 is busy, it doesn't happen.So again a PCIe 2.x motherboard will effectively double the data rate capable of being transfered to and from the card, increasing overall performance, vs upgrading to a newer card, the bottle neck isn't at the amount of lanes, it's at the amount of data being passed by each individual lane.
In a perfect world, all lanes would be on all the time, which if the article would have been clearer would be the case.
But on the downside would have had dramatic effects on the tests for each closing of a lane.
