I have a second graphics card in the system that is not SLI'd in use as a dedicated physx card. it does not have a display attached and does not have any SLI bridge stuff hooked up. Just in a second PCIE slot. When the Displaylink software is installed, the second card is not used for physx, only the primary card. When the displaylink software is removed, the second card is used properly for physx. In both cases the 240 is selected for physx offload in Nvidia display panel. Heres how to tell on Win 7:
1. get the GPU observer desktop gadget:
http://blog.orbmu2k.de/
2. select the 240 and turn on "show GPU load". You can run two of these, one for each GPU.
3. download the "nvidia physx particle fluid demo" from here (you can turn off the rest of the download from the power packs)
http://www.nvidia.com/content/graphi...s/download.asp
4. Run this demo with and without the displaylink software and watch the load on the second physics GPU (240 in this case).
5. I also saw this behavior on the "Dark void physx benchmark" and the "fluidmark" 1.1.1 benchmark:
http://www.ozone3d.net/benchmarks/physx-fluidmark/
Config:
Windows 7 32 bit Ultimate
Nvidia gtx275 graphics with driver 196.75 (and attached physx version)
Nvidia gtx240 graphics card with driver 196.75
2 22" monitors attached in windows multimonitor mode to the 275
Core2 duo 9650 system with 3.2 gigs (4 actually)
Displaylink software version 5.3 beta 2
MIMO 710
Note: Here is a card that should exhibit the problem since it is a very similar setup to my discrete solution:
http://www.evga.com/products/moreInf...eries%20Family