Are you saying you have a pair of GTX 680's to test this first hand? I could well believe it, given the cards have PCI-e 3 on them and because the driver has to to a lot of the schedulers work in software which would increase PCI-e traffic (over what Fermi required).

i have the platform to test them on when i get my second GTX 680
-- it's just a matter of time when i can do a thorough comparison of PCIe 2.0 vs. 3.0 using GTX 680 SLI

So for the moment you are just speculating that 8x+8x 2.0 PCI-e won't be enough, or has some other reviewer actually tested this?

Somebody has tested this recently and there wasn't too much difference

THG?

No game can stress PCIE 3.0,u have to use some compute intensive tasks.

Yeah THG from memory lane

Not PCIe 3.0, 8x ... PCIe 2.0 is constrained a bit

ANd i have a big surprise on the temps on the overclock
- decent. Maybe 4-8C more degrees with more than +20% overclock with the same cooler.
(under Max load with OCCT 4.2.0; it's tickling 70C)



And it is 80F inside the test room .. and i will not even attempt to find a max OC - trying for something different and rather specific

U call that hot? come over here and enjoy true summer

A lot of people have tested PCI-E bandwidth and in general it only made a 5-10% difference. With a GTX 680 the difference will probably be 3-5% going from PCI-E 2.0 to PCI-E 3.0 at 16X.

This is Spring

About ten miles from where i live (closer to sea level; i am in the mountains at 3,000 feet elevation), the temperature hits 120sF in the Summer

And i got lucky - my evaporate cooler is back in operation. i just turned it on .. and it is working at about 70% efficiency .. the temps were climbing in the room to 83F when the cooling went on
(it maxed out inside at 83F yesterday; 100F outside - i guess the house is "preheated" this morning - got a nice cool breeze going .. but with all the molds and crap from last year still in the old filters)