Two Very Different nVidia Cards, One as Graphics, The bigger one as Physics. How?

This is similar to a previous question: Two Different nVidia Cards, One as Graphics, One as Physics - Possible?

But I need addition information.

I have an unusually heterogeneous system with a GTX-580 and a new Quadro M6000 (obviously SLI is not applicable). What's also unusual is that I don't care much about graphics performance. My primary application is compute-bound, so my old GTX-580 is more than adequate as the dedicated display card. I want the much more powerful M6000 to be dedicated to the primary compute application. Obviously this is not a gaming application.

I previously stumbled on a configuration that made this work, but somehow lost it, so I know it's possible. So I'm more interested in answers about how to do this then answers presuming that it is not possible. Sadly, I did not back up the good configuration.

Currently, I can't seem to get NVidia Control Panel to recognize the existence of the M6000. The GTX-580 is the only detected hardware, and that card works fine.

It won't allow be to re-install the M6000 because it doesn't detect the M6000.

Now, I could just remove the GTX-580, since it doesn't add much compared to the M6000. I tried that, and things work okay, but when I run my compute application, it competes with the display. In particular I'm running keels that take up to 500-1000 ms to run (pushing the edge of the watchdog time limit). Ideally I'd like to disable the watchdog for the M6000 so that I am run longer keels, but obviously this is a problem if I'm using the same card for display. That's basically my motivation for keeping the GTX-580 as a display card.

Recent Questions...