A game using DX11 & PhysX at the same time would actually not be able to test either API's performance.
GTX 260 in Batman AA:
Without PhysX (Control):
105-42=63. 63/105 = 0.6 = 60% drop in performance.
In other words, the performance (FPS) is 40% of what it was without PhysX.
HD 5870 in STALKER: Call of Pripyat
DX11:
DX10.1:
DX10:
DX9:
Link:
http://www.pcgameshardware.com/aid,700136/Stalker-Call-of-Pripyat-Benchmarks-from-DirectX-9-to-DirectX-11/Practice/
Average changes:
DX11 / DX9:
Day - 44/108 = .41 = 41%
DX11 performs worse, 59% drop
Night - 38/119 = .32 = 32%
DX11 performs worse, 68% drop
Rain - 41/128 = .32 = 32%
DX11 performs worse, 68% drop
SunShafts - 29/54 = .54 = 54%
DX11 performs worse, 46% drop
DX11/DX10:
Day - 44/38 = 1.16 = 116%
DX11 performs better, 16% gain
Night - 38/33 = 1.15 = 115%
DX11 performs better, 15% gain
Rain - 41/35 = 1.17 = 117%
DX11 performs better, 17% gain
SunShafts - 29/27 = 1.07 = 107%
DX11 performs better, 7% gain
DX11/DX10.1:
Day - 44/37 = 1.19 = 119%
DX11 performs better, 19% gain
Night - 38/33 = 1.15 = 115%
DX11 performs better, 15% gain
Rain - 41/34 = 1.21 = 121%
DX11 performs better, 21% gain
SunShafts - 29/26 = 1.12 = 112%
DX11 performs better, 12% gain
The differences between DirectX 9 and DirectX 10/11 as well as DirectX 10 and DirectX 11 are quite obvious.
DX11 only performs worse than DX9 (around 60.25% worse), while being better than DX10 (13.75%), and DX10.1 (16.75%)