Re: Power consumption GTS 250 versus GTX 240/260/275 with the Linux drivers?
@hl_: the problem is that memory clock changes need to happen during the vertical blank period to avoid underflow and similar problems in the display engine. Since there's no guarantee that the display timings of both displays match and are synchronized (i.e. that the vertical blanks happen at the same time), memory clock transitions were disabled for multi-head configurations.
I believe that this has recently been relaxed, however, allowing p-state transitions in cases in which the two heads are raster-locked. I'm not sure off-hand what all of the relevant requirements are, but if I recall correctly, the use of identical display devices yields a good chance that the driver will do the right thing automatically.