Originally Posted by netllama
If your goal is to allow each head to be used independently, then you'd like need to run one X server for each, and not use Xinerama to control multiple heads and cards under a single X server. I don't believe that you can run more than one instance of gdm per X server.
How are you starting up each of the X servers?
Effectively, I have just checked Xinerama to see if my cards were working. I use gdm to start every X server independently. There is a [servers] place on the gdm.conf, there I declare each X server to be started.
The "****INVALID MEM ALLOCATION****" warning in your XFree86.1.log is a bug that I believe is resolved in a driver after 1.0-7174. Can you please provide a bug report that includes your use of the 1.0-7667 driver?
Sorry I am using 1.0-7174, not the 1.0-7667 (my fault). I think 1.0-7174 is the latests driver.
I think i've found what the problem is, but not how to solve it: The nvidia driver can only be used by the default graphics card and the other cards cannot use the nvidia driver: If I configure the BIOS to use the AGP card, then executing 'X -probeonly -layout X0' (where X0 is the layout for the AGP card) works fine, but 'X -probeonly -layout X1' (X1 for the first PCI card) fails. Again, if I change the default video card on the BIOS to PCI, then X1 layout works with the nvidia driver, but X0 doesn't.
I've found the following message on dmesg when AGP works and PCI doesn't:
NVRM: rm_init_adapter(1) failed
As attach, the full dmesg with that error and the nvidia_bug_report.log.