What did I already try? (... As far as I remember ...) Try 0A xserver standard; xf86-video-ati standard without ZaphodHead Result: Cloned displays - not useful Try 0B xserver standard; xf86-video-ati standard with ZaphodHead Result: Cloned displays - not useful [I've seen some misdetection in the logs.] Try 0C xserver mod (*); xf86-video-ati standard with ZaphodHead - Xinerama Result: Two seperate displays, but the VGA screen is always the primary one (and gets #:0.0) Config is robust against Screen number variations (in Display Section) Try 0D xserver mod (*); xf86-video-ati mod (**) with ZaphodHead + Xinerama Result: Two seperate displays, works as expected :-) Problems with painting-speed. Screen numbers in Display Section have to match exactly ZaphodHead identifiers. (-> kernel order has to be matched.) Try 0E xserver mod (*); xf86-video-ati mod (**) with ZaphodHead - Xinerama Result: Two seperate displays, works as expected :-) Problem: Split up in DISPLAY= :0.0 / :0.1 (I don't want that) Screen numbers in Display Section have to match exactly ZaphodHead identifiers. (-> kernel order has to be matched.) Try 0F xserver standard xf86-video-ati mod (*-) without ZaphodHead + Xinerama Result: Need to define 3 displays (as HDMI out is #1); HDMI as "dummy" (Problems with painting-speed?) Try 0G xserver mod (*); xf86-video-ati mod (**) without ZaphodHead + Xinerama Result: Two seperate displays, arrangable, but weired and sensible config (the exact order of mentioning the screens in ServerLayout matters) Problems with painting-speed. The try numbers are not chronological. mod (*) xserver: simple return value from array in xf86Bus (see Bug description) mod (**) xf86-video-ati: use the "Screen" number in Device section to choose the kernel device (pretty old style) (Bug description referred to this mod.) mod (*-) xf86-video-ati: use some other number from some struct available to choose the kernel device (this was a mistake by me, but made a first It-works!-Experience)