This isn't really something worthy of a actual forum post, so I figured I throw it up here on my blogs.

I ran a bunch of tests on my current setup, see specs below:

• Phenom II x4 940 @ 3.514GHz (16.5 x 213)
• RAM: 4GB Corsair XMS DDR2 800 @ 852 5-5-5-18
• GPU: 2x GTX 280 SLI (651/1406/1164)
• MB: ASRock K10N780SLIX3-WiFi)
• Video Driver: 260.99
• Resolution: 1680 x 1050

I then ran the benchmarks of several games 7x, then removed the highest scoring and lowest scoring and averaged out the remaining 5. My first set of data was with my GPUs sitting right next to each other in a 16x/16x configuration.

FarCry 2 (5 runs, Ranch small.): DX 10, Very High, AA 4x, Bloom:

• Avg FPS: 72.53
• Min FPS: 50.08
• Max FPS: 106.19

Metro 2033 (5 runs): DX10, Very High, AA 4x, AF 16x, PhysX On

• Avg FPS: 40.99
• Min FPS: 9.29
• Max FPS: 216.58

Crysis (5 runs): DX10, Ultra High settings, AA 4x

• Avg FPS: 37.23
• Min FPS: 19.52
• Max FPS: 48.54

I've finally grown tired of my top GPU (since they hug each other due to the MB design) running so hot and having the fan run at 75%+ all the time, I changed the PIC-E bridge chip from having the PCI-E slots run at 16x/16x/1x to 16x/8x/8x. I then moved the bottom GPU to the third slot. Then went through and reinstalled the video driver and ran the bench marks again for Metro 2033 with my GPUs now running 16x/8x. Here are the results:

Metro 2033 (5 runs): DX10, Very High, AA 4x, AF 16x, PhysX On (16x/8x PCI-E setup)

• Avg FPS: 40.48
• Min FPS: 8.62
• Max FPS: 208.44

Some of the runs with this setup ran just as fast or faster as some of the runs with my cards in 16x/16x. In the end, the results are within a 2% margin of difference, which means there is no difference in performance.

So, in the end, 16x/16x vs 16x/8x - they pretty much run the same as far as I can tell and as I'm concerned.