Ok, I am now pretty damn sure that the problem is a bad implementation of non-UEFI bioses on the 7000-series. I've contacted MSI to try to get a UEFI bios for my card to make sure.
Are we certain that 7000-cards don't get the white-screen-bug (WSB from now on) on real Mac Pros? Someone must have tested this, but I can't find anything about it.
If they do, then this might not be something we have to solve ourselves, it might just be an incomplete driver which will be fixed at 10.8.3 release.
If they don't, then that means that the bug is because of something hackintosh-specific, like the bootloader or one of the (few) kexts we all have.
Partly solved by using EFI injection instead of GraphicsEnabler, thus giving me full QE with HD3000 and a passive 7950 which works fine when I reboot to windows.
Not perfect, but a damn sight better than when I started
OK, after many (many!) hours of back and forth, this is now isolated to the AMD 7950. If i remove it, I get sleep and the HD3000 gets QE through GraphicsEnabler. If the 7950 is in the computer, GraphicsEnabler doesn't work and without it I get no sleep, and no QE.