AMD

[John P. Myers]

My Titan, which runs at 1071 MHz posts about 5.2 TFLOPS. I have had it installed in a PCIe x1 2.0 slot since i got it. Works perfectly fine

For gaming, again it wouldn't saturate even a 2.0 x16 slot. At least not on it's own. As i mentioned several posts back, there is no more Xfire bridge on these GPUs. All info is passed between the cards via the PCIe bus. I imagine running quad Xfire on a PCIe 2.0 bus would definitely cause some lag because of this, but 3.0 would have room left to spare.

The *effective* bandwidth of PCIe 3.0 is 985 MB/s per lane = 15.39 GB/s @ x16. To use it all, a single GPU running a 1080p monitor would have to achieve 1,992 frames per second.

[DrPop]

. . .The *effective* bandwidth of PCIe 3.0 is 985 MB/s per lane = 15.39 GB/s @ x16. To use it all, a single GPU running a 1080p monitor would have to achieve 1,992 frames per second.

Well, good to know I so desperately "NEEDED" those 40 lanes of PCIe 3.0 on these Socket 2011 mobos. I might actually use all that bandwidth in 20 years!

[John P. Myers]

Well, good to know I so desperately "NEEDED" those 40 lanes of PCIe 3.0 on these Socket 2011 mobos. I might actually use all that bandwidth in 20 years!

LOL well there is an advantage to having 3.0 over 2.0, which is time. Even if you only come up with 100 bytes of data that needs to be passed between the GPU and CPU, 3.0 can pass that data back and forth faster than 2.0, which speeds things up a little