Both the 680 and 580 should get a speed boost in 1.11 over 1.10.
There is one anomaly posted i'm trying to find out the cause of, but in general this should be the case.
It's really important to compare apples with apples here and make sure the different results are using EXACTLY the same benchmark scene (there are a couple floating around) with EXACTLY the same settings.
I have some results of a titan with the benchmark found here, http://render.otoy.com/forum/viewtopic.php?f=9&t=7144 on 1.10, changing to PT and no alpha shadows only:
3x Titan = 16.69Ms/sec = 5.56Ms/sec per card
On our 4x690 machine the exact same scene and settings gets:
4x 690GTX = 23.60Ms/s = 5.90Ms/sec per card.
On my machine here (just one 580GTX enabled @ stock 772 core 1002 memory)
1x 580GTX = 3.52Ms/sec
So the titan is almost as fast as a single 690 card. And a good 58% faster than a GTX 580.
I only have figures for 1.11 on the 690's and 580:
4x690 = 28.57Ms/sec = 7.14Ms/sec per card (21% faster than 1.10).
1x580 = 3.90Ms/sec (11% faster than 1.10).
So considering the titan is a kepler card i'm assuming it will get about the same speedup as a 690 on 1.11 so i would expect somewhere about 6.73Ms/sec per card for a titan on 1.11.
Thanks
Chris.